[SCM] Libav/FFmpeg packaging annotated tag, v0.5, created. v0.5

siretart at users.alioth.debian.org siretart at users.alioth.debian.org
Wed Aug 14 16:20:06 UTC 2013


The annotated tag, v0.5 has been created
        at  110c4b2a211fbf6779edb6361931680bf4e7b429 (tag)
   tagging  f8429ed58cefea1669bc127cf2b1905b4893e3f2 (commit)
 tagged by  Diego Biurrun
        on  Mon Mar 9 23:06:42 2009 +0100

- Shortlog ------------------------------------------------------------
Tag 0.5 Release

(no author) (1):
      New repository initialized by cvs2svn.

Adam Thayer (1):
      XviD Support patch by (Adam Thayer <krevnik at comcast dot net>)

Agent Smith (1):
      dont put flies in the stdout soup patch by (Agent Smith <asmith at wgz dot com>)

Alan Curry (18):
      vYCoeffsBank and vCCoeffsBank are allocated and initialized using incorrect
      AltiVec operations need to have memory aligned on 16-byte boundaries.
      altivec_yuv2packedX() ignores the requested output format and unconditionally
      Correct RGB vs. BGR confusion, the macros vec_mstrgb24 and vec_mstbgr24 each
      Fix big-endian color permutation problems.
      Move the v{Y,C}CoeffsBank vectors into the SwsContext, filling them in just
      make swscale-example compile patch by (Alan Curry <pacman  TheWorld com>)
      ng the fix for vCCoeffsBank patch by (Alan Curry <pacman TheWorld com>)
      Fix vec_clip for gcc's that don't handle (vector signed short){16} according
      don't call altivec_yuv2packedX() with a dstFormat that it doesn't support;
      Just a comment update, replacing a FIXME to reflect the new expectation of
      Fix rgb32tobgr16, rgb32to15, and rgb32tobgr15. All had the same problem that
      Support all 4 of ARGB/RGBA/BGRA/ABGR unscaled conversions, with more
      Reorganize vector constants to work around gcc 4.1 bug:
      Fix some bugs handling input that is not aligned at 16 bytes or has
      Fix typo introduced in the memalign->av_malloc conversion (there is no
      Fix compile error due to extra "FF" at beginning of file in the FF{MIN,MAX}
      Fix buffer handling for stereo mace files, fixes some random sig11

Alban Bedel (2):
      Added reverting support for -pp and -npp options
      printf to mp_msg

Albert Astals Cid (2):
      free moov and cmov when error occurs, fix memleak, patch by Albert Astals Cid, aastals at tv-wan dot es
      Fix compilation with -DDEBUG, patch by Albert Astals Cid, aastals tv-wan es.

Albert Lee (1):
      Simplify init preprocessor statements.

Alessandro Sappia (2):
      Adding support for libdc1394 v.2, patch by Alessandro Sappia a dot sappia at ngi dot it
      Enabling higher frame rates for dc1394, patch by Alessandro Sappia a dot sappia at ngi dot it

Alex Beregszaszi (317):
      multiple init bugfix (patch by Alex Beregszaszi <alex at naxine.org>)
      patch by Alex Beregszaszi <alex at naxine.org>
      added support for various app headers, and writin FFmpeg comment
      clamp intra matrix to 8bit for mjpeg (workaround for qscale>=25)
      BSD/OS support by Steven M Schultz
      applied 64bit patch from Ulrich Hecht <uli at suse dot de>
      pre-yvu9toyv12 converter, only grayscale Y-plane coping :)
      moved codebooks to svq1_cb.h
      added aspected_width/height for EXTENDED_PAR
      added missing palette8tobgr pairs (hope it's correct)
      fixed palette8to{rgb24,bgr24,rgb32,bgr32}
      aspect (ext. par too) support for h263 and mpeg4 (inc. build becouse of new vars)
      complete aspect support
      moved dct init out from mpv_common_init to dct_common_init (for less-uglier way for dv)
      added mjpeg
      reworked decode_frame marker searching, fixes many non-working samples
      10l
      minor speedup noticed by arpi
      2nd 10l (noticed by arpi)
      FOURCC removed, using ff_get_fourcc instead (should be big-endian safe), workarounded a restart interval bug (Spectralfan.mov) (rst support should be rewritten and moved from decode_sos)
      NetBSD fix by Bernd Ernesti <mplayer at lists.veego.de>
      fixed 10l
      fixed some bugs in app parser - some jfif and adobe jpgs fixed
      mjpegb support (need more samples)
      10l fix
      disabled aspect code due to it is extra buggy
      simplified
      10l
      killed an mp_msg occurance
      removed the last mp_msg :)
      1l
      non-x86 support
      update
      cs_test cleanup and fix compilation if gui enabled
      unused
      a bottle of coke (=0.5l)
      10l
      some warning fixes
      pal vs ntsc fix by stoyan k <stoyan at i-space.org>
      1000l
      warning fixes
      lossless jpeg support
      support for armv5tel (XScale used in iPAQs)
      tv standard selection support for dv1394 and grab (v4l)
      moved the tables into header files (and applied the 'static' patch). Nick: why do you like mergeing tables and code into one file, so making it unusable big?
      10l found by Curi Fabio Eduardo (SFL) <curif at TELEFONICA.COM.AR>
      unwanted second definition of CHECKED_ALLOCZ
      uniformization (now it uses the same trace functions as h264, defined in common.h)
      moved frame_count to wmadeccontext
      Optimized simple idct for arm by Frederic 'dilb' Boulay <dilb at handhelds.org>. Currently licensed under the GPLv2, but the author allowed to license it under the LGPL, feel free to change
      warning fix
      warning fixes
      get_bits -> skip_bits
      CLJR decoding support
      410p -> 411p
      remove cs_test and swscale-example too
      using av_abort instead exit
      initial sh4 support
      bump version numbers to 0.4.8
      initial nut muxer and demuxer (demuxer is not fail safe)
      some fixes
      move packetheader after startcodes
      nut and vcr2
      fix fabrice's broken get_bi and some minor changes in draft
      fixed time base and sample rate handling
      fixed codec specific data headers
      uyvy output support in swscaler
      theora decoding support (only keyframes for now, because by theora the frame isn't flipped so the motion vectors are getting screwed up)
      using av_abort
      theora comments skipping
      Sunplus JPEG codec (SP5X) support
      add some missing codecs
      removed the obsolete and unused parameters of init_put_bits
      10l
      init_put_bits changed
      make it compatible with older make (debian stable - make 3.79.1)
      10l
      100l
      10l
      1l
      initial support for zygovideo
      finally working with old theora bitstream (flipped image), the only sample I have is decoded successfully (theora.ogg)
      theora alpha3 support (with maintaining backward compatibility, maybe we should remove all backward compatibility codes after final theora release?)
      more correct header parsing
      cosmetics
      using LIBAVFORMAT_IDENT
      msb pts support
      FF_DEBUG_PICT_INFO and CODEC_FLAG_GRAY support
      correct C init tables
      freebsd fix for mpeg12.c (moving INT64_MAX to common.h)
      not needed anymore
      changed includes
      get_bit_count -> put_bits_count
      get_bit_count -> put_bits_count
      avoid negative array indices
      flac
      make gcc happy under hurd
      support for wider range of codecs
      unknown format chars
      conversion to av_log
      unknown format char
      fix compilation after michaels anti-time change
      support for -timestamp
      100l, and another one because it boils out with ICE under gcc 3.2.3
      use the AVV macro from gcc_fixes.h instead ifdefs
      disable hadamard for gcc below 3.3 - better fix awaited
      avoid macro conflicts
      ranlib cleanup by Dan Christiansen
      bigendian fix
      10l, non-x86 fix
      make the tables static
      skip_utf8, unused yet
      some debug
      warning fix
      altivec workaround
      unused
      some fixes
      simplify the init
      fix altivec.h inclusion (vector keyword in structure)
      new copyrights
      new opensource lossy/lossless audio codec based on speech compression techniques (actually based on bonk)
      unofficial ids for sonic for using in nut (fixme)
      snow and sonic
      ffmpeg short conversion got fixed somewhere, no need to workaround it - fixes amplitude bug
      more decorrelation types
      try to select w/h from tables
      new signed golomb routines
      forgot to commit this
      put the warning to the right place to avoid crashes
      lot of bigendian fixes
      greyscale decoding (option to skip u,v planes) support
      Decoding to bgr15/16 from 16bit input. Decoding of 24bit input added, not yet finished, but at least the picture can be recognized
      cws2fws, utility to convert compressed flash files to uncompressed ones, which can be than played back with ffplay. note: this doesn't gets compiled automatically
      report keyframes
      correction
      preliminary vc9 bitstream decoder, committing to make syncing and team-work on it easier
      preliminary vc9 bitstream decoder, committing to make syncing and team-work on it easier
      exported ff_mpa_synth_filter for upcoming qdm2 decoder
      10l
      correct pcm in flv handling
      macromedia flavour adpcm decoding (used in flv and swf)
      support adpcm audio, also some more verbose messages
      proper handling of samplesize
      pseudo codec tag for adpcm_swf
      changed the mailinglist pointers to the new place (mplayerhq)
      unused define
      some nut.hu hype
      1l
      some missing additions
      print 'not working' message
      parenthesing to avoid compiler errors in the future - compared asm outputs, behaviour didnt changed
      100l - this has only worked on big endian
      tell the user if compressed swf found
      10l
      10l
      FreeBSD ipv6 multicast patch from ports tree
      use av_log in debug mode
      initial MorphOS support
      lame bit_rate calculation
      removed unused CONFIG_AUDIO_NONSHORT
      sanity check whether dimensions are non-null
      set_se_golomb can only write 16bits, add a note about this (ok, maybe it's brain dead using it with more than 16bits, but..)
      fix feed read_header, avoid using put_tag in write_header, to be consistent with read_header, also some minor cosmetics
      fix ffserver, it is working for me now again, a more elegant fix is needed though
      10l, dont make bandwidth statistics negative when feed stops being sent (it was decreased every time feed sending was aborted)
      Ingenient MJPEG support, more at http://www.artificis.hu/files/texts/ingenient.txt
      support transcoding adpcm
      remove -e debug=off, unsupported my some wget versions (1.9), and doesnt makes any difference
      strict -1 is mandatory for jpegs because the colorspace thingie, this should be handled in a better way though
      Support Launch when BindAddress is not INADDR_ANY or localhost. Also fix a possible buffer overflow
      Make Launch more appropriate with support for full paths for the ffmpeg binary
      make ffserver regressions work again
      debug the launch commandline and new connections
      10l to michael not updating this after changing crc to output 0x prefixes (crc.c rev 1.10)
      support for unlimited feed size
      Support for streaming: dont write indexes and dont signal HAS_INDEX in header. Also set filesize to max in this case.
      updated regression reference
      mode_code_vlc_table size mismatch
      fixed long standing off-by-one bug (fixes playback on ppc)
      Fixing theora _again_. DONT BOTHER BREAKING THIS AGAIN. Added support for Theora 3.4 and stream created by Elphel cameras are decodable.
      10l, left debugging on
      better version reporting
      10l: now it works really, with 1k outbut buffers it failed more times, with 65k it looks safe with all samples i had
      tta decoder, tested with 44khz mono and stereo (for the latter use ffplay or change MAX_AUDIO_FRAME_SIZE)
      tta demuxer, also usable for moving tta audio data into an other container
      10l, fails to compile when debug enabled
      tta and theora improvements
      true audio
      improve video4linux2 detection (without sys/time.h it was complaining about struct timeval
      increase audio buffer to 1second of 48khz float audio
      10l, allocate bitbuffer with regard to the padding size
      explain extradata and pass avctx to av_log's
      check for CODEC_ID_NONE instead 0, tobe compatible with movenc.c and possible CODEC_ID_NONE value change
      make those nasty variables static, also fix a html bug
      added none/8bit/24bit audio formats and defaulting to s16 for now (once the system works like pix_fmt, the defaulting should be changed to SAMPLE_FMT_NONE)
      support correct ptses on other than version 3.2.0 aswell
      print a hint when trying V4L2 on V4L device
      report to user if mmf with midi found (and not supported)
      Make public domain. Hope Diego dont mind it, he had two cosmetics commits :)
      correct signaling of palette change
      Fix possible overflows. Found by Steven Johnson
      changed rgba32_to routines to support both alpha and non-alpha formats
      changed rgba32_to routines to support both alpha and non-alpha formats (see r7066 too)
      BGR555 has never been working as alpha supporting format. Remove the false setting.
      Remove alpha channel from RGB555
      10l memory corruption bug found by Steven Johnson
      support byte_run=0 case in DELTA_FLI (this case means only skip pixels)
      omg, major typo (fails if card supports rgb24)
      rename BE/LE_8/16/32 to AV_RL/B_8/16/32
      add AV_WB/WL for lswriting, similar to AV_RB/RL (also increment version)
      spit an error message in case of invalid chunk
      display line number when printing error
      use av_random instead random/srandom
      remove gettime_ms and use av_gettime instead of it
      change close() to closesocket() on sockets (compatible with libavformat)
      change write/read to send/recv on socket operations
      codec_tag settable via VideoTag, and transmit codec_tag in ffm
      if feeder is disconnected, clear the feed registration, thus a new feeder can connect
      change PF_INET to AF_INET to be consistent in the whole project. PF_INET is deprecated, while AF_INET is referred by the POSIX standards
      fix the server regression test (it produced for me identical results in more runs, really).
      1l, clean up feed1.ffm after finishing the server test
      warn that under windows, bash <3.1 may fail on configure
      synchronize to latest ffm changes
      give myself the glorious maintainership of ffserver and mingw port
      disconnect client if trailer is sent
      return error if av_write_header failed (it fails on some containers when codec is not supported)
      rescale pts/dts for output format
      store internal timestamps in int64_t instead long
      redirect http get requests of / to index.html (note, the http_parse_request may worth a rewrite)
      use AV_RB32, swf_mp3_info may be replaced with mpa_decoder_header
      add little endian 24bit read/write
      add little endian 24bit read/write
      disconnect new feeder if feed is already being received
      handle http error codes
      10l, audio_fifo is a static array
      typo: unkown->unknown
      simplification
      workaround for broken flvtoolized files
      move preroll_time from static variable into definition, might be a good idea moving this into the context and making user-settable
      print codec parameters for audio too
      remove dead code
      use av_strdup
      fix x-octet-stream mime type
      remove possibly broken wince cruft
      cosmetics: remove my email address
      ac3 parser requires tables from ac3tab.h, which are compiled into ac3.c
      compile mjpeg parser if requested
      compile rawvideo decoder on request
      remove useless code, file removal is handled elsewhere
      fixed InputFormat option, which was never working
      use av_strdup instead malloc/strcpy
      simplificate
      check ip port range from config
      support for hostnames instead ip addresses in ffserver conf (using resolve_host now)
      make the last globals static
      use resolve_host in acl handlign too, simplificates the code
      remove unused rtsp_callback
      cosmetics in resolve_host
      move resolve_host from tcp.c to os_support.c as it is used widely
      Use correct RTSP_DEMUXER checks instead of NETWORK.
      properly check for CONFIG_RTSP_DEMUXER
      remove CONFIG_NETWORK, av_read_play/pause are exported anyway
      properly compile code for REDIR_DEMUXER
      Don't compile RTSP_DEMUXER if not requested
      Make protocols configure selectable and also cleanup the CONFIG_NETWORK
      10l, rtp_muxer needs config_network
      use network.h instead network includes
      use ff_neterrno() and FF_NETERROR() for networking error handling
      implement ff_socket_nonblock and use it in networking code
      Use correct CONFIG_RTP_MUXER ifdef
      check for channels<=0 and print a reasonable error message
      cosmetics: fix indentation
      use DECLARE_ALIGNED from mem.h (which is included through common.h)
      link to ws2_32.dll only if networking is enabled
      initial mingw networking support
      Use usleep instead of sleep, some MinGW versions lack sleep.
      fix FD_SETSIZE handling in our poll() emulation
      r9095 invalidly removed the check for non-winsock
      fix compilation with assert
      10l to Benoit, poll.h is already included below, and this broke compilation on mingw
      Remove broken assert. ebml_read_element_id can handle level_up set to NULL.
      remove uneeded include as assert is never used
      h264_rtp_extra_data is not only used for debug purposes, asserts are using it also
      match doxygen comment with other parts of the file
      10l to Mans, assert is included later with proper DEBUG/NDEBUG ifdefs
      .memleak fix by Ronald Bultje. OK'd by Luca.
      use av_q2d as in the rest of the file
      set codec_type on init
      removed unused variable ti
      indentation fix
      cosmetics: indent correctly
      cosmetics: be sure all lines fit into 79 characters
      cosmetic: remove extra parenthesis
      cosmetic: indentation
      cosmetic: extra space
      cosmetic: simplify some constructs
      cosmetic: remove a huge number of useless braces
      check for closesocket
      use proper closesocket defines
      debug: tell how much bits are skipped with unknown nal code
      copy get_unary from vc1
      use get_unary from bitstream.h
      use get_unary from bitstream.h
      use get_bits1(..) instead get_bits(.., 1)
      use reget_buffer and remove internal copying of buffer - codec works again
      fix compilation with TRACE enabled
      use skip_bits where appropriate
      use skip_bits where appropriate
      Make sine_window static. Acked by Benjamin Larsson
      Change the "report sample" message from AV_LOG_DEBUG to AV_LOG_ERROR.
      Support 44.1kHz audio. Acked by Benjamin
      add myself as TTA guy
      Remove editor junk from the end.
      Prefer ADPM_SWF instead NONE if no MP3 encoding is compiled in.
      Honor the 8bit extension (now 32bit instaed 24bit) of the pts field introduced in FLV specification v9.
      Call prot->url_close only if it is present.
      Remove empty pipe_close

Alex Converse (47):
      Only read some of TNS bitstream data in the case that the TNS filter order is
      Correct order of parsing for pulse scalefactor band and offset to match the
      Apply pulses to unscaled coefficients rather than scaled.
      Validate pulse position and error out if an invalid position is encountered.
      Cast the return value of get_bits() to signed integer such that the subtraction
      Corrections to channel coupling code to attain conformance for appropriate
      Correct documentation of ch_select array.
      Rename variables to clarify the channel coupling element and corresponding
      Fix perceptual noise shaping scaling including an oversight of the offset for
      AAC: Frequency domain prediction and hence Main profile support
      AAC: Remove unnecessary fabsf()
      AAC: Unroll code within decode_spectrum_and_dequant(). Gives 12% net speed gain.
      AAC: Use a sign LUT rather than 1-2*get_bits1()
      AAC: Don't write an illegal sampling_index in a PCE to the MPEG4AudioConfig
      AAC: Fix AAC prediction when used in conjunction with the CPE
      Factorise enum of AC3 error types to be usable by AAC in the ADTS patch that
      Factorise id3v2 header parsing from mp3.c to be shared
      Fix probing of files with ID3v2 tags. Discussed at
      Fix a read past end of buffer crash in the mp3 probe
      Clean up FFT related comments.
      Mark ff_fft_init with av_cold.
      Mark ff_fft_end av_cold.
      Add the rdft family of transforms (fft/ifft of an all real sequence) to dsputil.
      Use the new RDFT code in the QDM2 decoder.
      VP3: Prevent stack corruption from an unset custom coding method.
      Add av_cold attributes to *_init and *_end functions.
      Make the MP3 probe weakly claim all files with ID3v2 tags to fix regressions on
      VP3: Do not ignore error from read_huffman_tree().
      Increase the number of packets required to identify as MPEG PES to keep from
      Remove bogus coupling_point assignment for non-CCE as this value is not checked
      Add support for sample rate index 12, 7350 Hz
      Give an error message when returning due to an unallocated channel element
      ADTS Muxer: Refuse to write illegal ADTS files by checking validity of header
      Correctly map ADTS profile_ObjectType to MPEG-4 AOT
      Report the illegal audio object type, not it offset by 1.
      Make set coup.gain[c][0] aand only coup.gain[c][0] is set for a independently switched CCE.
      Re-indent after last commit.
      Some buffer checks for indeo3
      Reindent after last commit
      Add Divio MPEG-4 FOURCC (M4CC)
      AAC in ASF does not need parsing.
      Re-indent after last commit
      Close the AAC parse context, discovered by Netgem.
      Use const to better allow the compiler to optimize AAC independent coupling.
      Refactor AAC unsigned band type decoding for a modest speedup.
      Re-indent after last commit.
      AAC: Make the CPE function more consistent with the functions for other channel elements.

Alexander Chemeris (1):
      fix crash when decoding a truncated H264 slice

Alexander E. Patrakov (3):
      Increase buffer size to 16384 patch by Alexander E. Patrakov" patrakov gmail
      Correct scaling factor in DCA decoder synthesis.
      fir_32bands_perfect[] contents were mixed, restore original order

Alexander Strange (57):
      Fix compilation with --disable-encoders.
      h264 optimization: common case hl_decode_mb patch by (Alexander Strange astrange ithinksw com)
      The configure test for ebx fails because it is run before the -mdynamic-no-pic
      Don't check the return value of decode_cabac_residual since it always
      Statements like a = b = c = d = e; store from right-to-left, so if
      Fix missing chroma in version 3 STR.
      Fix clicking sound in xa-adpcm.
      fix crash on non-AltiVec powered machines: MPV_common_init_altivec doesn't check mm_flags
      Fix configure to properly build without PIC by default on Darwin
      Check for and set HAVE_MMX2, it is used in several places.
      Check for sys/mman.h and set appropriate definitions.
      Split MANGLE macro into LOCAL_MANGLE AND EXTERN_PREFIX parts
      Fix 64 bit shared library compilation with MMX2 by properly using PIC mangling.
      Set mm_flags in dct-test to fix floating math.
      get rid of the useless DCT_ERROR() macro
      check mm_flags for each DCT and skips the ones that can't be run
      Altivec and SSE2 require 16-byte alignment, so make sure they get it
      Increase alignment for DCT block arrays from 8 to 16
      Add a header file to declare Xvid IDCT functions.
      Add a new IDCT permutation, used in xvid_sse2 and possibly future similar IDCTs.
      Add a new xvid-style IDCT using SSE2.
      Test idct_xvid_sse2() in dct-test.
      Xvid IDCTs are not under CONFIG_GPL, so don't require it when using them.
      Use check_cflags for -mdynamic-no-pic so third-party compilers will work on Darwin.
      Store mb_xy in H264Context and only calculate it once per MB.
      Use h->mb_xy directly in decode_cabac_residual().
      Remove newly unused variables.
      Set and use h->mb_xy in SVQ3 too.
      Previous commit missed an initialization - this fixes FATE's test sample.
      Define mm_flags/support to be 0 on architectures where they don't exist.
      Use a local variable instead of mm_flags in dct-test.
      Don't set mm_flags in fft-test, since it doesn't do anything anyway.
      Remove unused PutBitContext.
      Export qscale so that postprocessing works.
      Remove commented-out code.
      Remove duplicate emms.
      Remove pointless AVFrame* casting.
      Use MANGLE() instead of memory operands to read globals.
      Realign newlines.
      Don't test for H264 encoding unless an encoder exists.
      Don't generate two copies of hl_decode_mb() under ENABLE_SMALL.
      Clone decode_cabac_residual() for DC residuals, to get rid of a bunch of branches in the main loop.
      Reindent.
      Simplify loop condition for readability.
      Reorder the if/else for residual cats to match gcc's branch prediction.
      Use get_cabac_bypass_sign instead of an if/else.
      Cosmetics: split a long line.
      Make the function prototype visible to comply with C99 inline.
      Remove const vector casts.
      Cosmetics: reindent.
      Print the address of the AVCodecContext in av_log().
      Move the size of the H264Context delayed_pic array into a #define.
      Add a new -debug option for tracing calls to the default get/release_buffer functions.
      Allocate and clear FF_INPUT_BUFFER_PADDING_SIZE bytes at the end of mkv
      Clear FF_INPUT_BUFFER_PADDING_SIZE bytes at the end of NALs in rbsp_buffer.
      H264: Change left_block_options to uint8_t from int, saving a little space.
      Put dispatch_tab in the rodata section for macho64.

Alexander Strasser (16):
      Fix name clash when building in mingw environment.
      libavutil: Utility code from libavcodec moved to a separate library.
      Make out of source dir build work again.
      Reorganize lists to have one item on a line.
      Fix building shared version of lavf on some systems.
      Make ogg `muxer' expect new vorbis extradata format.
      Fixed format specifier in STOP_TIMER macro.
      Mention libav versioning in CVS policy section with a short explanation.
      Print libavutil version too.
      Added support for vorbis files containing floor type 0.
      10l (malloc check with the wrong pointer)
      Cosmetic: Fix indentation.
      Correct the relation between floors of type 0 and block sizes.
      Some Make variants complain if the LIB variable does not exist.
      Don't use pow/powf functions where we just need integer arithmetic.
      Build also ffserver (if configured) when doing `make test' .

Alexander Wichers (2):
      8HZ -> 8KHZ, cosmetics patch by Alexander Wichers development at wichersdot nu
      Nellymoser 8KHZ flv muxing fix, patch by Alexander Wichers development at wichersdot nu

Alexis Ballier (4):
      QT RLE encoder
      stricter constraints of asm() blocks
      Add missing sys/select.h #include, fixes compilation on FreeBSD 7.0.
      Remove useless inline qualifier, fixes linking with gcc 4.3.

Allan Hsu (4):
      Fix wrong flags for S16LE and S16BE audio in FLV files.
      Defines various common FLV format values between the FLV muxer and demuxer
      add FLV decoder metadata parsing
      Patch for the FLV muxer to supply more complete metadata

Allan Sandfeld Jensen (2):
      add support for information in CDXA format
      Revert r6124, the change was just wrong, other formats than MPEG-PS can be

Allen Day (2):
      verbosity patch by (Allen Day <allenday at ucla dot edu>)
      printf macro patch by (Allen Day <allenday at ucla dot edu>)

Anders Grönberg (3):
      Matroska uses tables from isom.c, so add the object file to its dependencies.
      Add a few const qualifiers in appropriate places.
      Prevent shift overflow, patch by Anders Grönberg, galileo.m2 gmail com.

Andrea Gianarro (1):
      fix AC3 5.1 being decoded as mono

Andrea Gualano (1):
      Enable avpicture_deinterlace() for PIX_FMT_GRAY8.

Andreas Öman (67):
      10l patch by (Andreas Hess <jaska at gmx dot net>)
      fixing output_example to work with current CVS libavcodec/libavformat patch by ("Andreas U. Trottmann": andreas trottmann, werft22 com)
      allow ffmpeg to read mp3s beginning with partial frames
      fix probe, patch by Andreas Oman, andreas at olebyn dot nu
      avoid crashing when trying to parse invalid files, patch by Andreas Oman, andreas at olebyn dot nu
      cosmetic v1/v2 renaming
      id3v2 reader
      id3v2 writer
      allocate PPS and SPS dynamically
      filter_mb_fast() doesn't handle h->deblocking_filter==2 (skip slice edges), so delegate that case to filter_mb().
      h264: use 'simple' in border backup / xchg
      Factor out init_scan_tables(), patch by Andreas Öman %andreas A olebyn.nu%
      indentation
      Decouple bit context from h264 context in decode_ref_pic_marking()
      Do not reuse the rbsp de-escape buffer if both
      Don't swap back un-deblocked lines for intra prediction when
      Compute the dequant coefficient table ptr outside the loop,
      Precompute a chroma_qp table with index offset for each pps,
      Add support for streams with different chroma_qp_index_offset
      fix visual painting of MVs from h264: they are painted twice too long
      add YUV440P and YUVJ440P support
      use YUV440P in swscale
      bail out on unknown jpeg pixel format instead of silently decoding data incorrectly
      regression tests for yuv440p yuvj440p
      increase buffer size of LAME MP3 encoder
      fix cabac decoding with some Blue-Ray sources
      Add slice-based parallel H.264 decoding
      Let H.264 decoder skip deblocking across slices
      Indentation
      fix image corruption when with multi-threaded decoding.
      Simplify H.264 decode_cabac_mb_cbp_luma(), giving a ~0.5% speedup.
      factor out dequant table lookup outside loops, gives a 1-2% speed-up
      reindent
      remove unused code, spotted by Jeff Downs, heydowns a borg d com
      reindent
      add support for reading duration from Xing-tag in mp3 files
      cosmetical change of mp3_parse_xing() to prepare for upcoming VBRI tag support
      fix indent
      add support for reading duration from VBRI-tag in mp3 files
      make sure the mpeg audio header is valid before passing it to ff_mpegaudio_decode_header()
      remove unnecessary -1 from snprintf len
      Bump lavf minor due to addition of
      Make AC3 decoder honor avctx->request_channels
      Make DCA decoder honor avctx->request_channels in a more advisory way.
      Notify the input coder about the number of requested channels.
      SDL does not seem to support more than two channels.
      Add missing '\n' to log format text
      Dont crash if resampling is requested but no audio packets are ever processed.
      Correctly use output format stream index for bitstream filter when doing stream copy.
      Fix a bug causing the generated stream to be corrupt if the buffer
      Fix the following warnings:
      Compute AC3 frame CRC for stronger raw AC3 format probing.
      Audio channels and sample rate must be set in order to skip
      Deprecate avcodec_build(), it returns the same value as
      Supply context argument to av_log() in mpeg audio decoder.
      Decrease severity of 'invalid new backstep' log message to warning
      The comment about channel order for AAC 5.1 audio was not correct, fix it.
      When setting codec_id during codec probe we must also set codec_type.
      dca: Do float -> int16 interleaving in-place using s->dsp.float_to_int16_interleave()
      dca: Set data_size outside block loop.
      5,1l: samples_chanptr[] should be (const float *)
      Change multichannel API define prefix from "CHANNEL_" to "CH_".
      h264: Implement decoding of picture timing SEI message.
      Include "libavutil/common.h" where we use llrint() in case ffmpeg's
      Revert r16257:
      Avoid allocating MPADecodeContext on stack.
      Gopher protocol should depend on networking being enabled.

Andrew Savchenko (1):
      Only warn about "Invalid and inefficient vfw-avi packed B frames" once.

Andrew Voznytsa (1):
      sample rate fix

Andrew Wason (2):
      Copy pts for each raw encoded frame.
      Fix endless loop when opening corrupt FLV files (issue 699).

Andriy Rysin (4):
      audio only avi patch by (Andriy Rysin <arysin at bcsii dot net>)
      RTSP "OPTIONS" command patch by (Andriy Rysin <arysin at bcsii dot net>)
      mp3 codec autodetection patch by (Andriy Rysin <arysin at bcsii dot net>)
      prefers MP3 codec (rather than MP2) for output if

Andy Gocke (1):
      Return an error when realloc fails.

Andy Parkins (2):
      ffplay seeking because url_feof() not working patch by (Andy Parkins: andyparkins, gmail com)
      do not truncate pts/dts if value is AV_NOPTS_VALUE, patch by Andy Parkins, andrew parkins at 360visiontechnology com

Anssi Hannula (6):
      Interplay C93 demuxer and video decoder
      CRYO APC demuxer
      Set correct value for libdir in pkg-config files.
      In pkg-config files, libraries not exposed through the library, but just needed
      Define private libraries as 'Libs:' when only static libraries are built,
      When shared libraries are built, make the pkg-config requires Requires.private.

Antoine Cellerier (1):
      Give a hint about what is the expected the data type of command line options.

Anton Khirnov (5):
      add support for chapters definition in lavf
      demux chapters out of matroska
      matroskadec: silently discard some element ID that we don't care about
      matroskadec: read TrackTimecodeScale and set timebase accordingly
      matroskadec: replace matroska_queue_packet with a single call to dynarray_add

Anuradha Suraparaju (13):
      Dirac decoding/encoding support via libdirac.
      Add Dirac support through libschroedinger.
      Disable libdirac decoder if the faster libschroedinger decoder is available.
      Add a parameter for queue size.
      Add entry for Dirac codec.
      Parse frames to feed the decoder with individual parse units.
      Fix pts handling when encoding with libdirac.
      Fix pts handling when encoding with libschroedinger, closes issue 453.
      Dirac encapsulation in MPEG-TS
      Change Dirac in MPEG-TS encapsulation to conform with spec changes.
      Add support for creating Simple Profile (I-frame only, no arithmetic coding)
      Fix incorrectly constructed Dirac parse units that caused A/V sync loss.
      Fix crash when encoding using libschroedinger.

Arpi (58):
      - removed #include ../config.h, because it's included from common.h
      MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
      inlineing common case of get_bits() -> gives 2speedup. more optim coming soon...
      get_bits() specialization, gives 4\speedup
      CONFIG_MPGLIB no more needed
      new postprocess code by Michael Niedermayer (michaelni at gmx.at)
      fixed a bug in the horizontal default filter
      small changes for mplayer integration:
      small changes for mplayer integration:
      Cleanup:
      c++ fix
      software yv12->rgb scaler - separated from fsdga
      header for swscale.c
      exporting qscale data for postprocessing (for MPlayer)
      16bpp code for atmos
      bpp changed from bytespp to bitspp
      added (f)emms
      (C) fixed
      xinc scaled by 16 instead of 8
      PP_FUNNY_STRIDE disabled
      dct_unquantize_h263_mmx() by Michael Niedermayer <michaelni at gmx.at>
      postprocessing array increased to handle HDTV resolutions (1920x1152)
      uyvy->uv12 added
      new IDCT code by Michael Niedermayer (michaelni at gmx.at) - #define SIMPLE_IDCT to enable
      new IDCT code by Michael Niedermayer (michaelni at gmx.at) - #define SIMPLE_IDCT to enable
      unrestricted MC fixed - thanks to Michael Niedermayer for idea
      avcodec_find_encoder_by_name() patch by Alex
      informal msg
      mpeg4 encoder fix by Michael Niedermayer <michaelni at gmx.at>
      rv10 sig11 fix by Michael Niedermayer <michaelni at gmx.at>
      (m)jpeg pad/flush with 1 instead of 0, fix by Rik Snel <rsnel at cube.dyndns.org>
      workaround gcc bug "m" (0)
      PATCH by Rik Snel <rsnel at cube.dyndns.org>
      PATCH by Rik Snel <rsnel at cube.dyndns.org>
      using Juanjo's way to pass forced type - especially as I_TYPE/P_TYPE defines aren't exported in the public interface, and the flagged method matches the way as qscale forcing work
      postprocess array increased to handle images up to 2048x1536
      Support for external huffman table and various fixes by Alex Beregszaszi <alex at naxine.org>
      ar -s has portability problems, anyway rc is enough without s -- suggested by Atmosfear
      10l - MMX rgb2bgr 32bpp expects num_of_bytes instead of num_of_pixels
      big cosmetics patch, cleanup of messages printed by mplayer and libs.
      sync with mplayer xp
      IF09 is alias for YVU9 (actually it has extra 4th plane containing MC change
      don't exit decoder after decoding first field -> fixes angels.avi interlacing
      warning fixes by Dominik Mierzejewski <dominik at rangers.eu.org>
      use detected WORDSIZE instead of warning, when inttypes.h doesn't define __WORDSIZE
      fdct_mmx -> ff_fdct_mmx (renamed to avoid namespace conflict with xvid)
      tons of warning fixes, also some 10l bugfixes, including Dominik's PVA bug
      DV ntsc support, 411P colorspace added
      10l - MMX/FPU state was not restored, causing nonsense fpu behaviour in caller (mplayer)
      3rd 10l - don't let it read data beyond the end of frame
      verbose can be negative
      handle init error in ffmpeg12
      *HUGE* set of compiler warning fixes, unused variables removal
      10l - set pixelfmt (colorspace) _before_ calling get_buffer() callback
      - removed nonsense *.d dependancy stuff, there was already a better 'make dep' support in it
      Utility functions (CRC calc & float->int converters)
      Compile a52 util functions too
      looks better (req. for mp-G2)

Art Clarke (12):
      free private streamcontext when closing demuxer, fix memory leak, patch by Art Clarke, aclarke at vlideshow dot com
      Check if there is at least a stream before writing trailer.
      Allocate static tables on stack, not heap.
      Use static vlc structure to decode layer 3 huffman tables.
      downgrade mov negative ctts log message to warning, patch by Art Clarke, aclarke at vlideshow dot com
      Fix static leaks in h264.c.
      Fix av_picture_copy missing pixels on packed planar AVPictures with odd size.
      Remove unused variable, patch by Art Clarke, aclarke vlideshow com.
      Add a warning log when trying to encode into a container requiring global
      av_write_trailer() must be done before closing codecs.
      Set global header flag for every container that requires it.
      Make output_example use av_interleaved_write_frame() instead of av_write_frame().

Artur Grabowski (1):
      64bit incompatibility bug, patch by Artur Grabowski <art at blahonga.org>

Assaf Yaari (1):
      block algn patch by ("Assaf Yaari" <assafy at mobixell dot com>)

Attila Kinali (1):
      add missing include of bswap.h

Aurelien Jacobs (789):
      Here are juste two added #ifdef CONFIG_ENCODERS to allow
      adapting existing mmx/mmx2/sse/3dnow optimizations so they work on x86_64 patch by (Aurelien Jacobs <aurel at gnuage dot org>)
      gcc 3.4.3 preversions do not appreciate invalid instruction and operand combinations anymore patch by (Aurelien Jacobs <aurel at gnuage dot org>)
      adapting existing mmx/mmx2/sse/3dnow optimizations so they work on x86_64
      10000l : fix a crash on x86 due to an horrible mistake in my x86_64 patch
      fix few x86_64 registers handling
      makes funnyCode pages executable (for CPU with NX bit)
      add an AMD64 specific implementation of rdtsc()
      drop most url_fileno() calls (allows to use ByteIOContext directly in caller apps instead of URLProtocol)
      install libav(codec|format).pc/libpostproc.pc for pkg-config (original patch from <j at v2v.cc>)
      add support for muxing subtitles in mpeg-ps
      add hdv2 fourcc to MPEG2 codec mapping
      fix unused variable warnings
      add a Creative VOC (de)muxer
      voc: set bits_per_sample correctly (fix mplayer playback)
      add Creative 8 bits ADPCM schemes support
      better match CVS policy in the indent recomendation
      Clarify creative ADPCM changelog entry
      fix ffserver regtest inconsistency between different arch (and even different kind of x86)
      update ffserver.regression.ref to the actual md5 values
      get ride of perl and sort in ffserver regtest (and avoid problem with behavior of sort changing with locales)
      don't try to test streams which are commented out in test.conf
      -loop was renamed to -loop_input
      Update ffserver regression tests:
      document clip functions
      clip_uint8 now return an uint8_t so no need to &0xFF
      simplify the voc demuxer using av_get_bits_per_sample()
      set proper bits_per_sample value for ADPCM codecs
      fix arm asm compilation in mpegaudiodec
      cook shouldn't try to enforce the bitstream reader type
      Fix A32_BITSTREAM_READER compilation on x86
      move some functions to bitstream.h to avoid conflicts
      force usage of ALT_BITSTREAM_READER where needed
      ALT_BITSTREAM_READER_LE is obviously not needed for the tta demuxer, as there was a typo in the define
      use the A32_BITSTREAM_READER by default on ARM (faster)
      VP5 and VP6 video decoder
      rename vector to vect to avoid clash with Apple gcc
      don't use dsp.put_pixels_tab[0][0] with unaligned ptr (fix artifacts on altivec)
      add a VP6F fourcc for apps which rely on fourcc
      Enables back the mmx/sse optimized version of the vp3 idct.
      replaces back some C code by dsp.put_pixels_tab[0][0] (revert r6231)
      disable vp3 mmx idct for theora files to avoid artifacts
      rename vector to vect to avoid clash with Apple gcc
      make some global var static (fix a build failure with shared lib on x86_64)
      10l inverted condition check generated an endless loop
      use coded_width and coded_height instead of width and height
      use the adjustment value present in FLV to crop VP6 video
      don't use ast before checking it's not NULL
      document the extradata protocol for VP6F
      fix flvdec.c file description comment
      replace coder/decoder file description in libavformat by muxer/demuxer
      fix mp3 in swf decoding
      Don't need to check for NULL before av_free().
      cosmetics: reindent after last commit
      Allow reading matroska title.
      Better way to detect cluster (fix files encoded with Haali's muxer).
      Add myself as matroska.c maintainer.
      add audio output sampling freqency reading in matroska
      add support for AAC in matroska
      Fix avidec.c compilation when dv demuxer is disabled.
      cosmetic: fix indentation after previous commit
      Allows playback of truncated matroska files.
      simplify free()+set to NULL using av_freep()
      matroska: properly handle real video extradata
      matroska: split real video frames so that each packet contains only one slice
      cosmetics: reindent after last commit
      add support for tta in matroska
      split voc.c into vocdec.c and vocenc.c
      move common voc muxer and demuxer data to voc.c
      add support for FLAC in matroska
      flac: allow reading of full metadata headers from extradata
      No need for special trick anymore for flac in matroska.
      cleanup the #ifdef mess in muxer/demuxer registration
      matroska: scale timecode according to time_scale
      cleanup the #ifdef mess in encoder/decoder/parser registration
      put_*() are not only useful to muxers and protocol
      small optimization
      Add support for VP60 and VP61.
      more simplification
      add support for another variant of vp6
      cosmetics: reindent
      cosmetics: indentation
      remove now useless codec_tag setting code in voc muxer
      flashsv encoder also depends on zlib
      fix a typo in deps
      ensure that data[0] is set to NULL for next frame
      simplify release_buffer using FFSWAP
      use av_noinline instead of __attribute((noinline))
      Use pointers to avoid copying AVFrame.
      Use pointers to avoid copying AVFrame.
      cleanup setting of returned data frame
      enable some 64bit optimized code on 64bit arch
      fix an off-by-one error in probing function
      split a matroska_parse_block() function from matroska_parse_blockgroup()
      cosmetics: fix indentation of the new matroska_parse_block() function
      add support for simple blocks (ie. matroska v2)
      don't rely on a specific appearance order for ebml elements at the same level
      use codec_get_id() instead of deprecated codec_get_[bmp/wav]_id()
      fix a stupid bug in ebml_read_sint()
      parse clusters until some packets are queued
      reindent after last commit
      extract ebml_read_binary() out of matroska_parse_block()
      check if current block contains a B frame and gives this info to parse_block()
      revert r7181
      reorder pts of packets from tracks using V_MPEG* codecs
      cosmetic: indent after last commit
      move timecode calculation out of laces loop
      scale duration and default_duration
      calculate default_duration for AAC tracks
      calculate pts for laced packets
      set pkt->duration for laced packets too
      remove no more needed parameters
      remove outdated comment
      disable pts reordering code and initialize a parser instead
      extract vorbis header spliting code into a reusable function
      use generic xiph header spliting func to split theora headers
      add support for V_THEORA into Matroska
      properly set *data_size when returning >= 0 values in shorten_decode_frame()
      cosmetics: indentation
      remove wrong and no more used packet reordering code
      remove unused context field
      properly set streams start_time
      cluster position is relative to segment_start
      move matroska_find_track_by_num() upper in the file
      add support for seeking in matroska files
      add an intermediate variable (prepare for next patch)
      add support for seeking to a keyframe instead of a random frame
      cosmetics: indentation
      add an enum for need_parsing
      move ac3 tables from a .h to a .c
      cosmetics: align
      set CODEC_CAP_DR1 for vp5 and vp6 decoders
      parser.h uses AVCodecContext so it needs to include avcodec.h
      move mpeg4video_parser in it's own file
      move mpegaudio_parser in it's own file
      avoid a 'discards qualifiers from pointer' warning
      move mpegvideo_parser in it's own file
      add a ff_ prefix to the now exported mpeg1_find_frame_end() function
      avoid a clash between two definitions of SEQ_START_CODE
      ff_frame_rate_tab is now declared in mpegvideo.h
      make some parser parameters const to avoid casting const to non-const
      fix display of theora videos with visible size smaller than encoded size
      cosmetics: fix inconsistent indentation
      avoid multiple inclusion of ac3.h
      move aac and ac3 parsers in their own files
      cosmetics: rename for consistency after previous aac and ac3 parsers move
      remove no more useful includes
      move defines and enums out of cavsdata.h
      move cavs parser in it's own file
      move ff_mpeg4_find_frame_end() from h263dec.c to mpeg4video_parser.c
      add missing rules for wmv2 compilation
      split msmpeg4 data to easily share some of it with vc1.c
      use header instead of copying table declarations
      those tables don't need to be extern
      disable reference to msmpeg4 and wmv2 code when those codecs are not compiled in
      loosen vorbis_enc.c dependencies by spliting vorbis.c
      rename vorbis.c to vorbis_dec.c and vorbis_common.c to vorbis.c
      split h264.c to move parser in its own file
      cosmetics: reindent
      move mjpeg parser in its own file
      move pnm parser in its own file
      make compilation of pnm.c optional
      rename pnm.c to pnmenc.c and pnm_common.c to pnm.c
      jpegls never call this code, so those tests are useless
      make compilation of mjpeg.c optional
      cosmetics: indentation
      replace some #ifdef CONFIG_* by if(ENABLE_*)
      cosmetics: indent
      split jpeg_ls into jpeglsdec, jpeglsenc and jpegls
      jpegls encoder does not call this function, so remove useless tests
      remove some unused mjpeg encoding variables
      better separation of vp3dsp functions from dsputil_mmx.c
      split mjpeg.c into an encoder and a decoder file
      add proper prefix to new mjpeg extern func/var
      prevent misdetection by using a stricter regexp
      remove libdts decoder, we have a native dts decoder
      split ljpeg encoder out of mjpeg.c
      add a proper prefix to all mjpeg encoder exported functions
      move the mjpeg_encoder struct from mpegvideo.c to mjpegenc.c
      move some common mpeg audio tables from mpegaudiodectab.h to mpegaudiodata.c
      add proper prefix to extern mpeg audio data tables
      move mp3_header_decompress bitstream filter in its own file
      move mp3_header_compress bitstream filter in its own file
      move noise bitstream filter in its own file
      move remove_extradata bitstream filter in its own file
      move dump_extradata bitstream filter in its own file
      move imx_dump_header bitstream filter in its own file
      move mjpega_dump_header bitstream filter in its own file
      allows to disable jpegls decoder
      move sp5x decoder in its own file
      10l: forgot to svn add mjpegdec.h
      move mjpeg-b decoder in its own file
      move MJpegDecodeContext declaration from mjpeg.h to mjpegdec.h
      loosen dependencies over mpegaudiodec
      add a ff_ prefix to some mpegaudio funcs
      remove dependency of mpeg audio encoder over mpeg audio decoder
      make compilation of mpegaudiodec conditional
      vorbis encoder missed rename to libvorbis
      define ENABLE_* for every CONFIG_ item
      cosmetics: alignment
      split rm muxer and demuxer in their own files
      cosmetics: indentation
      use FFSWAP()
      typo
      use if(ENABLE_SWSCALER) instead of #ifdef CONFIG_SWSCALER
      cosmetics: indentation
      fix an infinite loop in case one cluster is not enough to demux a packet
      add support for real audio in matroska
      cosmetics: indentation
      cosmetics: use one type per field in every structs
      move codecs declarations from avcodec.h to allcodecs.h
      use get_bits_long() where needed
      generate ENABLE_* defines also for all the HAVE_* values
      use if() instead of #ifdef and ensure all possible optimisations are enabled
      generate ENABLE_* defines also for ARCH_*
      use intreadwrite functions in vocdec
      use intreadwrite functions in matroskadec
      cosmetics: too much blank lines
      remove useless empty function
      simplify
      drop blocks which are part of unsupported tracks
      cosmetics: indentation after last commit
      split mpeg12 encoder out of mpeg12.c
      cosmetics: indentation
      add proper ff_ prefix to new extern symbols
      move mpeg encoder specific initialization in the encoder specific file
      make DCT_common_init() static
      forgot to svn add mpeg12.h in r9477
      include snow encoder only if it is not disabled
      move ff_copy_bits to bitstream.c
      simplify ff_copy_bits: merge 2 test branches
      split SVQ1 decoder and encoder in their own files
      simplify ff_copy_bits() when compiled with config-small
      properly set AVStream.language according to Matroska track header
      duration must not be scaled
      compile rangecoder.o only when needed
      split rawvideo encoder and decoder in their own files
      move ff_set_qscale from h263.c to mpegvideo.c
      don't compile some mpeg4 encoder code when encoders are disabled
      Split ff_set_mpeg4_time() and move the non mpeg4 specific part
      move init_vlc_rl from h263.c to mpegvideo.c
      Better place to disable some mpeg4 code and ensure that
      Avoid linking with h263.c functions when the relevant codecs
      cosmetics: indentation
      makes compilation of h263.o and mpegvideo_enc.o conditional
      conditional compilation of motion_est.c
      conditional compilation of ratecontrol.c
      Move public API functions av_codec_get_tag() and av_codec_get_id()
      add support for Matroska subtitle tracks
      comment explaining that text subtitle need to be UTF-8 for now
      add support for ASS like subtitles in Matroska
      The svq3 decoder depends on h264idct.o and also requires h264_parser.
      Allow selecting only the vp6f decoder.
      The mpegaudio parser depends on mpegaudiodata.o.
      The c93 demuxer depends on vocdec.o.
      All those demuxers use pcm_read_seek() and thus depend on raw.o.
      All those demuxers don't depend on riff.o anymore.
      The dv1394 demuxer depends on dv.o.
      The dxa demuxer depends on riff.o.
      Allow selecting only the mpegtsraw demuxer.
      Allow selecting only the vc1 demuxer.
      Avoid using undefined symbols.
      add support for S_ASS and S_SSA codec string
      help some gcc version to optimize out those functions
      reorder matroska codec tags so that the encoder picks up the appropriate tag
      use the most appropriate codec id for A_MPEG/L1 and L2
      remove wrong const qualifier
      copy packet duration when doing a stream copy
      copy language information from input stream to output
      add proper support for subtitles muxing/stream copying
      replaces hardcoded values by the equivalent enum definitions
      cosmetics: alignment
      merge some REGISTER_MUXER/DEMUXER into REGISTER_MUXDEMUX
      ensure that negative block_time are properly checked
      ensure av_noinline is always defined
      add a new av_clip_int16() function to libavutil
      use av_clip_int16() where it makes sense
      various simplifications around recent av_clip_int16() usage
      one more simplification
      more usage of av_clip()
      Try size 640x480 for FLC files which don't specify their size.
      calculate default_duration according to framerate only when not already set
      no need for special case default_duration for AAC
      Don't pre-scale default_duration to avoid precision loss in fps calculation.
      cosmetic: indentation
      ensure configure only accept valid --list-* parameters
      asv1.c directly uses Put/GetBitContext, so it should include bitstream.h
      cosmetics: some reordering
      create libavcodec/sh4 dir when building out-of-tree
      parsers and bsfs declarations are not meant to be part of public API
      move get_unary() to its own file
      cleanup get_unary()
      add get_unary_0_33() to help gcc with inlining
      add a warning message for unsupported kind of encoding
      *_protocol are not part of public API
      rtp and rtsp demuxer declarations are not part of public API
      move unrelated functions declarations out of allformats.h
      remove obsolete and wrong extern declaration
      loosen dependencies over allformats.h
      disable loads of debug messages to reduce object size
      Add support for VP6A in flv.
      remove last dependencies on allformats.h and allcodecs.h
      allcodecs.h and allformats.h are not used anymore
      add support for yuva420p colorspace (yuv420p + alpha)
      add support for yuva420p colorspace (yuv420p + alpha)
      some simplifications and uniformisation
      rename vp56_b6to3 to vp56_b2p
      move all model related tables into their own struct
      add a new vp6a codec (add alpha plan support to vp6)
      cosmetic: indent
      simplify
      revert r10527
      use the VP6A codec
      consistent spelling: plan => plane
      rename one variable to be consistent with the rest of the file
      simplify
      use a table to parse AVI file header
      add support for AVI files with On2 header
      remove useless #ifdef
      fix declaration name from MPV_common_init_ppc to MPV_common_init_altivec
      remove alternative declaration of h263_encode_init()
      merge some #ifdef CONFIG_ENCODERS sections
      dnxhdenc depends on mpegvideo_enc (usage of ff_convert_matrix)
      oops... mpegvideo_enc also depends on motion_est and ratecontrol
      VP6A needs its Makefile rules too
      moves fraps huffman decoder to its own file, making it more generic
      add support for VP6 with huffman encoded blocks
      Cygwin don't like this function declaration.
      simplify bitstream reading
      remove unused close function
      remove unused tags definition
      remove unused code
      remove unused context vars
      merge declaration and initialization
      remove useless &0xFF
      add support for VP6 in electronicarts demuxer
      cosmetics: reindent
      Remove wrong multiple inclusion guards.
      small simplification
      fix comment
      add a log message
      move audio header parsing in its own function
      set a default value for num_channels
      remove 2 temporary variables
      move vp6 header parsing into its own function
      better log message
      add audio and video codec context vars
      parse header chunk in a loop (don't depend on a fixed chunk ordering)
      cosmetics: indentation
      add some error processing
      add support for big_endian encoded headers
      make compression_type a function local var instead of a context var
      check compression_type to assign audio codec
      add support for PCM codec
      parse sample rate instead of setting a default one
      simplify
      fix big endian files support
      set bytes per sample in the context
      compute pts according to the audio codec
      check if we found an audio track before initializing the audio stream
      cosmetics: indentation
      skip unsupported audio track data
      better check for video track detection
      ensure all compiled in codec/muxer are registered
      cosmetics to make version.sh more VCS agnostic
      add support for git in version.sh (use short hash as version number)
      Make Matroska demuxer output full frames instead of slices for RealVideo
      remove unused context var
      cosmetics: Add #endif comment.
      cosmetics: reindent
      remove some useless initializations (the whole context is initialized to 0)
      convert a if() into a switch() to ease addition of new tags to probe
      avoid infinite loop at the end of files which are not properly terminated
      add parsing of 1SNh header in the EA demuxer
      don't use #define with names starting with _
      probe files starting with the 1SNh tag
      move framecrc muxer in its own file
      rename crc.c to crcenc.c
      cosmetics: sort
      don't register adpcm encoders which are not implemented
      cosmetics: alignment
      remove unreachable code
      remove useless #undef
      simplify dpcm decoders declaration using a macro
      remove useless initialization to 0 of adx context
      cleanup and simplify adx_decode_header()
      revert part of r10899 which was not intended
      remove useless debug func
      split adx encoder in its own file
      move adx.c to adxdec.c
      add Electronic Arts XAS ADPCM decoder
      add support for demuxing Electronic Arts .cdata files
      split definition of PCM_CODEC into PCM_ENCODER and PCM_DECODER
      probe more kind of electronic arts files
      rename some msmpeg4 symbols and make them non-static
      split wmv2 in its own file
      cosmetics: indentation
      correctly set audio codec when demuxing sofdec files
      fix libogg muxer dependency checking
      fix negative fseek
      fix potential buffer over-read
      add multiple inclusion guards
      make some tables static
      cosmetics: whitespace
      avoid some halfpel related #ifdef
      split wmv2 encoder and decoder in their own files
      add pcm_s16le_planar support for electronicarts files
      remove some empty close/init functions in avcodec
      cosmetics: alignment
      10l: init_get_bits want size in bits, not bytes
      make ff_p* vars extern so that they can be used in various *_mmx.c files
      use ff_ prefix for extern vars
      build vc1dsp_mmx.c in its own compilation unit
      convert some #ifdef CONFIG_ to if(ENABLE_
      cosmetics: indentation
      remove useless #ifdef around extern declaration
      remove some unused ff_p* vars from dsputil
      add ff_ prefix to all simple_idct symbols
      add required include to make this file self-contained
      move FLAC mmx dsp to its own file
      use proper url_is_streamed() API
      pb->buf_end is already set by url_open_buf()
      remove double check of pb->read_seek
      cosmetics: indentation
      move internal queue freeing code in its own function
      clear internal queue when seeking
      move num_packets=0 where it belongs
      mention libavfilter in warning about deprecated vhook
      improve CRC API
      install crc.h which is now part of public API
      Make ff_mkv_codec_tags lie entirely in .rodata section.
      Fix access to the last element of the table.
      Check return value of avc_parse_nal_units().
      Add one more parameter to avc_parse_nal_units so that it allows not to
      extract isom_write_avcc into a function
      Move isom_write_avcc() and related functions into a separate file.
      Use the isom avcc formatting for h264 extradata in matroska.
      matroskaenc now depends on avc.o
      Matroska muxer needs to format all NAL units, not only extradata.
      add a ff_ prefix to newly exported functions from avc.c
      nellymoser uses mdct functions
      and thus, nellymoser obviously also depens on fft.o
      add missing Makefile rules for zork pcm encoder/decoder
      add missing Makefile rules for pcm muxers/demuxers
      add proper matroskaenc dependency over xiph.o
      ensure av_rescale_q() can be calculated (won't divide by zero)
      const
      const
      don't cast const away
      use AV_RL* instead of le2me_*
      non-const 2nd parameter for strtol/strtod
      compile various fdct only when encoders are enabled
      split encoding part of dsputil_mmx into its own file
      Avoid infinite loop.
      makes really sure that no undefined behavior can happen
      simplification proposed by Rich
      avi.h does not use anything from avcodec.h.
      include required headers for int64_t and AV_RB16()
      include required header
      dsputil.h is a local header
      move draw_edges() into dsputil
      move ff_init_scantable() into dsputil
      remove old draw_edges declaration cruft
      Fix compilation on powerpc.
      fix wmv3 dependencies
      move ff_emulated_edge_mc() to dsputil
      move run length table constants to rl.h
      move relevant declarations to mpeg12data.h
      remove #include "mpegvideo.h" where it is not needed
      move EDGE_WIDTH definition allong with draw_edges where it belongs
      remove more useless mpegvideo.h includes
      Avoid linking with h263 and mpeg4 find_frame_end()
      add some required objects dependencies
      remove some unneeded artificial dependencies from configure
      cosmetic: align
      define BI frame type along with other frame types
      doxyfy frame type comments and add some more comments
      add dependencies for matroska audio and nut muxers over lavc
      move ff_mpeg1_dc_scale_table to mpegvideo.c
      make compilation of mpeg12* and mpegvideo conditional
      huffman: pass hnode_first as a flag instead of as an argument on its own
      huffman: add a zero_count flag and use it in fraps
      add FF_ prefix to all (frame)_TYPE usage
      cosmetics: alignment of the codec_bmp_tags table
      consistent use of MKTAG() in the codec_bmp_tags table
      remove useless and misleading comments
      cosmetics: remove some empty lines
      add a temporary variable to simplify the code
      add out of bound stream index checks
      check return value of matroska_find_track_by_num (fix a segfault)
      use common aac sample rate tables
      matroskadec: add support for track content encoding
      ensure vp56_decode_frame always return unchanged buf_size
      matroskadec: fail gracefully when seeking don't work
      remove useless comment
      matroskadec: add support for lzo compressed tracks
      cosmetics: indentation
      use enum type instead of int
      matroskadec: add support for zlib compressed tracks
      add missing \n in av_log message
      Add a check for bzlib; used in the Matroska demuxer.
      cosmetics: alignment
      matroskadec: add support for bzlib compressed tracks
      matroskadec: allows inserting chapters with unspecified end
      allows adding chapters with NULL title
      cosmetics: indentation
      don't set a dummy chapter title when title is unknown
      remove useless title check
      cosmetics: indentation
      compute the end of all chapters when end is not specified in the container
      matroskadec: use the proper timestamp scaling for index entries
      bump micro version after commit r13615
      matroskadec: move ARRAY_SIZE() macro at top of the file
      reorganize matroska_add_stream() to fix potential mem leak and buffer overflow
      matroskadec: avoid potential mem leak
      matroskadec: fix potential mem leak
      matroskadec: sane default value for track->time_scale
      adjust comment to reality
      Ensure the first audio stream is selected as the default stream
      matroskadec: don't remove dummy level when none was added
      cosmetics: indentation
      matroskadec: remove some useless code
      matroskadec: remove some useless code
      matroskadec: fix a memory leak
      VP3 parser can also handle VP6 without any changes.
      add some missing dependencies
      correct path in *-uninstalled.pc
      matroska: add support for V_QUICKTIME track type
      matroska: add support for A_QUICKTIME/QDM2 codec
      matroskadec: use more robust av_strlcpy() instead of strcpy()
      matroskadec: remove storage of various unused flags
      matroskadec: simplify flag default
      matroskadec: remove unused is_bframe flag
      matroskadec: add an ebml generic parser
      matroskadec: use generic ebml parser to parse ebml header
      matroskadec: store an AVStream pointer instead of a stream index
      matroskadec: create new AVStream earlier
      matroskadec: return pointer instead of index in matroska_find_track_by_num()
      matroskadec: use generic parser to parse index
      matroskadec: use generic parser to parse tags
      matroskadec: use generic parser to parse chapters
      matroskadec: use generic parser to parse attachments
      matroskadec: use generic parser to parse info
      matroskadec: use generic parser to parse tracks
      matroskadec: declare var where it's used
      matroskadec: use generic parser to parse seekhead
      matroskadec: remove useless dummy_level variable
      matroskadec: cosmetics: indentation of matroska_execute_seekhead()
      matroskadec: invert a test
      matroskadec: move declaration to upper block
      matroskadec: cosmetics: indentation of matroska_read_header()
      matroskadec: move context settings to matroska_read_header()
      matroskadec: use generic parser to parse matroska from toplevel
      matroskadec: use generic parser inside matroska_execute_seekhead()
      matroskadec: use generic parser to parse clusters
      matroskadec: remove some unused ebml funcs
      matroskadec: cosmetics: remove useless braces
      matroskadec: remove useless initialization
      matroskadec: remove now unused context fields
      matroskadec: remove unused struct name
      matroskadec: cosmetics
      matroskadec: remove some useless or out of date comments
      matroskadec: remove some useless empty lines
      matroskadec: cosmetics: placement of 'else if'
      matroskadec: remove useless check for cluster presence
      matroskadec: remove useless levelup check
      matroskadec: extract read_level_up out of ebml_read_element_id
      matroskadec: remove unused levelup parameter from ebml_read_element_id()
      matroskadec: expand useless ebml_peek_id() wrapper
      matroskadec: remove now useless expected_id parameter from ebml_parse()
      matroskadec: simplify/cleanup error logging in ebml_read_*()
      matroskadec: extract common code out of ebml_read_*()
      matroskadec: use an EbmlBin as parameter to ebml_read_binary()
      matroskadec: Use string instead of char table as probe data comparison source.
      matroskadec: cosmetics: consistent style for function declarations
      matroskadec: simplify check for hierarchy level
      matroskadec: don't care about the number of bytes read by ebml_read_element_id()
      matroskadec: rename ebml_parse() to ebml_parse_nest()
      matroskadec: new ebml_parse() function, centralize ebml_read_element_id() calls
      matroskadec: directly use ebml_parse() or ebml_parse_id() where it's enough
      matroskadec: remove now unused parameter 'once' from ebml_parse_nest()
      matroskadec: don't try to re-read already parsed ID in matroska_parse_cluster()
      matroskadec: remove now useless peek_id
      matroskadec: remove now useless wrapper ebml_read_seek()
      matroskadec: remove now useless wrapper ebml_read_element_length()
      matroskadec: reorder some functions in a more logical order
      matroskadec: update copyright year and related comments
      matroskadec: remove useless comments in ff_mkv_codec_tags
      matroska: expand useless define for MS compat codec id strings
      matroskadec: implement matroska_ebmlnum_uint() using ebml_read_num()
      matroskadec: simplify matroska_ebmlnum_sint()
      matroskadec: expand useless ebml_read_element_id() wrapper
      matroskadec: remove useless num_streams
      matroskadec: update seek regression test to match new ebml parser
      matroskadec: use av_freep(&x) instead of av_free(x);x=NULL
      matroskadec: add basic tags support (metadata)
      matroskadec: parse available blocks even when cluster parsing failed
      ffserver: SA_RESTART and unsetenv require _XOPEN_SOURCE
      define some math constants so as not to depend on _XOPEN_SOURCE
      move M_PI definition along with other math definitions
      strcasecmp() requires #include <strings.h>
      matroskadec: list some more ebml IDs found in the wild and that we ignore
      ensure we get explicit definition of various _XOPEN_SOURCE functions we use
      use h_addr_list[0] instead of h_addr as this is the posix hostent field name
      matroskadec: handle EBML_ID_VOID and EBML_ID_CRC32 in a generic way
      matroska: add support for most variants of PCM
      matroska: fix regression test according to r14862
      matroskadec: stop parsing when skipping en element crossing over the end of file
      matroskadec: fail cleanly when the required compression lib is not compiled in
      export sample_aspect_ratio read by the demuxer in a separate field
      convert every muxer/demuxer to write/read sample_aspect_ratio from/to
      matroskadec: make aac_profiles array const
      voc: add ff_ prefix to some global const data
      matroskadec: move setting of matroska->done inside matroska_parse_cluster()
      matroskadec: don't try to seek to negative timestamp
      matroskadec: on the fly index construction for index-less files
      matroskadec: full seeking support in index-less files
      matroskadec: unset matroska->done when seeking
      Fix the seek regressions to match changes in r14956, r14959 and r14960.
      Add a temp variable to ff_sine_window_init() to match the way wma.c computed
      mxfdec: set sample_aspect_ratio
      matroskadec: index timestamps are supposed to be in the timebase of the stream
      remove useless assignment (st is never read after this line)
      matroskadec: seek to the last position known in the index before resorting to
      matroskadec: cosmetics: indentation
      simplify ac3_probe() and eac3_probe(). patch by Aurelien Jacobs.
      matroska: add E-AC-3 support
      theoradec: small 4MV code reorganization to ease future improvements
      theoradec: skip decoding of uncoded MV in 4MV code
      theoradec: cosmetics: indentation
      matroska: subtitle display duration must be stored in pkt->convergence_duration
      matroskadec: fix ASS subtitle track packets before emitting them
      matroskadec: ensure that the timecode added to the index are is the one
      matroskadec: remove useless braces
      matroskadec: remove now useless vstream and is_video_key_frame
      bump lavf micro version to allow checking for the matroska/ASS bugfix in r15211
      fix seek regression tests after r15229
      remove one more stuff under #if LIBAVCODEC_VERSION_INT
      matroskadec: sanitize track time_scale before using it
      matroskadec: move timecode calculation a little earlier
      matroskadec: after seeking, skip up to the desired timestamp instead of
      matroskadec: ensure that overlapping subtitles are not added to the index
      matroskadec: demux relevant subtitle packets after a seek
      matroskadec: prevent seeking before the first keyframe of the reference stream
      matroskadec: simplify, first_timecode is already in the index
      assdec: don't remove CR, they are part of the spec
      matroskadec: pass duration as parameter of matroska_fix_ass_packet()
      matroskadec: set duration only for CODEC_ID_TEXT subtitles
      matroskadec: ass events lines must end with a DOS style EOL
      matroskadec: merge ASS events which share the same pts in the same AVPavket
      atrac3: ensure input frame is not overwritten (it is const)
      matroskadec: don't merge packets which have no timestamp
      matroskadec: cosmetics: indentation
      atrac3dec: cosmetics: indentation
      matroskaenc: remove useless mkv_block_size() function
      matroskaenc: pass the packet size directly as parameter of mkv_blockgroup_size()
      matroskaenc: use correct EBML ID for block duration
      matroskaenc: correct muxing of ASS packets
      Change one more variable type from int to enum PixelFormat.
      allows calling avcodec_find_(en|de)coder_by_name with NULL parameter
      lookup AVCodec by name
      cosmetics: indentation
      vp6dec: ensure we don't try to use a buffer with negative size
      uses FF_ARRAY_ELEMS() where appropriate
      matroskaenc: require global header
      10l: missing change from r15667 - global header also affect seek test results
      svq3: remove unused include
      enable zlib and bzlib only when useful
      remove useless ifdeffery, tscc won't do anything useful without zlib
      set correct number of fractions for multitable
      matroskadec: fix index timestamps for some broken files
      avoid POSIX reserved _t suffix
      avoid POSIX reserved _t suffix
      cosmetics: vertical alignment
      rename Vp56Model to VP56Model for consistency
      matroska: sort codec tags list
      Get rid of mmx_t.
      avoid POSIX reserved _t suffix
      indeo3: replace sizeof(struct ...) by sizeof(*var)
      vp56: don't reset dimensions to 0 in codec init
      Initialize s->opaque before calling url_resetbuf().
      Remove duplicated dot product code. Use dsputil's
      add a termination condition
      cosmetic: indent
      add metadata in AVStream, AVProgram and AVChapter
      rename AVMetaData to AVMetadata and meta_data to metadata
      add a typedef for AVMetadata
      Add a metadata compatibility layer, so that when a user application set
      ignore year=0 and track=0
      ensure that no GPL code is linked to LGPL only libavcodec
      free all allocated metadata structures
      Add a second metadata compatibility layer, so that metadata that gets
      move ff_metadata_sync_compat to metadata_compat.c
      rename ff_metadata_sync_compat to ff_metadata_mux_compat
      initialize AVFormatContext before calling av_write_header()
      fix warning: assignment makes pointer from integer without a cast
      add the necessary include for LIBAV*_VERSION_*
      fix version protection
      Change semantic of CONFIG_*, HAVE_* and ARCH_*.
      Change semantic of CONFIG_*, HAVE_* and ARCH_*.
      replace all occurrence of ENABLE_ by the corresponding CONFIG_, HAVE_ or ARCH_
      libavutil/lzo.h is a local header
      simplify: group all the AUDIO_NONSHORT parameters in the same place
      WORDS_BIGENDIAN must be checked with #ifdef
      matroskadec: ensure we only consider chapters which are properly ordered
      move timer related code in a new timer.h file
      move NULL_IF_CONFIG_SMALL() definition into internal header
      explicitly disable CONFIG_AUDIO_NONSHORT
      add -Wundef compiler flag, to catch typo in #if or similar situations
      export gcd function as av_gcd()
      remove ff_get_fourcc() and use AV_RL32() instead
      matroskadec: parse the WAVEFORMATEX header and discard it from extradata
      matroskadec: merge ByteIOContext declarations at upper level
      replace AV_METADATA_IGNORE_CASE flag by a new AV_METADATA_MATCH_CASE flag
      moves mid_pred() into mathops.h (with arch specific code split by directory)
      Remove detection of mmsh protocol in ffplay.
      matroskaenc: correctly handle h264 streams with 3 bytes nal startcodes
      'code' is used to index a table so make it unsigned
      make it clear that the new metadata API is a work in progress
      vp5: don't try decoding a P frame before any I frame was parsed
      add a ff_gcd() function again, for compatibility with old libavcodec
      typo
      don't support the mp3 extension in the mp2 muxer
      cosmetics: whitespace
      modify the way to pass parameters to av_metadata_set()
      use new metadata API in the mp3 demuxer
      use new metadata API in the mp3 muxer
      cosmetics: indentation
      simplify usage of dstlen as it is not a parameter anymore
      cosmetics: indentation
      it seems I never added myself in there before
      use new metadata API in rtp muxer
      use new metadata API in rtsp demuxer
      simplify
      use new metadata API in aiff demuxer
      use new metadata API in asf demuxer
      use new metadata API in asf muxer
      write all available metadata tags into extended_content_header
      vp56: alpha_offset is uninitialized on purpose
      move vp6_filter_diag4() to a new vp6dsp.c file and use it throught dsputil
      use new metadata API in mpegts demuxer
      remove now useless av_set_program_name() function
      use new metadata API in ogg/vorbis demuxer
      convert ff_pw_64 into an xmm_reg for future use in vp6 sse code
      use new metadata API in mov demuxer
      demux some more metadata tags in mov
      matroskadec: use new metadata API to export some simple information
      matroskadec: export track name through the metadata API
      matroskadec: read track and attachment uid
      matroskadec: add generic metadata support
      matroskadec: add needed definitions forgotten in r17331
      matroskadec: export language of metadata tags when available
      matroskaenc: use new metadata API
      use new metadata API in ape demuxer
      apedec: display a warning when truncating a metadata tag
      use new metadata API in rm (de)muxer
      remove now useless get_str16() from rmdec.c
      use new metadata API in mpegts muxer
      use new metadata API in nsv demuxer
      fix missing dependencies
      matroskadec: return AVERROR_EOF upon detection of end of file
      matroskadec: fix seek regression tests according to r17460
      Take account of struct size in total packet queue size tracking.
      don't trigger metadata compatibility code when user app already set metadata
      use new metadata API in mov muxer
      cosmetics: indentation
      return -1 for errors in ff_mov_iso639_to_lang() to allow for error checking
      movenc: add muxing of language along with metadata tags when available
      fix a memleak in av_metadata_set()
      rename asf-enc.c to asfenc.c and asf.c to asfdec.c for consistency
      move static tables from asf.h to non-static tables in asf.c
      use new metadata API in wc3 demuxer
      use new metadata API in rpl demuxer
      cosmetics: vertical alignment
      use new metadata API in nut demuxer
      use new metadata API in nut muxer
      add a metadata conversion API
      Add some basic metadata conversion tables for matroska and asf.
      simplify metadata conversion and fixes gcc-2.95 at the same time
      new metadata API is now officially part of public API
      convert ffmpeg to new metadata API
      convert ffplay to new metadata API
      convert ffserver to new metadata API
      use new metadata API in r3d demuxer
      use new metadata API in libavformat/utils.c
      fix last usage of old metadata API in ffmpeg.c (handling of AVStream language)
      fix missed usage of old metadata API in mov demuxer
      deprecate old metadata API
      add an API changelog with a first entry for the new metadata API

Axel Holzinger (1):
      Allow to uninstall a custom log callback

BERO (6):
      small optimize mpeg12.c/get_qscale patch by (BERO <bero at geocities dot co dot jp>) and the return idea by arpi
      bitstream reader optimize patch by (BERO <bero at geocities dot co dot jp>)
      decode motion & modulo optimize patch by (BERO <bero at geocities dot co dot jp>)
      sh4 optimized idct & bswap patch by (BERO <bero at geocities dot co dot jp>)
      fastdiv patch by (BERO <bero at geocities dot co dot jp>) with fixes & cleanup by me
      aligned dsputil (for sh4) patch by (BERO <bero at geocities dot co dot jp>)

Baptiste Coudurier (1349):
      Mjpeg interlaced decoding fix, height was wrong while decoding (/2).
      mp4 codecs objecttype handling patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      mov debug clean patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      mov split chunks removal patch by (Baptiste COUDURIER / baptiste.coudurier smartjog com>
      This patch makes output codec keeping input pix_fmt value when using
      add mov_read_ftyp to check iso media file type and set c->mp4 patch by (Baptiste COUDURIER <baptiste.coudurier 0 smartjog o com>)
      add 'in32' fourcc support patch by (Baptiste COUDURIER <baptiste.coudurier 0 smartjog o com>)
      simplify mov_read_hdlr patch by (Baptiste COUDURIER <baptiste.coudurier 0 smartjog o com>)
      Cleans reading stsd audio, and makes it more generic patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com>)
      CTTS support patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com>)
      compute output format before setting language patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      ftyp patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      Support mov stsd atom version 2.
      Make faac_decode_frame return -1 on error and not 0 which avoids
      add support for Motion JPEG 2000 file format patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      referencing more FCP fourccs patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      set track timescale to mov timescale if 0 patch by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      correct computing of DV fourcc in mov muxer
      Fix cropping, depending on enc pix fmt
      do nal reformating when muxing H264, create avcc atom
      create edts atom when muxing B frames, needed to compute cts
      fix a warning when DEBUG is defined
      fix conversion types in dprintf
      use of PRI macros in dprintf
      indention
      fix sample size for uncompressed sound
      use fseeko and ftello to be 64 bit offsets compliant
      use CFLAGS when compiling qt-faststart, to enable largefile support
      fix warning
      keep ftyp first
      > 2GB file fix
      free ftyp atom
      support 64bit date/duration
      clean
      dvcpro is 411p in fact
      use 64bit types
      vorbis seems to be 221, decodes borgcube_vorbis20.mp4, demux mi2_vorbis51.mp4 though not decoded properly (5:1), 249608749a89.3g2 is 225 and is Qclp
      Quicktime creates S8 that way too
      use av_d2q to parse frame rate
      need to dissociate pcm 16bit and 8bit cases
      fix demuxing of CDR-Dinner_LAN_800k.mp4
      let faad init parameters properly, fix many SBR samples
      use 64bit atom version if needed
      fix double free, priv_data is freed in av_open_input_stream
      add mp3on4 detection, fix red.mp4
      let vorbis decoder init too, fix mi2_vorbis51.mp4
      flush packets
      ignore wrong negative cts values created by quicktime
      skip 0byte packets
      fix edts for tracks without b frames
      decrease last_offset when ftyp is present
      dont write stss atom if keyframes only
      write mdat in write_header avoid check in write_packet
      get avcC profile and level indication from sps (from handbrake)
      clean and simplify mov_write_packet
      move declaration to mov.h
      always skip garbage at end of atom
      indention
      add AVID DV fourcc
      set width and height even for mpeg4, pix_fmt to NONE will force decoding anwyay
      skip null fourcc, fix ep2_mystery_m640_dummy.mov
      assing correct codec_id for mjpegb
      10l, undef DEBUG
      fix alac decoder on little endian
      fix damr atom, needed for quicktime player
      simplify, amr.c sets frame_size to 160, and sample size is always 8000
      fix spliting chunks, simplify
      remove unused variable
      add MPEG-2 intra vlc support
      clean and simplify
      add pcm 24/32 le/be support
      fix AMR muxing in mov
      do not write wave if not needed
      fix mulaw muxing
      add quicktime uncompressed 8bit 4:2:2 support
      fix pcm_s24be demuxing, simplify
      fix avc nal parsing
      MPEG-2 4:2:2 encoding support
      fix keyframe check, stss might be there but count set to 0
      fix seeking, typo provoking hard segfault
      fix mbd bits bug, finally
      fix closed gop flag
      simplify, completely ignore streams not recognized, that fixes seeking for some files
      fix mp3 muxing
      fix unused computed value
      check fseek return value
      export all tracks to libavformat layer
      make dump_format output more homogenous
      update libav regression tests for r5449 (adpcm)
      fix aac in mov for ipod
      simplify
      check only existing streams
      write exact values for audio stsd v1
      simplify and write correct values conforming to specs
      10l, all modes is 0x81FF
      typo fix
      clean
      skip others fourcc, if present
      add MS ADPCM fourcc
      av_get_bits_per_sample and due simplifications
      new mov_read_packet, mov_read_seek, fix seeking in ni mov, fix demuxing adpcm
      better handling of partial files
      clean up useless code
      simplify, use LE_32, BE_32 macros
      assign exact codec id, needed when stream copy
      GXF muxer
      Changelog update for GXF muxer
      MXF demuxer
      declare and use UID type
      demuxer reworked, more accurate parsing, prepare handling of other operational patterns, streaming demuxing, simplified codec detection
      add error messages
      rename variables
      remove useless variables
      use ternary
      cosmetics
      get rid of DECLARED_ALIGNED
      add some debug infos
      simplify PRINT_KEY
      add common codecs uls
      parse rgba descriptor (jpeg2000, uncompressed)
      add uncompressed picture support
      another MPEG-2 long gop codec ul
      another MPEG-2 long gop codec ul
      add information about codec uls
      use parenthesis around value
      parse aes3 audio descriptor
      simplify using function table
      skip run-in sequence during probe
      avoid reading beyond packet size
      negative to_pad won't provoke infinite loop
      add some useful debug info
      move common code from mov.c and movenc.c to isom.c
      more codecs uls
      fix ending null entry in table
      do not set unneeded fourcc
      remove unneeded audio fourcc as well
      resolve strong refs in parse_structural_metadata since objects may not be ordered, use object oriented approach
      add codec detection based on essence container ul
      attempt to demux clip wrapped content
      remove duplicate code
      10l, check container_ul existence
      again, disable DEBUG
      skip run-in sequence before reading header
      parse SONY hidden MPEG-4 extradata, fix C0023S01.mxf
      revert r5909, dont use non constant static variable, breaks multithreaded apps
      sample size is 1 if bits per sample is 8 and mono channel
      always write duration and file size, fix seeking, progress bar is now active
      r5939 (flvenc.c) rotozoom regression update
      vbr compression id is only used in mov
      use ternary
      mp4, 3gp field is reserved and value is 2
      sampleSize field to 16, reserved for mp4 and 3gp
      only set fields for mov, reserved for 3gp, mp4
      indention
      update regression tests for patches r5945-r5950 based on patch by Francois Draperi <ffmpeg at hp.com>
      detailed infos about d263 atom
      use correct tag for h263 in mov
      parse mpeg frame to get pict type and closed gop flag
      enable 'NONE' audio fourcc
      fourcc cleanup
      set elst start time to first pts
      update rotozoom regression ref after patches r5945-r5950
      export av_interleave_packet_per_dts
      update documentation
      use C99 standard constant, thanks to Foxy Shadis
      use packet dts as correct media field number and use av_interleave_pkt_per_dts
      add fullname
      init buffer
      gxf regression tests
      rename closed_gop to first_gop_closed
      set correct first gop closed value
      Alignment of the LT window, segfault fix patch by Baptiste Coudurier.
      support vdva fourcc (dv + dv audio in mov)
      put dv demuxing code around ifdef
      put dv demuxing code around ifdef
      dont set width/height for h261/h263/mpeg4 since they are oftenly wrong, let decoder init them
      move amr parameters adjusting like other codecs
      wide atom is only defined in mov, use free for mp4/3gp
      add codec id and tag for JPEG 2000
      correct uid, and enable JPEG 2000 detection
      fix mac3audio.mov, pcm_s8 and pcm_u8 can only have sample size of 1 and only if mono, dont rely on bits per sample which can be wrong
      move track timescale/sampleDuration init to write_header and return error if not set
      workaround for files, mainly d-cinema, with 0 sized tags especially UMID in source clip, 377M B.10,13 clearly mentions that this is non standard though
      remove unused, uninitialized free format frame size
      change codec id if sample size field is set to 24 in stsd, fix Sony-mx5p.mov
      add lpcm fourcc
      move sample size adjusting code after audio stsd v2 parsing to let v2 set correct channel number, fix Sony-hdv2.mov
      remove gcc warnings about void * used in arithmetic
      remove useless casts
      shut up gcc about incompatible pointer type
      add string param to PRINT_KEY for debug
      use a data definition uls table to fetch codec type
      add another data definition ul
      rudimentary binary seek
      return 0 as stream index if only one stream, this is completely non standard, fix Cars_TL4IO6_239_DEXX_MPEG_TDC_072006.wav.mxf
      fix audio for mog_op1a_prompeg_mpeg.mxf
      mjpeg 4:2:2 encoding support
      change extradata type to uint8_t *
      remove useless casts
      support D-10, XDCAM, fix ebu_small_d10_50_audio_resampling_problem.mxf, xdcam-pal-d10-imx50.mxf
      quiet gcc about enum value not handled in switch
      dont seek back if no sync to let eof happen
      rewind completely if sample time is negative
      assign myself to aiff maintainer
      fix d-10 mapping detection
      simplify by not writing unneeded codec name to fix segfault when stream copy
      use av_get_bits_per_sample, compute block align if not set
      use av_get_bits_per_sample
      fix bit rate
      support for Delphine Software .cin files demuxing / audio and video decoding, by Gregory Montoir %cyx A users P sourceforge P net%
      update doc, changelog, and increase libavcodec version for .cin files support
      update GXF status and add description
      add MXF description
      support for Tiertex .seq files demuxing/video decoding, by Gregory Montoir %cyx A users P sourceforge P net%
      skip run in sequence, fix C0023S01.mxf
      skip packet if no stream index is found, fix proxy_pal030926.mxf
      avoid overwriting pkt
      simplify pcm read seek, use av_get_bits_per_sample
      use AVFormatContext to log
      warn about zlib support needed if file has compressed moov atom
      disable annoying debug messages
      change gif demuxer to gif decoder
      add gif support
      indention
      printf -> dprintf
      remove unused variables
      use avctx->get_buffer
      removed useless variables
      activate gif decoder, seems to work fine on showdown2.mov
      indention
      remove mov_read_leaf and remove useless reading function table, mov_read_default will skip atom with no parsing function anyway
      fflags can/will be used for muxing too
      remove gcc warning about void * used in arithmetic
      mov vfr muxing
      simplify
      remove useless mp4 field
      clean, always write major brand for debug
      change gif muxer to simple gif encoder
      kill AVImageFormat
      reintroduce isom field and read version 1 audio stsd only if file is not isom, fix testclip_wrong_version_1_audio_stsd.mp4
      indention
      do not overwrite codec type if already known, fix conflict between 'raw ' fourcc used for audio and video, ae002_wordup.mov
      update docs to reflect AVImageFormat removing
      fix decoding position
      remove useless line variable
      use AVRational and ff_frame_rate_tab for frame_rate
      move check for output file streams num in av_encode
      move AVCodecParser prototypes and definitions to parser.h, and move mpegvideo parser to mpeg12.c
      rename ff_mpeg1_find_frame_end to mpeg1_find_frame_end and make it static
      include parser.h instead of mpegvideo.h
      add codec_tag and codec_stream_tag to MpegEncContext to not alter avctx values
      add option to set mpeg1/2 gop timecode start, and drop frame flag timecode flag
      typo
      support m1a hdlr mpeg elementary audio handler, fix Po-Slow_Death_64K.mov
      get rid of CODEC_ID_MPEG4AAC after next version bump, and change it to CODEC_ID_AAC where used
      report error when local tag has 0 size
      add low delay forcing option for mpeg1/2 encoding
      partial support for vobsub in mp4, used by nero, unsupported-embedded-subs-2.mp4
      add 422P at HL I-only universal label
      add movenc.c maintainers
      make option field const
      throw error if unsupported resolution is used with h261
      print error when unsupported sample rate is used with swf
      parse video frame headers to get key flag
      typo
      fix sigfpe with intra only encoding
      fix bitstream to be correctly decoded by quicktime
      readjust dts if negative
      move source_package declaration in the loop and reset it each iteration
      add bytestream big endian 16/32 writing functions
      mjpeg a dump header bitstream filter, modifies bitstream to be decoded by quicktime
      probe with some success image files not containing number pattern but having recognized image extension
      fix memleak, free stream codec allocated by av_new_stream
      cosmetics, remove braces, remove spaces just inside parens
      print a more correct message when container frame rate differs from codec frame rate
      check codec frame size directly and only for audio
      remove useless sampleDuration
      fix compilation, use the standard INT64_C macro
      dont segv is ap->device is not set, display error message
      documentation for X11 grabbing
      always print message when error, AV_LOG_DEBUG -> AV_LOG_ERROR
      change mpeg2 intra vlc test to advanced one, add qprd trell and b frames
      bt 0 is invalid, setting min to 1
      dont skip too many bytes if tag is TAG_STREAMHEAD2 and mp3
      simplify
      priv_data is allocated/freed internally
      humm, actually do what's in the doxy
      10l, only allocate priv_data if not already done
      remove #if obsoleted after last major version bump
      removed unused av_frac_set
      change regression tests for swf, use same as flv, swf muxer only supports mp3
      change regression tests for swf, use same as flv, swf muxer only supports mp3
      add official VC1 fourcc
      print error when decoding fails
      AnyType is needed, descriptor_ref can reference Descriptor or MultipleDescriptor
      always export videostreams
      I will maintain swf.c
      always export audiostreams
      use ch_id as AVStream id and simplify
      swf can contain only one audio track, simplify
      simplify
      enable mjpeg in swf
      priv_data is allocated internally
      simplify
      seek back at the end of file after updating header
      follow michael suggestion and simplify code at object level
      indention
      remove warning, key is const
      merge multiple descriptor parsing with generic one
      remove useless preface parsing for now, will reintroduce it if needed
      remove useless essence container data sets for now
      cosmetics, function -> metadata
      add png fourcc
      add msrle fourcc, quicktime_bmp.mov
      do not override rc_buffer_occupancy if set
      AVOption is now const, removes a few warnings
      add sanity checks
      make some fields unsigned
      remove obsolete comment
      remove useless field
      replace strcpy by memcpy
      stop if current sample is higher than sample count
      make vars unsigned
      remove some printf warnings
      add hint for closed gop and sc threshold error message
      TODO update
      do not force output audio channels
      add waveformatextensible support
      remove void * used in arithmetic warnings
      correctly set profile/level for hd
      fix regression tests due to pixel format name changes
      make key parameter const
      use aiff when we can, some players only support aiff
      indentation
      100l
      fix dts adjusting
      use avcodec_decode_audio2
      free faac extradata
      enable rtp payload size AVOption for video encoding
      get rid of rtp_mode field after next major bump
      define MPEG4AAC id to AAC, and get rid of it and MP3LAME id with next major bump
      get rid of CODEC_FLAG_H263P_AIC with next major bump
      quicktime 'raw ' support
      remove ps option and rename rtp_payload_size AVOption to ps
      check for any seeking failure during probe
      remove nonsense time base code, use swf ms per frame, only video frames contains pts
      factorize
      don't parse whole swf file in header, can now play ffserver swf streams
      update
      fix adpcm swf decoding
      support for adpcm swf, pcm
      fix typo
      use generic atom to extradata reading function
      support for mjpeg bottom field first in mov
      avid imx support
      partial avid meridien uncompressed support
      cosmetics, cleanup tables, add pcm 24/32 le tags
      use common tables for mov muxer and demuxer
      deobfuscate, clearly split mov case
      fix pcm u/s8 muxing
      add pcm s8 tag
      dummy support for mpeg2 non linear quant
      add R/WB24 functions
      add bytestream_get/put_be24 functions
      imx dump header bitstream filter, modifies bitstream to fit in mov and be decoded by final cut pro decoder
      change myself to mov.c maintainer in favor of Francois, who agreed
      simplify swf muxer
      cleanup useless fields
      more explicit message
      use AVFormatContext with av_log
      fix drop frame timecode flag
      fix pts handling
      seems safer to set pts timebase to sample rate, fix some mp3
      add message saying that pgmyuv format is deprecated
      remove codec_info_duration/nb_frames from AVStream
      remove wrong avid imx ntsc fourcc
      h264 complete frames parsing support
      reindent after last commit
      dnxhd decoder
      add dnxhd fourcc
      10l, fix debug, dprintf does not have log level
      treat frame_size > 1 as compressed audio
      set audio frame size based on stts
      cosmetics
      cosmetics, sample_size_v1 -> bytes_per_frame / samples_per_frame
      add specifications reference comments
      prefer MP3 id as mentioned in avcodec.h line 210
      add some codec object type ids
      sort
      cosmectics, mov -> mp4
      break if eob is reached to avoid reading one too much byte
      fix indentation
      simplify
      add some length checks
      actually return bytes read
      return error if loop has ended before decoding image
      actually return bytes read
      dont fail immediately when a somehow broken track is detected, some tracks might be good, fix mi2_vorbis51.mp4
      use MKTAG
      prefer MP3 id
      activate parser on MP3 id, fix [A-Destiny]_Konjiki_no_Gash_Bell_-_65_[71EE362C].mp4
      pass correct buffer size to lzw decode init
      remove TODO, its done now
      increment subminor version due to r8502 and r8504, those revisions make dnxhd and gif decoders in line with libavcodec API
      add IJPG fourcc, to play m_ijpg.avi
      fix decoding of big audio packets (48k 16bit 2 channels), needed size is related to samples which is short * while len passed to decode_audio2 is related to pkt->data which is uint8_t *
      revert r8538 (set bit rate for asf format) until regression tests issue is stated
      add hdv1 fourcc
      break if atom size is more than size left in container atom, fix shooter.mov
      unset sample count to disable track when is broken
      move atom size check before parsing function search
      typo, fix -fs
      1000l, revert r8547, im too stupid
      add OPT_INT64 option
      change -fs option to use OPT_INT64
      change -fs to use bytes instead of kbytes
      print error message
      add AVRn fourcc
      add D-10 PAL 40 ul
      add D-10 PAL 40 audio essence container ul
      remove MpegEncContext, patch by baptiste
      use INT64_MAX
      remove broken ctab code
      clean and simplify esds reading function
      cosmetics: mov_mp4 -> mp4
      long -> int
      done in r8635 by Xiaohui Sun
      fix compilation
      correct comment about seek function
      add 'wide' reversed tag in probe, detect broken xdcam files xdcam_hd_1080i60.mov
      add xdv2 xdcam hd tag, decode xdcam_hd_1080i60.mov
      remove now obsolete and no more true comments
      use dprintf with AVFormatContext and simplify
      cosmectics, use consistant and homogeneous type names for atoms
      remove no more true comment
      fix oss audio playback, broken since r7983, pass AVFormatContext filename to audio_open since audio_device when NULL is no more set to default value
      fix AVStream documentation: clarify start_time, start_time and duration are actually in stream base
      dmb1 fourcc, mjpeg_opendml.mov
      fix ctts index computation when seeking, check must be done against next ctts sample, thanks to Uoti
      simplify, no need for a function
      print and return error if output is non seekable
      cosmetics, reorder and classify uids
      match uids indenpendantly of the version byte
      sync to mxf klv key before trying to read klv packet
      do not print error message when klv_read_packet fails
      no need to seek on specific key anymore, since klv_read_packet will sync, tested on encrypted files
      remove now useless sync key
      set pkt pos
      typo
      remove duplicate entry since now we ignore version byte
      skip no more needed on error
      remove unused EssenceContainerData
      cosmetics remove struct name with typedef
      warn user if video codec timebase is very high
      add MMES fourcc, matrox mpeg2 in avi files
      split mpeg ps and variants muxer and demuxer, I'll clean more in a few minutes, lpcm freq tab is left static const in mpeg.h for now until we have more code in common
      remove now useless ifdef
      use allformats.h
      support files with COMM chunk after SSND, fix invalid_nocommon.aiff which is spec compliant
      according to specs, offset is before blocksize
      this is named BlockSize in specs
      compute absolute data offset after reading blocksize
      round timestamps up, k2 broadcast server seems to need it
      Smartjog granted permission to relicense as LGPL 2.1
      Calculate the correct blocksize for flash adpcm. Patch by Baptiste
      Set the step index correctly for flash adpcm. Patch by Baptiste
      treat NONE as U8 to fix qtNONE/000_0032.MOV, this should not any have side effect since if bps is 16 in stsd, codec will be changed to s16be, just yell if this breaks any sample
      use buf_ptr to correctly decode second field
      set block align to stsd audio v2 bytes per frame for adpcm ms and ima wav, fix surge-2-16-L-ms11.mov and surge-2-16-L-ms02.mov
      should be more correct
      add coding unit size for interlaced decoding
      10l, use correct cid table
      no need to duplicate cid table vars in context
      cosmetic
      dnxhd 185 interlaced support
      indentation
      dnxhd 120 progressive support
      move dnxhd data tables to separate file
      remove useless raw_write_header
      do not write id3v2 header when writing .mp2 files
      add streams on the fly, swf is a streaming format and has no real header, correctly detect audio in RamboMJPEGAVP6_112K.swf now
      cosmetics, indentation
      cosmetics, braces/parenthesis style
      remove now useless vars
      fix decoding of adpcm swf big frames, fix RamboMJPEGAVP6_1Mbps.swf
      cosmetic, indentation
      take care to use video track field number as duration
      dnxhd 120 interlaced support
      support swink created files which have soi/eoi broken tags reversed
      stop parsing udta if size is wrong/garbage, fix issue 154, fix RQ004F14.MOV
      extract correct d10 aes3 audio channels number
      add D-10 NTSC support
      HDYC fourcc, sample hdyc/Test2.avi
      export DCT_common_init and add ff_prefix
      export convert_matrix and add ff_ prefix
      DNxHD (SMPTE VC-3) encoder
      fix 10 bit per comp detection
      10l, check 'i' validity before using it
      merge
      preliminary 10 bit depth decoding support, still miss generic api to export picture, working on it
      typo weigth->weight
      append extradata atoms when parsing, fix OLOCOONS_O3.mov
      dnxhd 36mbit support
      better lambda search method
      indentation
      dont write udta atom for 3gp/3g2 for now, current written atoms (ilst) are itunes specific
      100l, fix dump_format, declare buf in dump_stream_format, sizeof(buf) is 4 because buf is char* since r10818
      fix issue 225, instead of stoping when wrong atom size is found,
      use FFMIN
      rename av_crc04C11DB7_update to ff_crc04C11DB7_update and move it to aviobuf.c so it can be reused by other (de)muxers
      ogg muxer
      files have been renamed
      ogg seek regression ref
      remove now useless libogg support
      fix corner case when qscale 1 bits < frame bits but max bits with worst padding > frame bits
      cosmetics, encode_variance -> encode_fast
      dnxhd 720p encoding and decoding support
      indentation
      add const to remove warnings, ffmpeg.c:2892: warning: initialization discards qualifiers from pointer target type
      fix dnxhd 720p in mov for quicktime player
      return error if malloc failed, found by takis, fix issue 286
      choose next sample by sample position
      return error when url_fseek could not read until desired offset in streamed mode
      support streamed mov/mp4
      cosmetics
      remove useless seek
      save pointer to parse table, it is unlikely to change
      remove useless declaration
      cosmetics: func -> parse, remove useless parenthesis
      remove duplicate field
      do not retain useless pointers to avstream priv_data, use it directly
      remove now useless fields
      useless 0L here
      avoid infinite loop if pixel format conversion does not exist
      add bitrate helper to choose all dnxhd variants
      supports glbl atom containing generic extradata for all codecs
      use generic 'glbl' atom if extradata is present but no specific method is known
      fix indentation
      reset manually selected codecs at the end of opt_input_file, to avoid propagating selection to output file
      cosmetics, split overly long lines
      check av_new_stream return value
      check av_new_stream return value
      simplify by checking min matching len
      add some dv container uls
      check malloc return
      support dynamically allocated local tags, used by encrypted files
      remove unused crypto context uid
      cosmetics
      cosmetics: read_metadata -> read
      simplify and detect better non frame wrapped mappings
      simplify with matching len
      check if read child returned error
      correctly handle sony mpeg4 extradata
      indentation
      init uid to 0
      reorder alphabetically
      unset extension, so code path, and guess format do not choose
      mention avm2 (flash 9) muxer in documentation and changelog
      remove non standard object type ids, if you find samples, report them
      this one is not standardized either
      comment vorbis objectype id
      document tag field
      reorder object types id by muxing preference
      write esds atom when mpeg-4 systems tags are used
      correctly and only set standard stsd fourcc for mp4/3gp
      indentation
      finally set codec tags for mov/mp4/3gp muxers, this should not break anything, if it breaks anything, yell
      spoke too fast, fix dv fourcc in mov accordingly if strict
      cosmetics, vertically align
      cosmetics, split long lines, remove useless braces, merge comments and if
      warning is only valid for mov
      use context in av_log
      fix prototypes, remove warning: i386/dsputil_mmx.c:3594: warning: assignment from incompatible pointer type
      add parenthesis, fix warning: i386/dsputil_mmx.c:2618: warning: suggest parentheses around arithmetic in operand of |
      add const, fix warning: cook.c:276: warning: passing argument 2 of 'maybe_reformat_buffer32' discards qualifiers from pointer target type
      cast to dest type, fix warning imgconvert.c:1958: warning: assignment from incompatible pointer type
      cast to dest type, fix warning: 4xm.c:304: warning: initialization from incompatible pointer type
      init to 0, fix warning: h263.c:4730: warning: 'dc_pred_dir' may be used uninitialized in this function
      fix indentation
      cast to correct type, fix warning: apedec.c:859: warning: passing argument 1 of 's->dsp.bswap_buf' from incompatible pointer type
      typo, also fix warning: asf.c:90: warning: passing argument 2 of 'get_buffer' from incompatible pointer type
      add parenthesis, fix warning: mpegenc.c:303: warning: suggest parentheses around && within ||
      init to 0, fix warning: mpegenc.c:1022: warning: 'avail_space' may be used uninitialized in this function
      typo, typedef uint8_t GUID[16] in asf.h, and fix warning: asf-enc.c:202: warning: passing argument 2 of 'put_buffer' from incompatible pointer type
      add parenthesis, fix warning: qtrleenc.c:257: warning: suggest parentheses around && within ||
      Add const qualifier, fix warnings:
      add parenthesis, fix warning: snow.c:2413: warning: suggest parentheses around && within ||
      revert r11677, it is considered speed critical
      support rawvideo in mov
      stupid typo
      fix field order export
      Do not print quality when stream copy,
      add const
      const
      const
      it seems ogg requires granule to be -1 on unfinished packets
      Fix issue 338:
      seek regression ref update for ogg
      cosmetics, remove whitespaces before and after parenthesis
      more correct and simpler
      indentation
      fcp imx 40 fourccs
      use correct demux mechanism when audio compression id is -2, dv audio does not matter, demux partially qt_dv_pal_test.mov which does NOT use cid -2
      split chunks in the case of big compressed audio frames,
      cosmetics, indentation, braces
      remove useless braces
      remove obsolete comments
      return error when dv audio extraction fails
      remove messy and not always correct chunk size workaround, use correct values instead
      remove now useless code
      warn use if edit list is not starting at 0, a/v desync might occur
      10l, forgot alaw/mulaw mono, use bits_per_sample in fact,
      honor stsd v1 first, fix mace surge-2-8-MAC3.mov, beware of mulaw,alaw
      ima4 in aiff support
      fix duration for ima4
      set bps to uncompressed original sound data for compressed audio
      remove useless tags, codec_get_id matches independently and qt uses lower case tags
      cosmetics
      remove useless casts
      indentation
      remove useless close function
      support QDM2 in aiff, QDM2.aif
      10l, protect malloc overflow
      fail if extradata cannot be allocated
      indentation
      protect malloc overflow
      fix mace duration
      move parameters reset after setting stream parameters, will need those infos
      rescale frame size, fix i.3g2
      indentation
      remove useless braces
      support mov reference files, ref.mov/ref.m2v/ref.wav and detect BrianCox.mov
      support keyframes indexes starting at 0, broken files created by omneon, fix ref.mov seeking
      fix possible overflow with memmove
      return error if malloc fails
      free in case of multiple type == 2
      add weird first atom to probe, finally fix bartjones.mov
      set codec bps to grayscale bits, fix 256grey.mov
      code is ok for 4bpp in mov too, fix 16grey.mov
      1000l parenthesis
      correctly compute bit rate for compressed audio, should fix seeking in aiff
      set block align for mace3/6
      fix interlaced decoding, original patch from reimar
      remove redundant SwScaler text since av_log uses AVClass context
      change sws_format_name to return const char*, supress many warnings
      support AVDISCARD_ALL
      do not reset duration_error for all streams, only reset current stream, fix wrong_fps.gxf
      do not display par and dar if not available
      indentation
      do not fail badly if aspect ratio is 0/forbidden, fix issue 369
      log aspect ratio forbidden value, and fail if error resilience is high
      remove useless img_write_trailer func
      useless null
      ipod/iphone compatible mp4 muxer
      fucked up alphabetical order again, sorry
      simplify, flags is 24bits according to specs
      just skip klv packet, when no corresponding stream is found
      cosmetics, remove superflous comments
      correctly set avctx->frame_size for mp3on4 and mp3adu
      omg, fix compilation
      move specific end treatment when trak is detected, will be needed for fragments
      move mov_build_index before mov_read_trak to avoid useless declaration
      free track data asap
      remove useless debug print since now mov_build_index will do
      cosmetics, move case up
      fix mpeg2 muxing, replace seq header if contained in extradata and key frame does not contain it
      indentation
      cannot free AVStream like this, and return ENOMEM
      return ENOMEM
      update description
      atoms are not mp4 only, and yes probe is sufficient
      remove ms audio tags from table, they will be chosen or detected with riff table
      simplify, and only stop parsing when non streamable
      cosmetics, split or move too long lines
      mp4 fragments support
      support fragments when non streamable
      yet again another hdv fourcc
      sample size to 1 might happen for data, fix GMCMidiaASX.mov, fallback is better than failing
      Only use chunk demuxing for old uncompressed audio mechanism specified by stts.
      according to specs, only color table id set to 0, have it in stsd, fix 4colors.mov
      print stream index in error message
      cosmetics, useless braces
      print essence path in error message
      gsm in mov, demux surge-1-16-B-agsm.mov
      change rle encoder to count up to 127, sgi does not support 128
      check frag track id before setting st
      support mp4 fragmented files without any samples in the moov atom
      prevent writing empty stss atom
      cosmetics, remove leading whitespaces
      missed one
      homogenous regression tests output for image formats
      mpeg4audio common code
      update ogg seek reg tests because of r12630
      use mpeg4audio common code in aac parser
      special mpeg2 handling is no more needed
      better error message
      remove cast
      set alac frame size from extradata
      support for alac in .mov
      qcelp fourcc
      demux qcelp, so this can work with soc decoder and stream copy
      cosmetics: remove useless whitespaces and put braces correctly
      set wrap bits to 32bit to actually use it
      pts are unsigned according to specs, fix negative pts when 32bit pts are used
      set gsm default frame size and bytes per frame, needed by old qt format, fix sample-gsm-8000.mov
      cosmetics, remove whitespaces in tags
      align and remove useless braces
      lower frame size to split amr/qcelp/gsm frames
      parser should not be needed anymore, merge cases and set frame size for amr too
      move vars decls where vars are used
      cosmetics
      error is -1
      update doc
      parse mpeg4audio config to correctly detect mp3on4, fix iso mp3on4 reference files
      cosmetics
      mov demuxer depends on mpeg4audio code now
      set channels and sample rate from mpeg4audio config
      fix adpcm ima qt decoding, channel at init is 0, correct stereo out since samples += avctx->channels
      remove ima qt decoding hack, decode whole frame at once
      remove unused var
      remove no more needed field
      identation
      move vars decls where vars are used
      remove unneeded variable
      yes it is true for mp4
      10l, alphabetical order
      add const
      values can fit uint8_t
      return error if buf_size is too small
      add FFMIN3
      simplify using FFMIN3
      fsize is 12 bits according to specs
      correctly compute out_size and samples number
      simplify
      use mpeg4audio common code
      mp3on4 decoder now depends on mpeg4audio code
      remove old draft last case not present anymore in iso specs
      fix 1 frame config decoding
      correctly patch syncword for samples rates < 16000, decoder now fully support all iso ref files
      move MP3On4DecodeContext def near the code and under ifdef
      simplify: remove useless fields and vars
      simplify
      simplify: merge loops
      cosmetics: merge decls
      merge init with declaration
      break if error in header so it can output already decoded frames
      set demuxers .value and use common audio_read_header function
      export dnxhd encoded picture quality
      cosmetics, remove space before opening parenthesis
      do not fail if correct is not recognized
      try to demux even if codec is not recognized, should demux M1F1-float32C-AFsp.aif and fix #454
      demux all stsd ids if codec is the same, fix premature_mov_ending.mov, closes #451
      warn about multiple fourcc
      only print debug info when sample is actually added
      split swf de/muxer
      remove useless close function
      remove useless/obsolete comments
      do not forget to free packet, fix memleak
      open logfile in append mode, don't overwrite it
      check pkt stream index before returning packet, prevent segfault
      no need to sleep to hide crashes
      simplify
      simplify
      use duration from pkt
      10l, fix compilation
      remove useless var
      flv use 32 bit pts
      in flv this field is dts finally
      add new h264 and aac flv ids
      simplify, remove useless vars
      latest libavc svn requires at least this value for bit_rate_tolerance, not using av_q2d because of float
      better error message
      simplify, use ffserver launch directive
      fix feed path according to regression script
      do not loop input, useless
      better error message
      use short option
      set correct duration when using b frames
      reenable test, data is available in this case
      split ffm de/muxer
      remove unneeded include
      move get_pts function to avoid useless declaration
      move DEBUG_SEEK definition before get_pts since func uses it
      move ffm_seek1 before get_pts function since func uses it
      move ffserver specific functions up and only include unistd.h in this case
      return error if len is negative, prevent segfault
      return error if frame_offset is negative, prevent segfault
      cosmetics, remove useless dot
      cosmetics, add demuxer word
      use dprintf
      use av_hex_dump_log
      replace printf by av_log
      update regression ref due to r13412
      aac chan config is 0 if bitstream contains program_config_element
      aac and h264 in flv muxing
      aac and h264 in flv demuxing
      register myself as ffm maintainer
      typo
      remove const qualifier, removes warning:
      use av_fifo_generic_write, old func is deprecated
      cosmetics, remove useless cases and braces
      simplify
      simplify
      simplify and use version instead
      merge init and declaration
      cosmetics
      use AVFifoBuffer
      simplify, use pointer to codec context in struct instead of only id
      simplify, use pointer to codec context in struct instead of only id
      cosmetics
      use internal video frame number
      remove dummy coded frame, no muxer should use it anyway
      use av_fifo_generic_write, old func is deprecated
      set avctx time_base in dv decoder
      check if extradata comes from mp4 and assume bitsteam is already formated, should fix streamcopy
      this should be valid for audio too
      reindent
      warn if pts is not set, prevent writing negative/weird cts, happens when using genpts and streamcopy
      init h, fixes CID54 RUN2
      check if stream has been set, fixes CID15 RUN2
      check that stream is set, fixes CID13 RUN2
      reindent
      check that redirect is set, fixes CID13 RUN2
      reindent
      use av_parse_video_frame_rate
      cosmetics, space before and after parenthesis, weird indentation
      cosmetics, move declaration where it is used, remove now useless block
      cosmetics, reindent
      set is_streamed before writing data to output file
      log some errors
      take over ffserver maintenance
      fix timestamps rescaling, index in source format might be different
      wake up connections to stop waiting when feed is being closed
      log error when writing frame to output
      sleep seems needed to wait for feed to be setup
      update regressions tests, should be stable now
      set got_key_frame when packet is video, fix first video packet with index 0 and 2 streams
      fix got_key_frame when stream is audio
      2nd try fix got_key_frame when video+audio
      add \n to error message
      give context to av_log
      memcpy considering output width, not src linesize, fix segv with av_picture_pad
      remove useless casts
      add const, suppress warnings
      add const, suppress warnings
      remove useless casts
      add const, suppress warnings
      cast to correct type, suppress warnings
      add const, suppress warnings
      cast to correct type, suppress warnings
      cast to correct type, suppress warnings
      restore needed cast to correct type with const
      add context_to_name func for logging
      return if bitrate is not specified or too low
      optimize, merge offset bits in vlc code
      cosmetics, remove table_ prefix
      rename last_frame_offset to next_frame_offset, due to r13615
      print correct mpeg-2 broken_link value, and print closed_gop
      fix continuity counter to start at 0
      add pretty common .m2t as ts extension
      update regression tests for ts due to r13641
      cleanup, remove useless parentheses
      x264 interlaced encoding option
      add esa/dia compat me options and new tesa for libx264
      cosmetics, rename first to header, more explicit name
      cosmetics, rename first to header, more explicit name
      Reset reading state after successful seek.
      fix pts handling in ffm
      update ffm seek tests now that timestamps are fixed
      remove ugly ffm_nopts hack, use AVFormatContext->timestamp
      add AVOption support in ffserver configuration file
      remove unneeded ffmpeg commandline switches, and enable quiet mode
      pass extradata through ffm
      uniformize trellis quant option
      revert wrong part of r13722 commit
      enable global_header flag for audio and video encoding, was wrongly part of commit r13722
      pass dts and pts through ffm, should fix streaming b frames
      remove now unneeded vars
      cosmetics
      remove obsolete requirements, quicktime 6 is from 2002, not sure if muxing scheme still fits
      indent
      10l, add parentheses
      rework udta handling, if any track set bitexact, don't write any metadata
      indent
      m4v/m4a are usual itunes/ipod/iphone extensions
      remove useless wrapper func
      support alac in ipod m4a format, tested on quicktime
      cosmetics
      cosmetics
      factorize psp metadata
      simplify
      write usual brand for ipod format
      move utf8 code before udta func, will be used by 3gp tags
      support for 3gp metadata
      indent
      set log debug when avformat debug is set
      don't write double sets of metadata in mov, only write usual ones
      cosmetics, reindent
      enable stream copy with ffserver, to be able to chain
      put avc1 brand in compat list when h264 is present
      fix warning, write_ftyp return int now
      simplify ipod brand selection
      change 3gp/3g2 brands if h264 is present
      simplify
      invert mode_mov condition
      set correct ftyp minor version for 3gp/2
      ffm has no per stream priv_data anymore
      remove useless close func
      remove useless close func
      remove useless close funcs
      factorize read_header failure freeing code
      uniformize AVStream->priv_data freeing in av_close_input_stream
      add static
      upgrade bandwidth to uint64_t
      upgrade bandwidth config to use int64_t, raise limit
      use http_log when log file is setup
      log accept error
      rescale duration when outputting packet
      always use stdout to log when running in debug mode
      use http_log when log file is setup
      remove useless init, logfilename is static
      change bandwidth to unsigned
      set default preload and max_delay for muxer to avoid mpeg ps muxer to spit many underflow errors, use ffmpeg default values
      return 1 when error during write frame, state changed
      update reg ref due to r13851
      free c->pb_buffer before overwriting it, fix memleak
      use av_seek_frame to flush demxu pending packets, fixes seeking, should have been done when av_find_stream_info was added
      do not reset ffm pts, this avoids setting last pts to 0 and screwing up seeking
      fix open return check
      http or rtsp only support
      reindent
      merge declaration and init
      remove useless init
      print date in http_log
      use ctime1
      no need to print date anymore since done in http_log
      add callback for av_log
      log to stderr
      enable feeder threads
      cosmetics, reindent
      do not allocate st->codec, it will be overwritten later; fix memleak
      use av_open_input_stream
      free pb
      remove return 1, we don't change state, we are still sending data
      check av_open_input_stream error
      move code after possible failure, so no need to free
      move code setting output parameters after possible failure
      update seek reg ref due to r13856
      remove pts set hack, this is no more correct
      store dts in ffm headers
      cosmetics, pts->dts
      chinese avs demuxing support, demux AVSFileFormat.mp4
      no need to print date anymore since done in http_log
      simplify
      fix source stream, source_index is refering to fmt_in, needs testing
      simplify, use ist
      10l, move init after possible goto
      fix compilation, this should be better
      10l, do not use ctx before being set, fix segv
      cosmetics, braces placement
      fix segv, fmt_ctx is not set for rtsp
      copy extradata if present
      reindent
      log when feed is already being received
      remove useless init
      move logfile setup right after config parsing
      use http_log now that logfile is setup
      make av_find_stream_info find frame_size for aac too
      revert r13867, - reference stdout, I was drunk
      give context to av_log
      clarify error message about codec tag
      unset Track sets, relying in AVStream->priv_data and will be freed in av_close_input_file
      cosmetics, align vertically
      cosmetics
      set frame_size in libfaad decoder
      include track id in error message
      print AVCodecContext address instead of AVClass, better when multiple instances of same codecs are used, based on r14237
      Return max score when ftyp atom is encountered.
      cosmetics, remove space
      update swf video frame number when muxing done, fix #439
      cosmetics, remove space
      add ogv extension
      update hdv fourcc descriptions
      more hdv fourccs
      add a bunch of new xdcam hd422 fourccs
      change assert test due to the new reducing of time_base in av_set_pts_info, fix #561
      skip flv video info / command frame packets, fix issue #546
      check url_fseek return value, update seek reg tests
      check url_fseek return value, update seek reg tests
      check url_fseek return value
      Only reset buffer state if seeking is successful; update seek reg ref.
      fix mpeg 4:2:2 vis_qp and vis_mb_type debug, fix issue #471
      move version specific parsing code before codec specific code, will be needed
      more complete audio stsd v2
      keep original codec/fourcc endianness, fix XDCAMHD.mov
      support big endian lpcm with audio stsd v2
      do not fail on qscale==0 if error resilience is high, fix broken quicktime h261, lotr.mov
      full lpcm support in mov audio stsd v2
      fix if/else case
      add mpeg2 4:2:2 regression test
      100l, fix test, ref values were done with mv0, dunno how this slipped in
      add dv 4:1:1 regression test
      rename mxf.c to mxfdec.c
      remove mxf.c, previously copied to mxfdec.c
      fix 4:2:2 psnr
      Prevent dts generation code to be executed when delay is > MAX_REORDER_DELAY,
      increase MAX_REORDER_DELAY and pts_buffer size to 16, max for h264 atm
      import ok'd hunks for mxf muxer
      enable fl32/64 le/be in mov with stsd audio v2
      I'll maintain mxf muxer
      simplify mpeg12 avctx->pix_fmt set
      fix version string ul
      cosmetics
      cosmetics, typos, case
      speex in flv demuxing
      remove vorbis dec hack, decoding will be done anyway to get frame size
      simplify size handling code
      factorize packet skip code
      skip empty flv data packets, fix issue #602
      make has_codec_parameters not returning true when sample_fmt is not set
      remove now useless casts
      timestamps in flv are 32bits _signed_ according to flash player, specs are wrong
      write strings as utf16be
      use strlen and do not write useless trailing 0 according to specs
      introduce mxf_write_local_tag_utf16 and factorize
      always write required version string local tag
      simplify
      only put version in tag
      simplify, and factorize
      cosmetics
      cosmetics, remove useless braces, move comments where appropriate, remove whitespaces
      cosmetics: header -> generic
      fix typos
      rename variable
      move per track code in mxf_write_header to be able to check for unsupported configuration
      get essence container ul in header and set it per track, check for unsupported codec
      compute essence containers in mxf_write_header, this simplifies the code
      change [0] to *
      factorize
      simplify mxf_write_essence_container_refs
      reindent
      merge declaration and init
      replace sizeof(type) by sizeof *ptr
      cosmetics
      introduce MXFContainerEssencePair to associate essence element key and
      fix track number ul in primer pack
      remove now useless mxf_essence_element_key array
      put back mxf_essence_container_uls in demuxer as static const
      remove now useless track_number_sign
      cosmetics, whitespace
      simplify, pass AVStream directly instead of index
      correctly write codec ul, mpeg-2 needs profile and level to be set
      set duration to distinguished value
      correctly set track duration
      cosmetics, mux_write -> mxf_write
      10l fix mxf_get_essence_container_ul
      always write footer partition according to specs
      header partition is open incomplete for now
      simplify current partition offset
      simplify and correctly rewrite metadata in header partition, mark it closed complete
      remove header_byte_count_offset from context
      fix type in prototype
      merge descriptor write table with essence container uls table and simplify
      store index in table in MXFStreamContext, simplify
      cosmetics, Pair -> Entry
      move mxf_write_package to permit function merge
      remove mxf_build_structural_metadata, simplify
      factorize
      do no write multi descriptor when only one track is present
      copy whole frame in vosData for dnxhd, this code will be used for other codecs too
      check for malloc failure
      make pcm encoder setting bits_per_sample field
      allow mp2 to be muxed into mp4
      correctly mux mpeg2 audio 13818-3 in mp4
      fix fourcc comment
      ac3 muxing support in mov/mp4/m4v
      cosmetics, vertical align
      cosmetics, change values to hex since mp4ra use hex
      cosmetics, move up mpeg1video mp4 objecttype
      dirac muxing support in mp4/mov
      forbid mp3 < 16000hz in anything except mov
      change the logic to mux 13818-3 and 11172-3 in mp4, less code
      cosmetics, remove whitespace
      print error when deinterlacing fails
      set block align and frame size for gsm in aiff
      Fix msgsm in wav, do not set blkalign to frame size, this is wrong,
      add -xerror option, ffmpeg will now exit right when an error occurs
      enable yuv422p to uyvy converter
      simplify function selection code
      simplify bps code by using av_get_bits_per_sample
      do not fail if audio is uncompressed and frame size is not set, just set it
      100l, compute correctly poll_table size when allocating
      set default bps to 16 in wav header
      cleanup, merge snprintf calls
      add missing html start tag to ffserver status html page
      terminate tag tables by CODEC_ID_NONE
      alac : fix case where bits_per_sample is not set.
      write copyright metadata tag
      set block align in av_write_header if not set, fix issue #489
      print stream index in decimal not in hex, easier for debugging
      set bits_per_coded_sample according to new lavc api
      sync to latest x264 git, field was renamed
      revert r14983, value is not sample aspect ratio, it is display aspect ratio
      do not override sample rate for amr nor qcelp, should fix some qcelp files at 16000hz
      only write trailer when state says so, should fix issue #646
      merge amr wb case with amr nb and qcelp
      stop parsing if tag size is wrongly < 8 to avoid infinite loop
      include sys/select.h instead of unistd.h to get select,
      include sys/select.h to get select, according to posix 2001,
      skip invalid audio samples in gxf packets
      cosmetics, remove GXF: in log messages since it is present in context
      only include sys/select.h if present, fix mingw compilation
      include sys/select.h, fix freebsd 5.5 compilation
      sync to latest mingw api, remove unneeded struct definitions
      support negative ctts in some way, unset wrong dts
      cosmetics, remove useless spaces
      set sample rate for amr to allow stream copy since no decoder is present,
      10l, fix amrwb sample rate
      get_pixels_sse2, ~+12% performance compared to mmx
      add state64 field to ParseContext storing last 8 bytes, to be able to check for longer startcodes
      dnxhd codec is intra only
      dnxhd parser
      raw dnxhd de/muxer
      force sample rate to 16khz for speex in flv, fix speexaudio.flv
      simplify sample rate code, flv_set_audio_codec already overrides it for nellymoser 8khz
      check if stream is set, fix potential segv
      cosmetics, indentation
      cosmetics: ffserver -> FFserver
      svq3 watermark decoding support, based on reverse engineering work by chrono
      svq3 watermark code now needs zlib
      fix start position UL
      fix sound essence compression UL
      svq3 decoder may need zlib
      fix umid base value, use revision 5 and generic material type
      add sse2 fdct
      convert fdct_mmx to plain asm
      remove useless mmx.h include
      timestamp is dts in wmv/asf
      simplify metadata reading
      read itunes metadata, code based on Reimar's patch
      set pts in flv demuxer
      decrease atom size, fix broken files missing version/flags
      move decoder initialization in separate function, earlier failure detection
      Implement the fields rc_max_available_vbv_use and
      partial WRAW fourcc support, might need image flipping
      WRAW fourcc
      use dsp clear_block
      use dsp clear_block
      cosmetics
      move structure definitions in their own header
      remove _t in typedef for POSIX compatibility
      parse pasp atom and set sample aspect ratio,
      rename 4x8 to 8x4 to match widthxheight consistently
      Move get_pixels_8x4 before init func to avoid useless forward declaration.
      dnxhd get_pixels_8x4_sym sse2
      check that nb_streams is valid before setting stream, fix crash with jp2 which do not use same structure
      activate jpeg2000 fourcc in mov/mj2
      set alac channels from extradata, fix alac mono in m4a
      cosmetics, remove useless parenthesis and whitespaces
      print more debug information if enabled, including index
      use lower case x and alternate format for hex in printf
      100l, add missing enum type, fix compilation
      only set le if value is 1, fix boom5.mov, issue #770
      fix segv, always set audio_buf to audio_buf1 if decoding fail
      check for negative dts value even if bitstream is already formated,
      reindent
      remove comment, this is not relevant anymore
      cosmetics
      remove warnings about uninitialized vars
      cosmetics
      move ffindex set before mov_read_default so it can be used in functions
      cleanup gif, use BISTREAM_WRITER_LE
      include intreadwrite.h, fix compilation, img2.c use AV_RL32
      compute pcr more accurately, regression tests need update
      set pts and dts of pes packets exactly according to specs
      use delay to avoid negative dts
      only write dts if they differ from pts, according to specs
      update seek regression tests due to ts modifications
      warn if dts < pcr which means TS is invalid
      allow user to specify muxrate
      calculate pat, pmt and sdt exact size and adjust total bitrate
      consider other codecs packets contain only one access unit, not just audio
      add aud nal unit if not present in h264 bitstream
      use delay for subs and dirac too
      return size written in ff_avc_parse_nal_units
      move actual writing before so new size can be taken into account
      export ff_avc_parse_nal_units_buf
      10l, fix prototype
      use ff_avc_parse_nal_units_buf because output size might differ from input size
      support h264 extradata with 0x000001 startcode
      match format in names list separated by commas
      jpeg2k is intra only codec
      add jp2 formater in image2 muxer
      R3D REDCODE demuxer
      fix gxf time code mark out value
      set correct value for starting line
      Do not use avctx->frame_number which might not be set,
      change sample aspect ratio test to be more clear, print old one
      exit with error code if av_encode failed
      fix seek reg tests, Im always forgetting this
      Add const, remove warning:
      print stream sar if present and different than codec sar
      fix endianness of time code, and calculate it in fields, according to specs
      stored height is a multiple of 16 for mpeg2
      reduce display aspect ratio
      write essence container data
      set pts and dts when only pts is present, according to specs
      seek at 0 when timestamp requested is negative
      second try fixing time code in gxf
      split av_interleave_per_dts, to avoid code duplication
      use new ff_interleave_add_packet
      function is static
      correctly pack and interleave pcm samples in mxf
      delay header writing and parse mpeg2 frame
      parse mpeg2 progressive frame flag
      add size param to write_generic_desc
      stored height is stupidly /2 if interlaced
      local tag 3201 is picture essence coding, so do not write it for wav
      thanks to interleaving, audio is now locked
      split wav_desc in generic_sound_desc
      add aes3 pcm wrapping and use it by default
      add frame layout local tag, best effort
      write best effort video line map
      cosmetics, align vertically
      fix doxygen tag
      add my copyright as well
      cosmetics, remove empty lines
      do not use PAL_samples_per_frame in init, to make init independant
      change prototype of mxf_interleave_get_packet to make it compatible
      use st->priv_data to store per stream context
      free audio fifo
      fix decoding 720p50 audio samples
      remove some warnings due to avctx->execute
      return eof/error if only ffm header has been written, should fix #815
      write display height/width tags
      write bitrate tag
      do not write f2 if not interlaced
      adjust size when not writing f2
      write req local tags for wav
      factorize common wav local tags
      write index table and rip, muxer works pretty well now
      fix flush and interleaving by edit units
      fix video line map tag size
      fix temporal offet of ipbb sequences, simplification welcome
      more flexible mapping selection
      24bit pcm support
      mxf does not have timestamps
      cosmetics: whitespaces, empty lines
      be more flexible with frame rate check
      prevent reading more than container atom size, fix broken file broken_by_rev15830.MOV, fix #818
      exactly compute index byte count
      let packet pass through when no timestamps
      use sample rate as audio input time base
      always write mxf time base in descriptors
      fix key frame offset with open gop
      use timestamp as creation date
      use timestamp as modification date
      format timestamp correctly according to specs and set it
      use 512 kag size needed for d10
      declare ff_interleave_add_packet
      write system item
      force first track to be video and simplify
      only use 2 slices for index, one video(vbr) and one audio(cbr)
      offset dts according to edit list, hackish but works, based on patch by Reimar
      extract audio interleaving code from mxf muxer, will be used by gxf and dv
      use new audio interleaving generic code
      free index entries
      remove implicit header inclusions from mxf.h
      reuse MXFCodecUL for data def and simplify
      remove warning about incompatible pointer
      remove useless cast, it does not remove warning, encode prototype must be changed
      Remove 'const' qualifier from variable in av_parse_video_frame_size().
      fix typo in h264dsp_mmx (no effect currently as the function is not used), approved by Dark Shikari on IRC
      fix doxygen
      remove useless return
      rewrite pts also, as comment says
      remove useless field
      cosmetics: sample_to_ctime -> ctts
      cosmetics: sample_to_chunk -> stsc
      document ff_audio_interleave
      Declare ff_interleave_new_audio_packet as static, it is not used
      Move declaration of ff_interleave_add_packet to internal.h.
      remove no more needed comment, declaration is in internal header
      rename ff_audio_interleave to ff_audio_rechunk_interleave
      fix compilation with DEBUG defined, field does not exist anymore
      correctly set multi track operational pattern
      fix primer pack local tags uls
      move up mxf_write_klv_fill and align index, rip and primer pack
      fix header byte count
      write component depth and horizontal subsampling best effort local tags
      cosmetics, associate mxf_write_klv_fill to aligned elements
      10l, edit unit offset is after klv fill
      retrieve flags from mxf_parse_mpeg2_frame using a param
      remove check, mxf muxer does not have track number limitation anymore
      compute index byte count in mxf_write_partition
      write body partitions
      fix timecode value
      fix partition offset when only one stream present
      fix body offset
      fix last keyframe index accross body partitions
      adjust time rate according to elst time offset
      write timecode track
      parse mpeg2 gop header time code and use it
      do not set start time to 0, it is not the case with elst time offset
      fix frame size, time rate is not stts duration anymore with time offset
      extend resampling API, add S16 internal conversion
      rename wrongly named b_per_gop to b_per_i_or_p according to specs
      return error if read failed
      check body partitions count before writing
      set last packet next pointer to null
      write active picture height for d-10/imx in mov
      encode klv fill item length in fixed ber 4 bytes
      write essence elements klv packets with ber4 length, more interoperable
      split mpegvideo descriptor into cdci descriptor and wav common in sound common
      store display height without vbi
      MXF D-10 muxer
      cosmetics, reindent, add/remove some empty lines, redundant comment
      simplify
      set mime type for mxf
      fix slice offset computation
      cosmetics, rename nb_frames to nb_fields
      compute body offset and index entry offset correctly
      fix mxf interleaving packet purge
      merge mxf_write_d10_packet into mxf_write_packet
      fetch display aspect ratio from mpeg-2 essence
      simplify descriptor size computation logic
      fix rip body sid when muxing cbr
      use index 0 for timecode track and write it in source package also
      flush pending write in mxf_write_footer
      do not write slice count when muxing cbr
      remove redundant comment
      start counting track number from 0
      check fifo size and realloc if needed
      fix preface version
      always use multi track since timecode track is present
      fix partition index byte count
      parse aac extradata to fetch channels and sample rate, patch from Netgem
      return AVERROR(ENOMEM) if alloc failed, set count after alloc, use variable when reallocing, based on patch by Netgem
      check av_get_packet return value
      do not set error if an unknown parameter is in config file, just skip
      add bandwidth before failing if feed is already being received
      mention MXF D-10 Mapping muxer in documentation
      start track id at 2, track id 1 for timecode
      do not set frame start bit since marked as not used and factorize
      set channel number even for inactive channels
      simplify, use codec->block_align
      fix index edit rate, inverse num and den
      dv probe
      revert r16717, r16718, r16719, EAGAIN handling, this causes FFserver to hang

Bartlomiej Wolowiec (17):
      using stream type in eac3 parser
      using EAC3_STREAM_TYPE_* instead of numbers
      adding lacking include to aac3_parser.c
      removal of stream_type in AACAC3ParseContext and adding AACAC3FrameFlag
      change name from stream type to frame type in AC3 code
      change of aac_ac3_parser, so it is able to send complete portion of data to decoder
      Corrections of errors in aac_ac3_parser
      Correction of typo in aac_ac3_parser
      undo changes in aac_ac3_parser
      change ff_ac3_parse_header() to take a GetBitContext instead of const char*
      add a ff_ac3_parse_header_full() which calls ff_ac3_parse_header()
      Nellymoser handling in FLV
      Separating from nellymoserdec.c parts required by encoder
      Little-endian bitstream writer
      Okayed parts of nellymoserenc.c
      Okayed parts of nellymoserenc.c
      Remaining parts of Nellymoser encoder

Bartłomiej Wołowiec (1):
      Use of new bitstream in vorbis_enc.c.

Ben Hutchings (1):
      assume widescreen when ratio also > 17, patch by Ben Hutchings, ben at decadent dot org dot uk

Ben Littler (30):
      fix url typo pointed out by Guillaume LECERF
      obfuscate urls and fix note about them
      IntraX8 (J-Frame) sub-decoder for WMV2 and VC-1
      jtype frame implemented
      update adding codec docs
      removed gif decoder in r14209
      only animated gif (gifdec.c) was removed)
      add FFDS fourcc, ok'd by kostya
      add fourccs from avi files
      add VX1K h263 and DMK2 mpeg4 avi fourccs
      add mpeg2 fourcc and aac format tag from asf
      add another mpeg4 avi fourcc from uncommon codecs list
      add rpza avi fourcc
      QIVG fourcc, fixes http://rodweb.dip.jp/partner/check.avi
      two more mpeg4 fourccs and one msmpeg4v3 from uncommon codec list
      add MPNG fourcc
      add amr in avi riff tags
      fourcc CJPG is sequential jpeg
      cosmetics: CJPG is from Creative Webcam
      add SLMJ fourcc to mjpeg
      add svq3 watermark support entry
      add LM20 mimic in avi fourcc
      hybrid wavpack support
      add T263. works on samples/V-codecs/T263/
      another mpeg4 fourcc
      add EM4A fourcc , works for:
      pvez avi fourcc is truemotion1
      ogg muxer is vorbis only
      vorbis and flac ogg muxer only
      figure out which muxers support which codecs? new api required!

Ben Taylor (1):
      Check for support of -fomit-frame-pointer before adding it to CFLAGS.

Benjamin Larsson (139):
      ss patch by (Benjamin Larsson <banan at student dot ltu dot se>)
      Ffmpeg supports AAC doesn't it patch by (Benjamin Larsson <banan at student dot ltu dot se>)
      Liba52 can be compiled to use only integers and FFmpeg can handle the
      printf -> av_log patch by (Benjamin Larsson <>banan student.ltu se)
      printf-> av_log patch by (Benjamin Larsson, banan: student ltu se)
      Cook compatibe decoder, patch by Benjamin Larsson
      This patch adds some support for non-joint stereo streams. It also
      Fix an out of array access and some minor cleanup of the code.
      Dual mono stereo strems sound ok now, added sanity checks and removed
      Incorrect check removed.
      get_vlc -> get_vlc2 transition.
      get_vlc -> get_vlc2 transition.
      get_vlc -> get_vlc2 transition and get_vlc() removed from bitstream.h
      Silence warnings, these came when some tables got declared as const.
      gcc 2.95 fix
      Fix for title field in PSP compatible mp4 files. Patch by
      Fix the wrongly commited regressions stuff for psp compatible mp4 files.
      Small cosmetics and better variable names.
      Fix broken cosmetics commit and add a check for valid headers.
      64bit fixes for the tta decoder, sizeof(long) is 8 on amd64, 4 on x86, the patch
      Set 'MJLS' FOURCC for JPEG-LS in AVI and documentation fix. Patch by
      Init simplification and 2% faster wma_decode_block on amd64 with tables use
      Use a more descriptive error message.
      AC3 encoding volume fix, by Justin Ruggles jruggle earthlink net.
      Patch to generate the ac3 window, based on code by Justin Ruggles.
      Fix samples to bytes relation.
      Revert bad ac3enc commit, the commited code was bigger and slower.
      Add support for raw flac decoding based on the .flac suffix of input files.
      Cosmetics. Add the defines needed for debug output.
      Flash Screen Video decoder (fourcc:FSV1)
      Fixed a possible bug, checked against the assembly.
      ff_fft_init now double checks that the system has mm3dnow.h before
      Possibility to skip bytes in the beginning of a file.
      SIMD vector optimizations. 3% faster overall decoding.
      -async and -vsync documentation. Based on info from the mailinglist.
      Added codec id for QCELP.
      Added Maintainers entry
      Added -map usage to docs.
      PPM vhook documentation, by Victor Paesa.
      Added qcelp faq.
      -interlace -> -ilme This has been wrong in the docs for years,
      Remove log2() usage.
      Add myself as maintainer for the imc* files.
      Missed one log2().
      Bug fix for crashes when SSE is used on unaligned arrays.
      av_log(NULL,... -> av_log(avctx,.. where appropriate.
      Correct value for the test.
      VP6 and flashsv stream copy and muxing support.
      Get rid of #ifdef CONFIG_ZLIB in the code. Code cleanup.
      Fixed a typo, cosmetics.
      The block_size might be used incorrectly if it is not updated.
      Don't break compilation on systems without zlib.
      Really fix my never ending mistakes.
      Flash screen video encoder.
      Add flashsv encoder to changelog and fix the encoder so it actually works.
      Remove debug line.
      Don't copy more then needed, bugfix by Gus Scheidt ellocogato at gmail dot com and Michel Bardiaux mbardiaux at mediaxim dot be.
      Flash screen video encoding support.
      Get rid of one warning.
      Cook stereo (MONO_COOK2) bugfix, by Ian Braithwaite.
      Sort out the diffrent cook versions.
      decode_subpacket cleanup by Ian Braithwaite ian braithwaite dot dk.
      Kill a warning and don't use modulus.
      Get rid of the COOKextradata struct. And use valid C to parse the extradata.
      Check that js_vlc_bits from the extradata is in a valid range.
      Don't use byte swaps and use the more portable way
      Clairyfication of the gain tool struct.
      Cosmetics.
      Complete the video format enum.
      Set FLV_SAMPLESSIZE_16BIT for ADPCM.
      Simplify, use avctx->frame_number.
      Respect the gop size (-g) for marking I frames. Use -g 0 gives the old behaviour.
      Per reference swf/flv adpcm encoder.
      As usual Real actually took something existing and rebranded it.
      Atrac3 decoder.
      Let maintain things. Added myself as atrac3 maintainer.
      Version bump, atrac3 added.
      Deobfuscade decodeTonalComponents calling logic.
      Added support for instances where linesize[0] is negative.
      Change default 1 pass blocksize to 64x64.
      Remove lots of unused commented code. (Cleanup)
      Use FFABS instead of abs.
      Only compile fft and mdct when needed.
      Better error reporting.
      Don't encode the first sample twice.
      Trellis quantization support for adpcm_swf.
      Cosmetics, indentation.
      Nellymoser ASAO decoder
      Added myself as maintainer of the Nellymoser codec
      Unaligned memory access rcrash fix
      Remove unused variable
      Small simplification
      Use memcpy instead of for loop
      4 months of hard development and finally it's done. Closes issue 58. Adpcm_swf regressions tests.
      Correctly handle FLV_CODECID_NELLYMOSER_8HZ_MONO files
      Reindent
      Enable proper output from the LFE channel, based on patch from Mathias Rauen dear dot madshi at net
      silence wmaenc.c:181: warning:suggestparentheses around assignment used as truth value
      Riff ids for wmapro and wmavoice, this adds support for stream copy of these codecs between asf,avi and wav containers.
      Add more riff IDs to support remuxing.
      Support setting the debuglevel (-g level) with --enable-debug=#
      Don't use unchecked data from the bitstream.
      Only set channels in the stream if previously unset, fixes resampling crash on broken dca frames
      Raw dts muxer
      Long name spelling fixes
      dts probe
      Update copyright and text
      Sample tested and the code worked so the comment is removed
      Oma demuxer
      We have an Oma demuxer now
      Forgotten oma docs stuff
      Add more codec IDs and tags.
      Rearrange parameter logic
      Add support for mp3 contained in oma
      redundant assignment found by Clang
      Fix for possible null pointer dereferencing, closes Coverity report 68 run 2.
      Prevent the qdm2 code from overreading/overflowing. Fixes Coverity ID 112 run 2
      Reduce the loglevel on a log message in the Nellymoser decoder. Related to Roundup 447.
      FOURCC addition: added 3IV2, another mpeg4 id. incoming/video_3iv2/video_3iv2.avi
      Add support for Acelp.net fourcc and codecid, remuxing wav to avi should work
      add ok parts to the rm demuxer for SIPRO support
      Fix compilation with TRACE
      Fix dca decoder with non simd float2int16 conversion
      Disable codec downmix when not using simd instead of silently produce silence
      Fix huffman coded dca samples.
      Fix selection of quant step table
      Fix compilation, remove stray ;
      Add channel layouts for dca.
      Guess channel layout based on the number of channels if a custom layout is specified in amode.
      Change the dca channel layout, S -> SIDE not BACK.
      Remove channel layout table that only should be used by encoders.
      Comment typo
      Proper channel output reordering for the dca decoder.
      Set channel_layout
      Silence warning in imc decoder
      channel layout support for ffmpeg.c
      Add size that is needed for the wmapro codec
      I'll take care of qdm2 also
      Allocate trellis tables on heap only when needed.

Benjamin Zores (3):
      remove dts_internal.h
      monkey audio demuxer now can parse ape tags
      Fix libswscale build after r27561 if --enable-runtime-cpudetection is used.

Benoit Fouet (147):
      Remove useless check, patch by Benoit Fouet, benoit.fouet purplelabs com.
      Move the aframes option from the video "subsection" to the
      Clarify FAQ, section "ffmpeg does not work".
      add CODEC_TYPE_NB to enum CodecType, will be used to declare option arrays in ffmpeg.c
      duplicate *_opts for every codec types, in order to have the  "ab" and "b" options work in the same manner
      change management of "ab" option in ffmpeg.c, so that it works as "b" option.
      Move the initialization on context definition.  In amr.c, only the
      Cosmetics in amr.c.
      Remove useless casts as recommended by Måns Rullgård.
      Use - instead of _ for AMR options in help output.
      Remove -ab 64k from ffmpeg options, this is now default.
      Accept only existing bitrates.
      Remove commented-out while loop.
      Remove useless pointer shift.
      Return correct decoded size, decoder is called with only one frame at a time.
      add a commit-rule precision when applying patches discussed on the mailing list
      more format abbreviations
      updates the x11grab doc example, adding -s cif, as it is mandatory
      add rgb suffix to rawvideo muxer
      to complete r8807: free vstats resources
      Revert my two previous deliveries (back to r8806) to aply the good patch for vstats
      test if color conv fonction exists prior to using it
      cosmetics: indentation after previous patch
      return right format in set_codec_from_probe_data
      modify order in which files are diff'd
      update reference file for seek test after delivery 8826
      add AVERROR_NOENT
      Use url_fopen error code when opening input file
      add rgb suffix handling for demuxer
      seek tests only available with GPL enabled
      Indentation
      add and remove some parentheses
      comment endif directive token
      aanscales is used only with CONFIG_ENCODERS
      remove unused constant definition
      move dct_quantize and denoise_dct function pointers initialization to C
      remove clear email addresses
      Update copyright year
      Make globals static
      add an AVClass pointer in SwsContext context
      test context before accessing it
      Indentation
      Description: split mpegvideo.c
      update regression checksums after r9447
      set thread_opaque to NULL when freeing it
      bring back h264 build
      map warning levels between libx264 and libavutil
      Add Balatoni Denes as vorbis dec maintainer
      order is name surname
      fix tools compilation
      revert my accidental previous commit
      remove useless check
      remove warnings
      use sscanf to parse address
      fill title, author, copyright and comment fields by parsing udta atom
      Avoid showing banner twice.
      right filename in doxygen comment
      add Dénes Balatoni to CREDITS
      add a close function to bitstream filter
      Add a bitstream filter for H.264.
      Forward error code when reading a v4l2 frame
      Check read_packet before using it.
      Indentation
      Increment minor after r10709.
      Merge audio and video bitstream filter options handling.
      Re-enable DEBUG_SEEK flag.
      Prevent mov muxer from writing empty udta atoms.
      Cosmetics: indentation.
      Fix mov checksums following r10852.
      Change ff_rm_read_mdpr_codecdata to get back to old behavior.
      Update show_banner date.
      Initialize "saved" buffer to zero.
      Description: remove superfluous parentheses.
      Remove non cosmetic spaces inside parentheses.
      Cosmetics: whitespaces
      Remove some useless parentheses.
      Change (a == NULL) condition to (!a) and (a != NULL) condition to (a).
      More verbose error log message.
      Inform user when bitstream filter fails.
      Remove integer.h inclusion.
      Update regression test checksum for ffm.
      Define block size for AMR-WB (use IF2 table).
      Make AMR-NB block size const and 8 bits.
      define VOF as double of VOFW.
      revert r12489.
      cosmetics (by Björn Axelsson)
      Support for BMP os2v1.
      Indentation
      fix a typo
      Add missing path to #include.
      Simplify mm_support() for ARM (by Måns).
      Cosmetics: indentation
      Remove a useless directive.
      Remove useless 'channels' assignment.
      Remove unused redefinition of av_log for test.
      Change OSS long names.
      Change table types to uint8_t
      Remove useless casts and use variable instead of its type in sizeof.
      Corrector type fits in 8 bits.
      use uint8_t instead of unsigned char
      Cosmetics
      Remove useless test.
      Check allocations on init.
      Remove useless initializations.
      Move allocated buffer out of YUV buffer structure.
      Simplify iv_free_func().
      Indent.
      Reindent whole file.
      Remove useless lines.
      Move iv_decode_frame function to remove a forward declaration.
      Merge declaration and assignment.
      Remove useless context variable.
      Fix indentation.
      Rename some variables and add some comments to try to be a bit more clear.
      Add a space between frequencies and their units.
      Add Jason Garrett-Glaser as libx264 maintainer.
      Add a macro to get the number of elements in a table.
      Cosmetics: fix indentation after previous commit.
      Bump minor version after addition of FF_ARRAY_ELEMS macro.
      Revert r15653.
      Change email address to a valid one.
      Fix warning "format not a string literal and no format arguments".
      Add const qualifiers to decode functions.
      Add jp2 support to img2 demuxer.
      s/FFmpeg-devel/ffmpeg-devel/
      Add my GnuPG fingerprint.
      Pass avf context to audio_open().
      Use avf context in av_log, instead of NULL.
      Enable building r3d demuxer when its config is set.
      Rename audio.c to oss_audio.c in libavdevice.
      Fix an exploit in indeo by checking we are not writing out of the strip array.
      Remove unneeded check.
      Add a Changelog entry for the ALSA stuff.
      Check buffer is inside what is passed when probing for flac.
      Whitespace cosmetics to align return statements.
      Allow demuxing of audio substreams stored as 0x06 type.
      Add a context to av_log() calls and extend ape_dumpinfo prototype to do that.
      Add a context to av_log() call.
      Add a context to av_log() calls.
      Add a context to av_log() calls.
      Add a context to av_log() call.
      Add a context to av_log() calls.
      Change av_log() calls to fprintf() to stderr calls.
      Add a context to av_log() call and update the trace not to be redundant.
      Add a context to av_log() calls and modify a function prototype to allow it.
      Cosmetics: reindent after previous commit.
      Add a context to av_log() call.

Bernhard Rosenkränzer (2):
      Better ARM support for mplayer/ffmpeg, ported from atty fork
      Fix PIC compilation, some defines were under #ifdef !PIC but used

Bill Eldridge (4):
      mingw patch by (Bill Eldridge <bill at rfa dot org>)
      mingw32 configure patch by (Bill Eldridge <bill at rfa dot org>)
      win32 received_sigterm patch by (Bill Eldridge <bill at rfa dot org>)
      quiet/verbose patch by (Bill Eldridge <bill at rfa dot org>)

Bill May (1):
      GCC 4.0 on Mac OS X 10.4 patch by (Bill May <wmay cisco com )

Bill O'Shaughnessy (2):
      bring AC3 encoder output up to input volume level
      AC-3 encoded volume is too high.  Revert revision 7160:

Bitbreaker/METALVOTZE X (1):
      Fix documentation of the return value of avcodec_encode_video().

Björn Axelsson (22):
      store asf streaming bitrates consistently
      Documentation fix for url_f(d)open()
      Add functionality to set the direction of a ByteIOContext buffer.
      Allow propagation of stream selection through the ASF demuxer to the
      ffplay currently needs special handling for pausing in some protocols.
      Use dynamically allocated ByteIOContext in AVFormatContext
      Extend URLProtocol with new function pointers and api functions for
      Extend ByteIOContext and add the buffered IO functions:
      Enable av_read_pause(), av_read_play() and the ASF demuxer's av_read_seek()
      Add AVClass to URLContext at next major version bump
      Add libavutil to DEP_LIBS so that PROGS get rebuilt upon libavutil changes.
      Add a generic write function to av_fifo.
      Fix vertical resolution of subtitles.
      Fix blend_subrect for subrects positioned on odd rows.
      Fix blend_subrect for some subrects positioned on odd rows.
      Add missing ffplay subtitle cycling keybinding.
      Fix timestamp parsing in XSUB decoder.
      Fix blend_subrect for even-width subrects positioned on odd columns
      Add ffplay subtitle stream selection command-line parameter
      Check that start_display_time is 0 in avcodec_encode_subtitle()
      Check that there are subtitle rects to encode in avcodec_encode_subtitle()
      Set AVSubtitle format to 0. Neither dvdsubdec nor xsubdec intializes format

Björn Sandell (1):
      various openbsd and general warning fixes - patch by Björn Sandell <biorn at dce.chalmers.se>

Bobby Bingham (5):
      Targa (.tga) encoder,
      Add 15 bit support, patch by Bobby Bingham, uhmmmm gmail com.
      Move the encoding of the image data to its own function.
      Add RLE encoding support, patch by Bobby Bingham, uhmmmm gmail com.
      Change the vhook code to send real timestamps to the filters instead of the

Bohdan Horst (4):
      this include isn;t longer need and now generate a lot of warnings patch by Bohdan Horst <nexus at hoth.amu.edu.pl>
      bsd patch by (Bohdan Horst <nexus at irc dot pl>)
      INT MIN/MAX patch by (Bohdan Horst <nexus at irc dot pl>)
      FreeBSD fix

Brad (2):
      V4L2 support for OpenBSD, patch by Brad, brad comstyle com
      AltiVec detection support for OpenBSD, patch by Brad, brad comstyle com.

Brendan Kehoe (1):
      buglet in libpostproc Makefile patch by (Brendan Kehoe <brendan at zen dot org>)

Bret Hughes (1):
      gcc 4 compilation fix

Brian Becker (2):
      Correcting the header size for ADPCM_IMA_WAV files patch by (Brian Becker <brian dot becker at palmone dot com>)
      Parse stss info in MOV files to get key frames patch by ("Brian Becker" <Brian dot Becker at palmone dot com>)

Brian Brice (5):
      fix ODML AVI Duration for files > 2GB
      Fix wrong ASF duration of generated files
      * fixing a memory leak in DV format.
      Flush mp2 as well on seeking. Part of a patch by Brian Brice bbrice gmail com.
      Clear synth_buf on flush(). Part of a patch by Brian Brice bbrice gmail com.

Brian Foley (7):
      first cut at altivec support on darwin patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      altivec build tidyup patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      altivec optimizations patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      * altivec and pix_norm patch by Brian Foley
      altivec accelerated v-resample patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      -map and sanity checking patch by (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      Move copy_block* functions to dsp dsputil.h

Burkhard Plaum (4):
      Memory leak fix patch by (Burkhard Plaum <plaum >at< ipf.uni-stuttgart )dot( de>)
      Call av_free_static automatically before DLL unload patch by (plaum:ipf uni-stuttgart de)
      Fix for memleak in dv.c patch by (Burkhard Plaum; plaum, ipf uni-stuttgart de)
      Fix DV memory leak and allow multiple instances patch by (Burkhard Plaum?)

Calcium (10):
      This patch corrects the bug, that if padbottom IS NOT specified and
      08_limit_in_filesize.patch by (Calcium | calcium nurs or jp)
      04_AMR_no_debug_output.patch by (Calcium | calcium nurs or jp)
      12_finish_encoding_at_shortest_stream_was_finished.patch by (Calcium | calcium nurs or jp)
      19_audio_volume_adj.patch by (Calcium | calcium nurs or jp)
      28_fix_parameters_in_G726.patch by (Calcium | calcium nurs or jp)
      29_vtag_in_ASF_not_effective.patch by (Calcium | calcium nurs or jp)
      30_clean_up_global_header_flag.patch by (Calcium | calcium nurs or jp)
      31_ASF_compatibility_with_WMP_and_add_index.patch by (Calcium | calcium nurs or jp)
      25_fix_problem_that_unsynced_with_ss_option.patch by (Calcium | calcium nurs or jp)

Carl Eugen Hoyos (100):
      Move base64.[ch] to libavutil.
      Fix crash when trying to convert from or to RGBi (i<15) as in bgr8torgb24
      rename attribute_unused to av_unused and moves its declaration to common.h
      rename attribute_unused to av_unused as it is declared in common.h
      Improve understanding ofavcodec_find_decoder()
      kill warning
      Ensure av_tree_insert() is defined before using it.
      Remove error raising when explicitely skipping frame
      Silence warnings about NAL_SVC_NON_IDR in H.264.
      Enable PAFF decoding
      Set pict->interlaced_frame if returned frame is MBAFF or PAFF.
      Allow compilation with icc 10.1.
      Make dither4 & dither8 const.
      Discard two symbols from libswscale.
      One const per declaration is enough.
      Do not return -1 from void functions.
      Silence useless icc warnings.
      Allow icc to compile with optimizations (pentium3, pentium4 or core2).
      Replace some occurrences of -1 with PIX_FMT_NONE.
      Do not use variable frame_info before its value is set.
      Replace some occurrences of 0 with CODEC_ID_NONE.
      Replace one more occurrence of -1 with PIX_FMT_NONE.
      Correct type for pcx_rle_decode().
      Add MATROSKA_TRACK_TYPE_NONE.
      Fix icc warning #188: enumerated type mixed with another type
      Add VP56_FRAME_NONE and use VP56_FRAME_CURRENT.
      Do not loop endlessly if converting into dv with timestamps Jan 1st not
      Fix compilation for buggy icc.
      Warn about unimplemented H264 feature.
      Revert r13542: Spatial direct mode probably also works with PAFF.
      Show result of yasm test.
      Remaining parts of GSoC MXF muxer by Zhentan Feng.
      MXF is a valid output format now.
      Whitespace cosmetics
      Fix a warning about an undefined function when compiling h264.c
      Fix memleak for currupt input.
      Rename copy_picture to ff_copy_picture.
      Allow using DECLARE_ALIGNED with Sun cc.
      Remove g6Dither from libswscale.
      Remove b5Dither, g5Dither and r5Dither from libswscale.
      Change variable types from int to enum PixelFormat.
      Change variable types from int to enum PixelFormat.
      Don't include byteswap.h anymore.
      Use PIX_FMT_NONE instead of -1 in AVCodec.pix_fmts.
      Fix some icc warnings by using enum PixelFormat instead of int where appropriate.
      Enable AC-3 decoding regression tests.
      Revert r15615: Enable AC-3 decoding regression tests.
      Remove unused code that can't be compiled without svn archive.
      Fix H.264 decoding.
      One const is enough.
      Support picture output without -f image2 for most codecs.
      Add forgotten file format pgmyuv.
      Use PIX_FMT_NONE to silence icc warning #188:
      Recognize pnm file as images.
      Use AV_GCC_VERSION_AT_LEAST() to simplify gcc version checks.
      Silence the following icc warnings:
      Fix a warning when compiling with icc:
      Fix two identical warnings when compiling riff.c and nuv.c with icc:
      Fix h264 decoding on SSE2 cores with icc compilation.
      Silence one warning when compiling with icc:
      Reduce number of warnings when compiling with icc by two.
      Revert r16375: It broke aac decoding.
      Silence two icc warnings:
      Fix small errors in configure --help's output.
      100l: Add a deleted '{'.
      Cosmetics: Fix indentation.
      Remove leading underscore from union declaration.
      Ministry of English Composition treatment: s/Alloced/Allocated
      Rename vdpau_render.h as vdpau.h.
      Export VDPAU support by installing vdpau.h.
      1l: Forgot to rename header guard.
      Factorise setting of avctx->pix_fmt.
      Rename mpeg_set_pixelformat() as mpeg_get_pixelformat().
      Rename bitstreamBuffers* as bitstream_buffers*
      Rename ff_VDPAU as ff_vdpau.
      Add PIX_FMT_VDPAU_H264.
      Add refdist_flag to VC1Context.
      Move VDPAU check, make future VDPAU patches smaller.
      Move vdpau_h264_set_reference_frames() call from vdpauvideo.c into h264.c.
      Fill range_map* in VC1Context.
      Cosmetics: Break long line.
      Use only MpegEncContext in ff_vdpau_h264_add_data_chunk.
      Cosmetics: Rename ff_vdpau_h264_add_data_chunk as ff_vdpau_add_data_chunk.
      Fix OBJS-$(CONFIG_H264_VDPAU_DECODER).
      Remove H264Context from vdpau_internal.h.
      Fix compilation.
      12l: Fix compilation with --disable-vdpau.
      Add mpeg1_vdpau decoder.
      Remove CODEC_ID_H264_VDPAU.
      Fix compilation with --enable-libxvid.
      Add PIX_FMT_VDPAU_WMV3 and PIX_FMT_VDPAU_VC1.
      Cosmetics: Fix indentation after last commit.
      Sort VdpPictureInfo alphabetically.
      Silence one icc warning:
      Silence following warning if only zeros were not consumed:
      Add myself as vdpau maintainer.
      Fix VC1 and WMV3 playback in MPlayer.
      Apply correct pix_fmts to WMV3_VDPAU and VC1_VDPAU.
      Completed documentation of reference field.
      Cosmetics: Fix indentation after last commit.

Cedric Vincent (1):
      The atom 'skip' should be handled by 'mov_read_leaf()', not by

Charles Yates (5):
      stdin patch by (Charles Yates <charles dot yates at pandora dot be>)
      vhook patch by (Charles Yates <charles dot yates at pandora dot be>)
      ppm vhook follow up patch by (Charles Yates <charles dot yates at pandora dot be>)
      support BUILDSUF patch by <charles dot yates at pandora dot be>
      Take build_suffix into account in pkg-config files.

Chip (3):
      inttypes.h patch by (Chip <szarlada at freemail dot hu>)
      make ffmpeg/libavcodec/liba52/crc.c compile on systems where no inttypes.h patch by (Chip <szarlada at freemail dot hu>)
      Altivec test on AmigaOS4 patch by (Chip <szarlada at freemail dot hu>)

Chris Flerackers (3):
      inttypes emulation cleanup patch by ("Chris Flerackers" <cflerackers at androme dot be>)
      rtp callback context patch by ("Chris Flerackers" <cflerackers at androme dot be>)
      - Support empty GOB's (no mb's) and skip all mb's

Chris Halse Rogers (1):
      This patch to configure makes the --tune option recognise k8 core cpus

Chriss (3):
      DVDNav4 patch by ("Chris" <chris at garveycocker d0t com>)
      This patch takes into account that fifo_realloc may adjust fifo.wptr
      -target dvd minimum vobu length patch by ("Chris" [chris garveycocker com])

Christian Lohmaier (1):
      fix error: indeo2.c:202:1: directives may not be used inside a macro argument,

Christophe Gisquet (12):
      Drop ff_ prefix for static functions
      Replace function declarations with macro
      Pass modes as parameters instead of calculating them inplace.
      Make bicubic interpolation standard compliant
      Use put_signed_pixels_clamped where appropriate
      add VC-1 MMX DSP functions, under MIT license.
      Strip debug stuff from vc1dsp_mmx.c, patch by Christophe GISQUET %hristophe P gisquet A free P fr%
      Typo fix. Previous version had some picture error building up until next keyframe.
      add MMX version for put_no_rnd_h264_chroma_mc8_c, used in VC-1 decoding.
      Fix issue #301:
      Factorize some duplicated code from CAVS and H.264 into a common file.
      Select scan tables for 8x4 and 4x8 blocks only once.

Christophe Lorenz (1):
      Add dv50 tag to support samples with dv50 fourcc in AVI.

Christophe Massiot (2):
      miscalculation of header_bits when using the -thread option bugfix, patch by (Christophe Massiot <cmassiot at freebox dot fr>)
      Border processing adaptive quant patch by (Christophe Massiot |cmassiot freebox fr)

Christophe Mutricy (1):
      Avoid "HAVE_CMOV not defined" warning.

Clemens Ladisch (1):
      fix end-of-file detection

Colin Leroy (2):
      fixing RGB32->RGB16 on big endian patch by (Colin Leroy <colin at colino dot net>)
      fix compilation on non-x86 with gcc 2.95

Colin Ward (1):
      Move CONIG_MUXERS below struct frame_attributes, it's not muxer-specific.

Corey Hickey (13):
      DTS is 0x2001 according to: http://www1.mplayerhq.hu/MPlayer/DOCS/codecs-status.html
      sc_factor support for encoding.
      Add a newline to b_frame_strategy error message.
      Marcus Engene is author and maintainer of watermark.c
      make ff_rate_control_init() bail out if rc_strategy==1 and lavc wasn't
      - Add new file internal.h for common internal-use-only functions.
      Fix regression tests to match changed output of AC3 encoder
      Make b_frame_strategy not fail on pass 2, but still print a notice.
      Make B-frame reduction sensitivity by b_strategy/vb_strategy = 1
      error message spelling typo
      second pass encoding changes:
      Correct description of mblmin and mblmax.
      Don't forget about audio remaining in buffer, in case there is enough

Craig Nicol (1):
      Corrected pause handling when using the 's' step by frame key.

Cédric Schieli (9):
      make MMX registers parametrized in the WRITEBGR32 macro,
      splits various YSCALEYUV2xxx macros into YSCALEYUV2xxx_UV,
      move zeroing of alpha channel register out of YSCALEYUV2xxx macros,
      Factorize some code in yuv2rgb_template.c to ease further yuva2rgb patch.
      Refactor code for upcoming alpha patches.
      Use YUV420P code path for YUVA420P where appropriate
      Add an isALPHA macro to check if pixel format has alpha channel
      Fix avpicture_layout to not chroma shift the alpha plane when outputting YUVA420P
      When converting from a non alpha format to an alpha format, defaults

D Richard Felker III (15):
      10l found&patch by D Richard Felker III <dalias at aerifal.cx>
      new ffserver options patch by (D Richard Felker III <dalias at aerifal dot cx>)
      C implementation of the median deinterlacer (seems to be the only one
      yuv4mpeg pipe reader for libavformat patch by (D Richard Felker III <dalias at aerifal dot cx>)
      this isn't actually stupid, but it's not valid C and gcc 3.5 rejects it as such
      consistent use of types patch by (D Richard Felker III <dalias at aerifal dot cx>)
      1000l
      Patch from Gianluigi Tiesi (sherpya at netfarm dot it):
      100l -- mismatched type after changing sizes to type long!
      10l to michael for breaking gcc 2.95 compile :)
      fix constraint violation: libavformat is not allowed to modify state of caller, including rng state
      100l to loren. at least this fixes the compile error, but the code is still broken (decoding buggy/crashing)
      signedness error caused broken decoding
      sync message with modern ffmpeg bitrate option syntax.
      Do not use GNU-specific (or BSD-specific or whatever they may be)

Damiano Galassi (1):
      skip 2 bytes more when parsing ac3 frame, syncword 16bits + crc 16bits,

Dan Christiansen (6):
      ranlib patch by (Dan Christiansen <danchr at daimi dot au dot dk>)
      Apple GCC 3.3 compile fixes patch by (Dan Christiansen <danchr at daimi dot au dot dk>)
      PowerPC warnings & optimisations patch by (Dan Christiansen <danchr at daimi dot au dot dk>)
      - Moving -Wall from the individual Makefiles to configure (and only
      Using the extension m4a for audio only MPEG-4 files is increasingly
      gcc->C99 and warning fixes patch by (Dan Christiansen <danchr at daimi dot au dot dk>)

Dan Maas (3):
      size[0-3] are not initialized (and can get random negative trash
      AC coefficient weighing (per SMPTE 314M) for the DV encoder and decoder. The
      Assign class 2 to most macroblocks by default, instead of a more conservative

Dan Villiom Podlaski Christiansen (1):
      Fix detection of IPv6 support on Darwin by ensuring that

Daniel Cardenas (1):
      give the position offset of the file in the packet

Daniel Kristjansson (5):
      part of the 'libavformat/utils.c doxygen documentation' patch by (Daniel Kristjansson: danielk, cat nyu edu)
      "removes definition for old crc func" patch by Daniel Kristjansson / danielk mrl.nyu edu
      "converts printf's to avlog's" patch by Daniel Kristjansson / danielk mrl.nyu edu
      mpegts-documentation-only.patch  -- adds documentation patch by Daniel Kristjansson / danielk mrl.nyu edu
      fix mem leak

Daniel Maas (1):
      DVCPRO50 support.

Daniel Serpell (2):
      Fix the PSP examples in the FAQ to use libx264 as codec instead of h264.
      Add an example of higher resolution video encoding for the PSP to the FAQ.

Daniel Verkamp (9):
      Fix one peculiar case in Smacker decoding.
      remove dead netscape loop header code in gif encoder, patch by Daniel Verkamp, daniel at drv dot nu
      remove dead rgb24 code, gif encoder now directly takes palette, patch by Daniel Verkamp, daniel at drv dot nu
      remove no more needed fields in GIFContext, patch by Daniel Verkamp, daniel at drv dot nu
      Add missing av_cold in static init/close functions.
      Document 'ffmpeg -formats', patch by Daniel Verkamp, daniel drv nu.
      cosmetics: Break image format lines before comments.
      Split PGM and PPM in image formats section.
      Add missing image formats, patch by Daniel Verkamp, daniel drv nu.

Dario Andrade (1):
      set more x264 encoding parameters

Dave Yeo (12):
      Recurse into libswscale on uninstall.
      ln is not available on all platforms, so introduce a variable for it
      Add SLIB_UNINSTALL_EXTRA_COMMAND to uninstall the things installed via
      Add #include "os_support.h" to restore OS/2 support.
      OS/2 support restored in cleaner form.
      Add small OS/2 section with a link to further information.
      Add portable way to check for the existence of definitions and
      Add attribute_align_arg to os2thread.c.
      Create object files in a.out (aout-emx) format on OS/2. This works around
      sys/resource.h needs sys/types.h on OS/2, patch by Dave Yeo, daveryeo telus net
      struct timeval requires #include <sys/time.h>
      Add sys/time.h header #include, fixes compilation on OS/2.

David Bateman (1):
      IWMMXT-specific dsputils clear_blocks function.

David Bolt (1):
      Print timestamp with increased precision.

David Bryant (2):
      Add the handling of the INT32INFO block to the WavPack decoder.
      Support for WavPack version 0x410 (false stereo chunks)

David Conrad (171):
      Support AAC audio in AVI with tag 0xff
      AVI tag reading and writing patch by David Conrad.
      Fix for AVI files with odd sized 'movi' list
      The NSV demuxer assumes that a video frame's timestamp increases by one on each
      Change from using a seek table internal to the TTA demuxer to using the standard lavf index.
      Support for seeking in TTA files.
      split some common code from the mkv demuxer that will be useful to the muxer
      Add EBML ID for Segment UID
      Set dts in tta demuxer
      Use PRId64 in seek test
      Add av_get_bits_per_sample_format()
      Add V_SNOW
      SimpleBlock keyframe flag is the most significant bit
      Beginning of mkv muxer, only EBML head is written correctly
      Simplify
      Cosmetics: align nicely for better readibility
      Get rid of put_ebml_utf8(); the error checking that would have distinguished it from put_ebml_string() belongs elsewhere in lavf.
      Const correctness
      put_ebml_float()
      Remove useless debug write
      First stab at writing the tracks element, still needs some additional cases for certain codecs
      Write one cluster and SimpleBlocks for the frames. Should now create playable mkv files for some video codecs (H.264 and VP3 checked)
      Compile fix; missed a variable declaration in the last commit
      Correctly write CodecPrivate element for Vorbis and Theora
      Fix writing Xiph header sizes
      Start a new cluster every 5 MB or 5 seconds
      SimpleBlocks are Matroska v2
      Track number and UID only have to be nonzero
      Write unknown size if the size given is too large for EBML (greater than 2^56-1)
      Remove unused prototype
      put_ebml_void()
      Write the duration of the file
      Move Xiph's CodecPrivate writing code to its own function
      Move writing the tracks element to its own function
      Make sure to return a value in functions that return a value
      Write the Seek Head element
      Indentation
      Calculate the size of a SimpleBlock before writing it so that bytes aren't wasted in writing the size
      Doxygenize comments
      Write the cues element
      Write wav header if there is no native audio codec ID
      Write bit depth for PCM audio
      Use av_get_bits_per_sample
      Simplify
      Keyframe is the first bit not last
      Simplify
      Determine the output sample rate for SBR AAC and write it
      Only write extradata if it exists
      Move writing a block to its own function
      Write subtitle tracks
      Indentation
      Simplify
      Floats aren't used enough to justify writing any with single precision
      Using LIBAVFORMAT_IDENT for both muxing and writing app is good enough
      Cosmetics
      Write the creation time
      Write segment UID
      Move writing Xiph-style sizes to its own function
      Write FLAC codec private correctly
      Codec Private should be correct for all but Real codecs now
      Revert writing the file creation date
      Write the display size elements
      Add some debug logging
      Add mka muxer
      Set the language to undefined if no language specified
      Doxygenize some comments
      Revert using the time to generate the segment uid
      Use a MD5 hash of some frames to write the segment uid
      Correct handling of smaller unknown sizes
      Move ebml_id_size()
      Simplify put_ebml_id()
      Make a byte always mean a byte
      Move calculating the bytes needed to represent a size in EBML to its own function
      Simplify
      Set default subtitle_codec to CODEC_ID_TEXT
      Modify put_ebml_size() so that the bytes parameter is exact rather than minimum
      Use cluster_pos when checking cluster position
      Calculate the size of key EBML master elements beforehand so only just enough size is reserved for the size
      Use PRIu64/PRId64
      Clarify comments
      Calculate and store DAR rather than SAR
      Use sample format for bit depth if av_get_bits_per_sample() doesn't give one
      Break long comments
      Grammar nits
      Simplify
      Use assert for conditions that must be true
      Indentation
      Return AVERROR(ENOMEM) when appropriate
      Check that the seekheads were created
      Pass the returned error code of functions up
      Always check the return of mkv_add_seekhead_entry
      Correct message
      Split some really long lines
      Make sure that the calculated duration doesn't decrease
      Doxygenify comments
      Use num_entries directly rather than mixing it with a local copy
      Cosmetics
      Move writing codec private element to its own function
      Write codecprivate to a dynamic buffer so that seeking isn't required
      Match the behaviour betwen the bmp and wav codec tag lookups
      Indent
      Get rid of useless braces
      Don't seek when streamed
      Indentation
      We shouldn't be passing in sizes larger than 2^56-2, so use an assert
      Rename ebml size functions to indicate that they can be used for more kinds of numbers
      Always use AVFormatContext for av_log messages
      Grammar nits
      Enable Matroska muxer
      Matroska regression tests
      Use V_MPEG4/ISO/ASP for mpeg4 when muxing mkv
      Update mkv regression tests after r10421
      Use correct asm object type and necessary -DPIC for yasm on 64-bit Darwin
      Let ff_pw_8 be used as an SSE constant
      Use ff_pw_8 in MMX/SSE VP3 IDCT
      Remove unused dequantization code from SSE VP3 IDCT
      Don't declare SSE vp3 idct data static, so it can be used in the mmx version
      Declare ff_vp3_idct_data to be uint16_t
      Use ff_vp3_idct_data in vp3dsp_mmx.c rather than duplicating it
      Fix compilation on Mac OS X 10.4: Defining _POSIX_C_SOURCE hides the u_char &
      Take idct permutation into account when initializing dequantization matrices
      MMX/SSE2 VP3 IDCT are bitexact now that the dequantization matrices are permutated correctly
      Rewrite MMX VP3 IDCT in inline asm
      Rewrite SSE2 VP3 IDCT in inline asm
      Cosmetics: reindent
      Ensure MMX/SSE2 VP3 IDCT selection isn't disabled when only Theora is enabled
      Cosmetics: remove excessive blank lines
      Factorize SSE2_(Row|Column)_IDCT into one macro
      Make a separate macro for writing an entire block from registers to memory
      Remove duplication of TRANSPOSE8 macro, also removing some redundant load/stores
      Use skip_frame for keyframe-only decoding rather than #ifdef
      Downgrade severity of leftover bits in header packets, and don't check for the comment header
      Remove unused vp3 debugging info
      Missed hunks of last commit
      Merge (vp3|theora)_calculate_pixel_addresses
      Cosmetics: indentation
      filter_limit_values only needs 7 bits, make its tables smaller
      Visible width/height fields were added in Theora 3.2
      Cosmetics: indentation
      Return if unpacking motion vectors for a keyframe
      Unneeded braces
      Unneeded else
      Cosmetics: reindent
      Move VP3 loop filter to DSPContext
      Remove unused variables
      MMX VP3 Loop Filter
      VP3 loop filter is mmx2 not mmx
      Combine non-bitexact sections
      Cosmetics: reindent
      Convert typeof keyword into __typeof__; this is a gcc extension and the former is not accepted in C99 without GNU extensions enabled (e.g. via -fasm).  This fixes compilation on PPC.
      Speex decoding via libspeex
      Documentation and changelog update for libspeex decoding
      Fix compilation with only Theora decoder enabled
      Workaround for shared libs on OS X with Apple gcc 4.0.1
      AltiVec version of h264_idct(8)_dc_add
      add AltiVec implementation of biweight_h264_pixels(16|8)x(16|8|4)
      offset and weights are signed, fixes some non-bitexact issues.
      fix compilation with GCC-4.3+
      Fix float_to_int16_altivec prototype to match float_to_int16's in dsputil.h
      add AltiVec implementation of weight_h264_pixels(16|8)x(16|8|4)
      Add Altivec version of vector_fmul_window.
      Add AltiVec versions of h264_idct_add(8|16|16intra),
      Read extradata (justification, colors, fonts, etc) for mov/mp4 timed text
      Add support for muxing mov/mp4/3gp timed text streams
      Use the dimensions in tkhd for timed text tracks
      Flush the buffer after writing the header and when done with writing the files
      Treat disposition==0 as undefined and don't write the default flag
      Cosmetics: indent
      Update regression tests after r16651
      Copy width/height for subtitle streams (needed for scodec copy of mov/mp4
      Workaround for gcc 3.4 to align sh properly

David DeHaven (1):
      Handle ID3v2 tags in raw FLAC streams by skipping them.

David Liu (1):
      Close parse context.

David S. Miller (2):
      Alignment fixes, caused problems on SPARC.
      sub_mb_type[] needs to be 8-byte aligned because it is referenced

Denes Balatoni (6):
      denoise_dct_sse2() patch by (Balatoni Denes <pnis at coder dot hu>)
      1.) LGPL license mentioned
      vorbis 5:1 decoding fix, fix mi2_vorbis51.mp4, patch by Denes Balatoni <dbalatoni at programozo dot hu>
      Don't segfault if an empty codebook gets used.
      UltraSPARC VIS yuv2rgb
      add simple_idct implemented in Sparc VIS

Denis Fortin (4):
      add some comments
      printf --> av_log, patch by Denis Fortin, fortin nerim net
      Rename code012() to ff_code012(), and remove static qualifier: paves the way
      static (inline) X => ff_X

Dennis Melentyev (1):
      FreeBSD malloc aligns on 16 byte boundaries, so no need to use memalignhack.

Derk-Jan Hartman (1):
      Add the flags for faac and mp3lame to $extralibs in configure, so they

Diego Biurrun (1884):
      k6 pop fix patch by (Diego Biurrun <diego at biurrun dot de>)
      Syntax errors fixed.
      Make -pphelp output consistent with the other instances of -xxx help.
      Quote path expression to take care of $(prefix) possibly containing spaces.
      Mention the l5 pp filter in 'mplayer -pphelp' output.
      Replace stray spaces by a tab.
      Indeo 2
      Mac OS X 10.4 compilation fix by Steven M. Schultz <sms at 2BSD dot COM>
      Add missing pp=ac to pp_help.
      Cygwin now has a native inttypes.h that works better than the custom one.
      Ignore .pc files.
      spelling/grammar/wording
      The default MPEG-4 fourcc is now FMP4.
      spelling/grammar/wording
      spelling/grammar/wording
      Cosmetics: Break overly long lines.
      Further spelling/wording/grammar fixes suggested by the Wanderer.
      spelling/wording/grammar
      spelling fixes suggested by the Wanderer
      spelling/grammar/wording
      Further spelling/grammar/wording fixes as suggested by the Wanderer and Mans
      consistent spelling
      Talk of MPEG-4 instead of DivX.
      Clarify the sentence about WMP object IDs.
      Cosmetics: Removed trailing whitespace, converted tabs to spaces.
      spelling/grammar/wording/phrasing
      typo
      spelling/phrasing fixes, mostly by Mans
      Rephrase a paragraph for ease of understanding, suggested by Mans.
      spelling fixes
      spelling fixes
      Reduce verbosity.
      spelling/grammar/wording
      Add missing columns to multicolumn declaration.
      typo: viedo --> video
      Wrong comment after #endif.
      Reduce unnecessary swscaler verbosity.
      spelling/grammar/wording
      CAREFULL --> CAREFUL typo fix
      minor typo fixes
      Rename put_string to ff_put_string to avoid a symbol clash on Mac OS X.
      Move lrintf implementation from libavcodec/dsputil.h to libavutil/common.h
      Misplaced #ifdef, the MMFContext struct is needed outside CONFIG_ENCODERS.
      Add some files generated by regression tests.
      Replace CONFIG_ENCODERS/CONFIG_DECODERS with CONFIG_MUXERS/CONFIG_DEMUXERS
      Make configure bail out when confronted with unknown options.
      spelling fixes
      support for building dynamic libraries on Mac OS X
      Remove references to release or CVS versions, they are misleading.
      Ignore AMR directories.
      Remove .pc files on distclean.
      DESTDIR support , patch taken from the Debian package
      Add DTS_INC to the CFLAGS for DTS.  This is only set by external configure
      FFmpeg and feature requests
      Unify include paths, -I.. is in CFLAGS.
      ffserver fixed
      ffserver should be fixed now.
      Elaborate on the Visual C-- situation.
      Unify include paths, -I.. is in CFLAGS.
      cook
      QDM2
      COSMETICS: Remove all trailing whitespace.
      Add some comments to #ifdef #else #endif blocks and fix wrong ones.
      Reportedly FFmpeg fails to compile on Cygwin with vhook enabled, but FFserver
      COSMETICS: tabs --> spaces, some prettyprinting
      updates
      spelling/wording/grammar
      spelling/grammar
      Replace buggy --disable-codec/--enable-codec with non-buggy
      Add entry about release frequency.
      TrueSpeech compatible audio decoder by Konstantin Shishkov
      bashism--;
      bashism--;  Use POSIX-compatible string handling instead.
      Simplify source_path handling, avoids some warnings with non-bash shells.
      Make all tables static const.
      RealVideo 1.0/2.0 and TrueSpeech are supported.
      Comment out WMV3 decoder for now, it crashes on every file and causes
      --enable-shared-pp should directly enable pp and not depend on --enable-pp.
      consistency cosmetics
      Move Darwin shared library build peculiarities to configure.
      Update licensing information: The FSF changed postal address.
      Update licensing information: The FSF changed postal address.
      minor consistency fix
      Change this Makefile to be consistent with all the others.
      Point at new bug report page.
      CamStudio video added.
      Add a note about what to do when adding a new codec.
      CamStudio video decoder
      Create symbolic links for all library sonames and clean up Makefiles.
      Clarify codec addition instructions.
      Fix a bunch of spelling/grammar mistakes in doxygen comments and output.
      Move library name generation to configure, simplifies build system a bit more.
      Remove redundant --enable-shared-pp configure option and make the PP code
      Define separate version number for libpostproc.
      Don't use 'rm -f *$(SLIBSUF)' in a clean rule, $(SLIBSUF) might be empty.
      Make clean/distclean consistent throughout all the Makefiles.
      Remove erroneous extralibs line in a --disable option.
      typo
      Put muxer-specific code parts in #ifdef CONFIG_MUXERS.
      Skip all includes for WinCE, instead of falling through to the #else case.
      Remove all stray tabs and trailing whitespace, this time for good.
      Make sure that ffmpeg_g and ffplay_g get removed on 'make clean'.
      warning fixes by Luca Abeni, lucabe72 ##@## email ##.## it
      Remove obscure Makefile.bak entry from distclean rules.
      Make clean/distclean behave uniformly in all directories.
      Simplify LDFLAGS handling for MinGW.
      Update for new style guidelines regarding TABs and trailing whitespace.
      On MinGW it is not possible to build shared and static libraries at once.
      typo, RealAudio 14.4/28.8
      Fix silly typos.
      JPEG-LS codec fixed.
      --enable-shared now needs --disable-static on MinGW.
      GNU Make is necessary on *BSD.
      List of people maintaining different parts of FFmpeg.
      typo
      cosmetics: indentation fixes
      Install libpostproc headers if CONFIG_PP is set.
      Install static libpostproc if CONFIG_PP is set.
      Add installlib and install-headers targets like everywhere else.
      Separate program installation into its own target.
      The install-progs target should depend on $(PROG).
      Separate header file installation from the general install target.
      Separate library install into static and shared installation.
      Make sure that install-progs has the proper dependencies when compiling
      Add incdir to configure.
      Simplify header installation.
      10l typo: libdir --> incdir
      100l typo: forgotten $
      Use the incdir variable.
      libpostproc installs into a different directory that needs to be created.
      Quote all installation paths to protect against spaces in the path.
      Move libpostproc compilation to the top-level Makefile.
      Do not add quotes around SRC_PATH, the quoting has to be done at a later
      Move SRC_PATH out of the conditional, it is always added to config.mak.
      Remove obsolete big-endian FIXMEs.
      Unconditionally overwrite config.h. The current behavior where config.h is
      Revert unconditional config.h overwriting, it may cause unnecessary rebuilds.
      link fixes
      Use BSD-compatible expr syntax and change test -ne to test != for portability.
      Remove outdated FAQ entry about A/V sync issues when grabbing.
      Make -xy help output consistent, output an empty line before and after.
      Use AV_LOG_DEBUG loglevel instead of AV_LOG_ERROR where appropriate.
      Disable the NUT (de)muxer until it has been synced with the specification.
      Reenable NUT and NUT regression tests as requested by Michael.
      Enable ffserver on Darwin, non-AltiVec regression tests pass.
      Use '-' instead of '|| true' to ignore errors.
      EXTRALIBS already contains -lm, no need to add it explicitly.
      Ignore apiexample and test programs.
      output cosmetics
      Allow passing CFLAGS through the environment on all platforms.
      Revert Windows "warning" message.
      Have EXTRALIBS appear after other link options, fixes shared build on Cygwin.
      Add mphq2 admins.
      CLEANAMRWB --> CLEANAMR typo
      Misplaced CLEANAMR.
      cosmetics: Group all AMR variables together, add some newlines.
      cosmetics
      Disentangle dependencies between amr_nb and amr_nb-fixed: amr_nb-fixed does
      amr_nb should disable amr_nb-fixed and vice versa.
      Add EXTRADEPS variable for extra dependencies that need to be built first.
      Fix amr_nb-fixed build.
      Enable AMR_NB decoder/encoder when either AMR_NB or AMR_NB_FIXED is defined.
      Add missing newlines at end of file.
      Unconditionally clean AMR subdirs.
      Ignore errors from recursive AMR cleans.
      Further simplify AMR build.
      Honor the LDFLAGS environment variable.
      Simplify SLIBNAME_WITH_VERSION and SLIBNAME_WITH_MAJOR.
      Add MPEG FourCC for MPEG-2 video.
      Remove some lines that are duplicated in common.mak.
      Add Mans as maintainer for oggparseogm.c.
      cosmetics: Prettyprint some commands.
      typo fixes
      Simplify vhook build.
      uninstallation support
      Rename singular variables to more sensible plural form.
      Generate the list of all programs, unstripped programs and man pages in a more
      typo/grammar fixes
      Remove all .cvsignore files, they are no longer necessary.
      AMR URL fixes
      Add a small comment about a part that should be in configure.
      libavcodec/libpostproc ---> libpostproc
      'make depend' should recurse into libavutil as well.
      Remove now obsolete .cvsignore files.
      CVS --> Subversion
      www1.mplayerhq.hu --> lists.mplayerhq.hu
      CVS --> Subversion
      Add support for GNU/kFreeBSD systems.
      Fix compilation with all combinations of --disable-(de)muxers.
      Rename version from CVS to the more sensible HEAD.
      Avoid binary stripping rule code duplication.
      Remove unused VERSION definition from config.mak.
      Use Subversion revision for FFMPEG_VERSION.
      Remove unused FFMPEG_VERSION_INT.
      .depend needs version.h, noticed by Uoti Urpala
      Produce correct version string even when not run in a Subversion working copy.
      'uname -m' returns IP<number> for MIPS CPUs on IRIX.
      spelling fixes courtesy of "Hervé W." H.O.W.aka.V+ffmpeg (at) gmail (dot) com
      simplify
      simplify: $(SRC_PATH)/tests --> $(V_PATH)
      Remove erroneous -I../.. flag that points outside of the src directory.
      cosmetics: Break overly long lines.
      100l: V_PATH vs VPATH typo
      spelling typo
      Add proper LGPL header.
      spelling/grammar
      Chinese AVS decoder
      Remove bogus IPP section that is not referenced anywhere.
      FastUVMV vs FastUVMC typo, noticed by bond
      Move UltraSparc CFLAG addition to configure.
      Move MinGW LDFLAGS setting to configure.
      Remove silly TEST variable that is used in one line but not directly above.
      Move gprof compiler flag handling to configure.
      Move MinGW section before the combination of FFLDFLAGS and LDFLAGS.
      MinGW should set FFLDFLAGS and not LDFLAGS so that the LDFLAGS env var works.
      100l: Add missing quotes to FFLDFLAGS in the MinGW section.
      Not every 64 bit SPARC is an UltraSPARC so optimize for v9 instead.
      CONFIG_AIFF_DEMUXER vs CONFIG_AIFF_MUXER typo
      Remove old MSVC remnants.
      cosmetics: Fix indentation after last commit.
      Split CODEC_LIST into ENCODER_LIST AND DECODER_LIST.
      CONFIG_WIN32 implies MinGW and Cygwin and possibly more, so use just
      Also remove versioned .so files on make clean.
      Remove MPlayer-specific include option hacks.
      Simplify the setting of some defaults.
      10l: the ${prefix} variable should not be expanded when setting path defaults.
      cosmetics: Reorder some entries for more consistency and readability.
      Properly handle vhook build flags for platforms that require special handling.
      Fix DESTDIR handling.  DESTDIR got prepended twice for $(DESTDIR)$libdir
      Create infrastructure to install shared and static libs into different places.
      Remove unused variable.
      Remove obsolete and misleading warning.
      Rename FFSLDFLAGS variable to FFSERVERLDFLAGS to lessen confusion.
      Fix --extra-ldflags when cross-compiling for MinGW.
      Skip the MMX/SSE and memalign() check when running on OS X/Darwin, *alloc
      Remove unnecessary darwin variable.
      10l: wrong quoting
      Fix prefix vs PREFIX handling in pkgconfig files.
      Fix FSF postal address.
      Remove redundant TARGET_OS setting.
      Comment some #endif lines.
      Remove spurious references to Ogg Theora encoder, such a thing never existed.
      Cope with localized Subversion clients.
      Set custom SLIBNAME_WITH_MAJOR and SLIBNAME_WITH_VERSION for MinGW.
      Use LC_ALL instead of LANG since the former overrides the latter.
      Create $(shlibdir), not $(libdir) for shared libs.
      prettyprinting cosmetics
      Subversion allows renaming files.
      Change one itemized list to be bulletted, remove non-referenced node
      VC-1/WMV3/WMV9 decoder
      Move MinGW special casing for shared lib creation to configure.
      Don't hardcode .dll in the MinGW section, use $SLIBSUF instead.
      Shared libraries now go in $(shlibdir), not $(libdir).
      Add quotes where necessary.
      Remove MinGW specialcasing and make shared library installation compatible
      Shared libraries should go in $PREFIX on MinGW.
      Do not assemble MMX, MMX2 or 3DNOW code unconditionally on X86 and X86_64.
      editorial changes
      Cygwin compilation and Cygwin/MinGW crosscompilation sections
      Fix some "'static' is not at beginning of declaration" warnings.
      Fix compilation with MMX disabled, the mmx_null and mmx_one constants don't need to
      Replace asmalign.h hack by ASMALIGN cpp macros from config.h.
      Mac OS X requires running ranlib on static libraries after installation.
      Simplify expression by using single quotes and avoiding backslashes.
      Remove duplicate -I option.
      Remove unnecessary -I. option.
      Simplify dsptest* rule and add libavutil to the dsptest* includes.
      Make vhook CFLAGS handling less confusing.
      Remove unnecessary include and variable declaration.
      Revert last commit, VPATH is necessary for out-of-tree builds.
      simplification cosmetics
      Move all internal -I parameters to the front of CFLAGS to avoid using external
      Move all internal -I parameters to the front of CFLAGS to avoid using external
      Explicitly include fastmemcpy.h from libvo/.
      Explicitly include libmpcodecs/img_format.h and libvo/fastmemcpy.h.
      spelling fixes for the last commit
      FFMPEG_VERSION was moved to version.h.
      Work around Solaris tr not understanding "tr a-z A-Z" as it should.
      Some gcc versions dislike version.h to appear in the list of files to
      Remove superfluous variable setting, the script will exit anyway.
      Add license from zlib.h instead of referring to it.
      Add copyright notice from the Independent JPEG Group instead of referring
      Switch license from GPL to LGPL, this file originates from libdv, which
      The license of this file is LGPL, as clarified by Fabrice.
      Exchange informal LGPL notice by official license header.
      Exchange informal GPL notice by official license header.
      Mark this file properly as LGPL as allowed by its author Leon van Stuivenberg.
      Switch license from GPL to LGPL. The original author agreed to this as
      Add official GPL header to make license explicit as discussed on ffmpeg-devel.
      Move required libs after the object file for the shared lib compilation.
      LDFLAGS_$@ contains libs, not LDFLAGS, rename it accordingly.
      Add -g to LDFLAGS instead of adding it to the command line.
      Remove empty variable declaration.
      FFLIBDIRS --> LDFLAGS
      simplify
      Remove pointless variable indirection.
      FFLIBS --> EXTRALIBS
      Replace informal GPL notice by official header.
      Add proper GPL headers.
      Add proper GPL headers.
      Add official LGPL license headers to the files that were missing them.
      Remove unused variable.
      Fix 'make depend' for libpostproc.
      Remove superfluous ifeq.
      Remove the LGPL exception clause as discussed on ffmpeg-devel
      Fix Indel --> Intel typo.
      Change VHOOKSHFLAGS (which defaults to SHFLAGS) from a shell to a Makefile
      wording/spelling
      Remove 404 link.
      Remove unused variables.
      Remove bogus .PHONY declaration from version.h target.
      make depend should recurse into libswscale as well.
      Update copyright year and mention other copyright holders.
      Fix quantiSer --> quantiZer typo.
      Add replacement for libavcodec/fdctref.c to the list.
      Merge identical targets.
      Add codectest to test targets.
      rtp.o needs mpegts.o
      Fix unused variable warning when compiling with MMX disabled.
      Remove unused variables and the corresponding warnings along with them.
      Mark all phony targets as such.
      depend and dep are really the same target.
      make dep should also work.
      Remove bogus dependency information.
      Also remove cpuid_test on clean.
      Slightly simplify apiexample compilation command line.
      Move EXTRALIBS handling to common.mak.
      Replace .. by $(SRC_PATH), otherwise there might be problems with
      Fix warnings:
      Fix warnings:
      Simplify compilation call by using $(EXTRALIBS).
      Remove redundant EXTRALIBS declaration.
      Remove cs_test and swscale-example on clean.
      Libraries are created in the build tree, not in the source tree.
      cosmetics: Put all OBJS on a line of their own for nicer-looking diffs.
      FFmpeg uses -D_ISOC9X_SOURCE and not -D_GNU_SOURCE in CFLAGS.
      Switch idct_mmx_xvid.c from GPL to LGPL as permitted by the
      Simplify EXTRALIBS handling.
      imgresample-test depends on $(LIB).
      Test programs depend on $(LIB).
      Fix imgresample-test compilation.
      Simplify test program build.
      Simplify test program build.
      message typo fixes
      cosmetics: Add backslashes to the last line of continued lines to make
      Remove bogus commented out include line.
      Replace -I../-I. by -I$(BUILD_ROOT).
      Replace -I.. by -I$(BUILD_ROOT).
      Compilation fix, printf gets redefined to please_use_av_log.
      Change license headers to say 'FFmpeg' instead of 'this program/this library'
      Change license headers to say 'FFmpeg' instead of 'this program'.
      Compilation fix, printf gets redefined to please_use_av_log.
      Compilation fix: pix_abs16x16_c was undeclared.
      output typo fixes
      Fix GPL/LGPL license version mismatch.
      Fix default license header.
      Remove gcc-workaround that is already present in libavutil/common.h.
      Use ABS macro from libavutil.
      Move sign macro to libavutil.
      Change libpostproc to use the FFMIN/FFMAX macros from libavutil.
      BRANCHLESS_CABAD --> BRANCHLESS_CABAC_DECODER
      Remove unused variables.
      Rename SIGN macro to the more fitting UNFOLD.
      printf --> av_log with some help from Oded
      Switch to the LGPL as agreed to by the author according to the
      Move CFLAGS handling to common.mak.
      Move CFLAGS handling to common.mak.
      Rename SIGN macro to FFSIGN to avoid clashes with system headers.
      Rename ABS macro to FFABS.
      Rename ABS macro to FFABS.
      Rename internal.h --> xvid_internal.h, it clashes with libavutil/internal.h.
      Add codec ID for Shorten audio.
      Add a note about illegal codec IDs.
      Remove intrinsics checks, they are no longer used.
      Put libmpeg2 IDCT functions under CONFIG_GPL, fixes link failure
      Consistently use $(BUILD_ROOT).
      Remove unused variable.
      0.pgm is now 00.pgm.
      Warn about --enable-gpl.
      Consistently use $(BUILD_ROOT).
      Revert putting EXTRALIBS in common.mak, linking fails on MinGW.
      Don't mix declarations and statements.
      Add libswscale to the include path when enabled.
      Replace FLAC FourCC by TwoCC.
      Fix #endif comments.
      Expand some #endif comments.
      Fix crash with illegal instruction, cmov is available on 686 and later only.
      Don't run libavtest if CONFIG_GPL is not set.
      Fix license header error notice by Steve LHomme.
      spelling/grammar fixes
      wording/grammar
      Rename dvdsub.c to dvdsubdec.c.
      Many systems have TEMPDIR mounted noexec, warn about this.
      Fix compilation with PIC enabled, BRANCHLESS_GET_CABAC is defined under
      Fix dependencies for 'make dep'.
      cosmetics: Restore alphabetical order.
      cosmetics: Restore alphabetical order.
      better #endif comment
      Add ARCH_X86_32 as a new define for 32 bit x86 architectures and change
      typo
      Use common define for x86_32 and x86_64.
      Use common define for x86_32 and x86_64.
      Make tests compile on non-x86.
      Do not mix declarations and statements.
      Rename SWAP macro to FFSWAP.
      Accept multiple --extra-ldflags and --extra-libs options.
      Detect some more ARM CPU types.
      Add option to disable ffmpeg build.
      Do not set audio codec_tag to 1, that would be PCM audio.
      Use a variable for FFmpeg options.
      Simplify regression test execution.
      Use MMX only under x86_32, there are compilation problems.
      Remove illegal FourCC values.
      slight simplification
      Sync documentation with the current ffmpeg -h output.
      Remove ifneq MINGW condition around man page installation.  We don't do
      Remove redundant #ifdef HAVE_THREAD, the file is only compiled
      Remove wrong/misleading comment.
      Remove some unused variables.
      Make man page installation dependent on documentation generation.
      Update list of supported codecs.
      Remove empty section.
      Move AMR support instructions into the documentation.
      Move BeOS-specific code into os_support.c.
      Remove tar target, it's hardly useful.
      MinGW compilation fix, inet_aton and #includes depend on networking support.
      Move BeOS extralibs handling into configure.
      Remove ifneq MINGW condition around man page uninstallation.  We don't do
      Remove duplicate *.lib from rm command.
      Do not build output_example and qt-faststart by default.
      Also remove cws2fws on clean.
      Remove stray newline.
      Clean up XAN DPCM hack and set codec_tag to 0 for XAN DPCM, AVI files
      Fix misplaced comment.
      typo
      Nuke config.log support.
      Kb --> kb
      typo
      Fix compilation with VC-1 decoder disabled.
      Conditionally compile some of the AltiVec optimizations.
      Remove redundant comment.
      Use a function for audio-only tests.
      Use a function for streamed images.
      Use a function for image formats.
      Use a function for libav.
      Use a function for audio encoding and decoding.
      spelling cosmetics: cliped --> clipped
      Merge a repeating variable declaration into do_video_encoding.
      Make do_video_decoding general enough to accomodate all cases.
      Make do_video_encoding general enough to accomodate all cases.
      Fix some obvious mistakes in the option descriptions.
      Don't hardcode the shlibdir on Cygwin, it's supposed to be bindir.
      Improve option description.
      Improve broken shell warning.
      libpostproc has its own version number and should not use libavcodec's.
      typo
      Generate pkgconfig files via two functions.
      grammar/wording
      vqmin/vqmax --> qmin/qmax in the option description
      Remove useless #include.
      Use $< instead of explicit names.
      Make the X11 frame grabber disabled by default.
      Remove duplicate line.
      slightly more consistency for the Darwin linker options
      Create libpostproc pkgconfig files only if libpostproc is enabled.
      Fix regression tests with out-of-tree builds.
      Register x11grab only when it has been enabled.
      typo: Properly escape a $ sign.
      Add AMR directories for out-of-tree builds.
      Disable Win32 threads by default.
      Slightly shorten libvorbis check.
      Remove redundant variable setting.
      Replace fail test with the die function.
      Properly check for all required bktr headers.
      Align configure output.
      more consistency for the output naming
      one more naming consistency change
      Reorder configure output.
      Rename variables: a52 --> liba52, a52bin --> liba52bin
      variable renaming: dts --> libdts
      Rename variables: faad --> libfaad, faac --> libfaac, faadbin --> libfaadbin
      variable renaming: mp3lame --> libmp3lame
      cosmetics: Reorder external libraries.
      ordering cosmetics
      Unconditionally enable Theora decoder. It works fine now and there is no
      Add a FIXME note about paths in pkgconfig files.
      Bail out immediately with software scaler instead of failing the tests.
      typo: pallete --> palette
      cosmetics: typo pallete --> palette
      comment grammar fixes
      Add two more variables for temporary files.
      Merge redundant comments.
      100l: Fix damage from a buggy emacs macro.
      Remove unnecessary variable indirection.
      Remove leftover variable.
      spelling, formatting
      Revert broken removal of variable indirection.
      Remove outdated Linux section.
      Make nb_streams unsigned to avoid an ugly cast.
      Rename do_libav var so as not to clash with the do_libav function on Solaris.
      Add/update some comments.
      Declare zlib dependency in the Makefile, not in the source.
      Remove useless #ifdef.
      cosmetics: List all encoders together.
      cosmetics: Sort decoder/encoder/parser variables lists.
      Clarify that Xvid is used for encoding only.
      Remove BeOS hack.
      Fix some more license headers.
      Fix FSF postal address.
      Fix license headers and FSF postal address.
      cosmetics: sort
      Theora encoding via libtheora.
      Move conditional CFLAGS and OBJS setting to common place.
      Simplify conditional compilation handling.
      PROGS depends on PROGS_G so depending on just PROGS is enough.
      comment typo
      Do not run (broken) ffserver test by default and warn about its brokenness.
      Update gcc sections.
      Add disabled_any() and disabled_all() functions.
      Make use of disabled(), disabled_all() and enabled_all().
      Make use of enabled_any().
      Use disabled() to check for GPL.
      Use ! disabled() where appropriate.
      spelling and formatting cosmetics
      Fix wrong condition check in the memalign error output.
      Use a function for repeated AltiVec output and condition check.
      simplify condition
      Clarify --help output about Vorbis and Theora.
      Clarify that libogg is for muxing only.
      Clarify usage of external libs whose functionality exists natively.
      slightly reformat --help output for ease of understanding
      Exit on failed libfaad check.
      Simplify quoting.
      BMP encoder by Michel Bardiaux, mbardiaux mediaxim be
      Simplify and speedup FFMPEG_CONFIGURATION.
      Put Visual C++ question in a better location.
      Link to the FFmpeg Windows Help website.
      cosmetics: indentation
      a bit more consistency for for the imlib2 check
      Use enabled/disabled functions where possible.
      WMA encoder
      WMA encoding
      spelling/wording
      Add seektest to list of PHONY targets.
      Replace deprecated PIX_FMT names by the newer variants.
      1000l: Revert accidentally committed change.
      missing spaces
      Add a comment about swapped numerator and denominator.
      Properly separate native and libnut NUT (de)muxers.
      cosmetics: Reorder NUT entries.
      Remove commented-out NUT muxer entry, can be added once we have a native muxer.
      Remove obsolete imgformat1 leftovers.
      happy new year
      Remove unused libac3 AC3 decoding interface.
      Force liba52 to yes if liba52bin is set.
      consistency cosmetics for the option handling routine
      Remove internal liba52; external lib still works, native decoder coming up.
      Rename ac3 decoder to liba52 to prepare for native decoder.
      Restore the possibility to link liba52 instead of dlopening.
      minor help output formatting fix
      consistency renaming: rgba32 --> rgb32
      consistency renaming: uyvy411 --> uyyvyy411
      consistency renaming: yuv422 --> yuyv422
      Ogg is not Vorbis.
      Remove unused file.
      liba52 was removed, libswscale is also GPLed.
      some updates
      Move xvmc_render.h to libavcodec where it belongs.  Blessed by Ivan.
      link fix by Dujardin Bernard, dujardin.iut numericable fr
      cosmetics: Fix a common typo, sepErate --> sepArate.
      Move MinGW extralibs setting into MinGW section.
      Remove unused variable.
      cosmetics: Fix another common typo, gurantee --> guArantee.
      simplify i386 object handling
      cosmetics: fix usefuLL --> usefuL typo
      cosmetics: aswell --> as well typo fix
      cosmetics: typo fixes, usefuLL --> useful and aswell --> as well
      Merge bfin add_ldflags lines.
      Update (lib)nut maintainers.
      Simplify .depend file inclusion.
      cosmetics: Fix another common typo, dependAnt --> dependEnt.
      Add standard license header.
      Fix wrong conditional, Snow decoding, not encoding, was SIMD-accelerated.
      Only compile snowdsp_mmx.c if the Snow decoder is enabled.
      Add LGPL header as requested by Oded.
      Collect the different thread variables in a nice easy to reuse list.
      10l: Use double, not single, quotes when variables need to be expanded.
      Simplify .depend file inclusion.
      Remove unused #define.
      spelling/grammar fixes for a Doxygen comment
      Revert previous commit that contained an unrelated hunk.
      spelling/grammar fixes for the Doxygen comments
      Remove useless conditional.
      Remove useless conditional.
      The Snow DEcoding routines should be under CONFIG_SNOW_DECODER.
      Fix wrong license header.
      Fix 'rm -r' vs 'rm -f' typo.
      Add Doxygen author and file description, rephrase a Doxygen comment.
      Remove superfluous comment.
      Add missing license headers.
      Add standard LGPL header with Fabrice's permission.
      wording/spelling
      Improve Doxygen documentation, inspired by Michael's description.
      Relicense to LGPL and add standard license header.
      Clarify that this decoder uses liba52.
      Don't mention the number of vhooks, it's a volatile set.
      Fix UNKOWN --> UNKNOWN typo
      markup/wording/spelling/formatting improvements for the last commit
      small improvements to the help output
      Remove superfluous header include.
      Recommend building qt-faststart using make instead of a manual cc invocation.
      libswscale header files are put in a subdirectory named ffmpeg, not swscale.
      misc wording/spelling fixes
      cosmetics: Fix indentation.
      blackfin: Audio buffers are no longer allocated on the stack,
      dct-test works on non-x86 now.
      Force lowercase targetos value for MinGW, we check against lowercase strings.
      Merge redundant comments.
      Add '#undef exit', the function is now forbidden.
      Always lowercase targetos.
      Use $(command) instead of `command`, it's POSIX, clearer and already used.
      support for the THP game format by Marco Gerards, mgerards xs4all nl
      Fix copy and paste error in comment.
      Replace forbidden exit() by return() Víctor Paesa, wzrlpy arsystel com.
      bfin optimizations
      spelling fix as suggested by Mans
      Remove #if 0 code.
      cosmetics: Sort list of input/output formats.
      cosmetics: Better variable name, the ENABLE_* lists contain more than codecs.
      Explain which configure options are necessary for which AMR variant.
      Audio (de)muxer dependency on OSS is handled by configure.
      cosmetics: Put audio (de)muxer entries in alphabetical order.
      CONFIG_NETWORK depends on CONFIG_PROTOCOLS, no need to nest conditionals.
      Add CPP_OBJS-yes to CPP_OBJS, same as the other *_OBJS variables.
      Get rid of unnecessary conditionals.
      cosmetics: indentation
      There were two files, libavformat/grab.c and libavformat/grab_bktr.c declaring
      cosmetics: Fix indentation.
      typo/grammar
      Add some Doxygen comments, by Kamil Nowosad, k.nowosad students.mimuw.edu pl.
      spelling/wording/grammar
      typo
      typos/grammar
      typos
      Add THP video and audio decoders.
      Remove unused variable.
      Fix compilation when Theora decoder is disabled, but VP3 is enabled.
      Remove #if 0 code.
      cosmetics: Move one code block to save an #ifdef in the next commit.
      Only compile in Theora-specific functions if the Theora decoder has been
      cosmetics: indentation
      Update download URL with a link to a working version.
      Update AMR instructions for last round of changes.
      typo
      Add complete instructions for the brave people wishing to fiddle with the
      Add a note about how to enable external libraries.
      spelling
      cosmetics: Fix indentation.
      Remove redundant 2.95 branch from the Darwin gcc check.  It is already
      Simplify Darwin gcc flags setting.
      Update preprocessor directive comments to take recent changes to this file
      Move AC3 header parsing code together with the rest of the AC3 parsing code.
      The AC3 demuxer depends on the AC3 parser.
      Make the declaration of AVInputFormat ac3_demuxer conditional
      Mention configure parameters for enabling AMR.
      Doxygen comment wording/spelling
      Remove the build instructions from the doxygen comments, they are duplicated
      typo
      spelling/wording/punctuation
      Remove unused variable.
      Add a note about license headers to the patch submission checklist.
      Remove unused code + variable, fixes a warning.
      Move ebx_available and ebp_available from CONFIG_LIST to HAVE_LIST,
      cosmetics: Reorder endianness macros by bit depth, alignment prettyprinting.
      typo
      Remove hackish support for the non-free fixed-point AMR-NB reference code.
      cosmetics: Put AMR entry in alphabetical order with all external libraries.
      10l: Rename missed occurrences of CONFIG_EBX_AVAILABLE to HAVE_EBX_AVAILABLE.
      Policy addition: New files should have proper license headers.
      Don't declare -lm as a dependency of swscale-example.
      Remove redundant dependency from ffserver_deps.
      proper dependency on mpegts_demuxer for rtp_muxer
      cosmetics: sort()
      cosmetics: Remove trailing whitespace and tabs.
      cosmetics: Remove trailing whitespace.
      cosmetics: Align some lines.
      Mark code parts that cannot work on AMD64 due to broken relocations as such.
      cosmetics: Replace manual tab indentation by \n\t as in the rest of the file.
      cosmetics attack, part I: Remove all tabs and prettyprint/reindent the code.
      cosmetics attack, part II: Remove all tabs and prettyprint/reindent the code.
      cosmetics: Fix stray wrongly indented line.
      cosmetics attack, part III: Remove all tabs and prettyprint/reindent the code.
      cosmetics attack, part IV: Remove all tabs and prettyprint/reindent the code.
      cosmetics: Group all external library decoders and encoders together.
      cosmetics: Sort entries by CONFIG_ name.
      Remove unused variable.
      Rename dts_decoder to libdts_decoder.
      Give libamr decoders/encoders a lib prefix in the name.
      Move dca parser to its own file.
      Move VC1 parser to its own file.
      Rename oggvorbis decoder to libvorbis.
      Group benchmark items in the patch submission checklist together.
      Move H.261 parser to its own file.
      Add a note about tabs + trailing whitespace to the patch submission checklist.
      parser.c is compiled unconditionally, there is no need to compile it
      Fix warning:
      Fix multiple "‘inline/static’ is not at beginning of declaration" warnings.
      parser.c is compiled unconditionally, there is no need to compile it
      Remove superfluous #includes, parser.h now includes its prerequisites.
      Move dvbsubdec parser to its own file.
      Rename dvbsubdec_parser.c to dvbsub_parser.c.
      Move dvdsub parser to its own file.
      Add a note about maintaining alphabetical order to patch submission checklist.
      RoQ audio encoder
      Add a few explanatory comments.
      Move H.263 parser to its own file.
      Allow conditional compilation of H.263-related decoders.
      Replace general CONFIG_ENCODERS by more fine-grained CONFIG_WMV2_ENCODER.
      RoQ muxer
      Remove redundant #inclusion of common.h, avcodec.h already #includes it.
      Remove unused variable.
      Remove duplicate #includes, avcodec.h #includes common.h.
      Merge RoQ entries
      cosmetics: Group all test targets together.
      Remove redundant extern declaration of j_rev_dct that is already
      Use TARGET_ARMV6 variable instead of HAVE_ARMV6 for consistency.
      Drop unnecessary libavcodec/ prefix in #include path.
      Remove extern C declarations for C++.
      Remove extern C declarations for C++.
      Make documentation build nonrecursive: Merge it into the top-level Makefile.
      Add explanatory comments to some #endifs.
      Fix imgresample-test linking.
      Rename motion_test.c to motion-test.c, which is the naming scheme used by all
      Remove mdct.o and fft.o from fft-test prerequisites list.
      extern C declarations were removed.
      Remove obsolete and non-working DSP test program.
      typo
      Use proper license header.
      Replace multiple rules for compiling binaries by one pattern rule.
      Also remove seek_test upon clean.
      Add seektest to top-level targets, remove individual codec tests.
      Do not fail or print an error if the logfile already exists.
      'grep -q' is a GNU extension.
      whitespace cosmetics
      Remove portability comment, it should be pure POSIX now.
      --enable-amr-* --> -- enable-libamr-*
      Fix seek_test compilation for out-of-tree builds.
      Do not hide the commands make runs.
      Remove comments that are neither very enlightening nor particularly up-to-date.
      better variable names
      whitespace cosmetics
      Rename variable for consistency.
      Exit with error when running libavtest and not configured with --enable-gpl.
      10l: Revert accidental removal of unused code.
      doc/Makefile no longer exists.
      Remove redundant fastmemcpy.h #include, it is indirectly #included by avutil.h.
      Remove redundant fastmemcpy.h #include, it is indirectly #included by avutil.h.
      Mark phony swscale_error target as such.
      Rename faac_encoder to libfaac_encoder for consistency and to be able to
      Rename aac_decoder to libfaad_decoder for consistency and to be able to
      readability cosmetics
      spelling
      Remove unused variables.
      Give all wrappers for external libraries names starting with lib.
      Give names of external library decoders/encoders a lib prefix
      Update for recent file removals/renames.
      one more rename
      Consistent naming and lib prefixes for x264, xvid and mp3lame within the code.
      Add some forgotten lib prefixes to Makefile variables.
      Fix linking: emms_c is a macro, #include the correct header file.
      FAAN is not x86-specific.
      Fix linking when GPL code has been disabled.
      Only run MMX2 tests when MMX2 is available.
      '-acodec mp3' --> '-acodec libmp3lame'
      Preprocessor #defines starting with __ are reserved for the system.
      Use consistent names for all test programs.
      The server configuration file is passed as a command line parameter.
      Merge tests Makefile into the top-level Makefile.
      Remove stray commented out line, fix individual regression test dependencies.
      Remove redundant #include, avformat.h indirectly includes common.h.
      Remove unused variable.
      Simplify command with automatic variable.
      Add appropriate headers for exit and printf.
      Make main() return 0 at the end and mark as int.
      Compile all binaries with default CFLAGS.
      Simplify ffplay.o CFLAGS handling.
      whitespace cosmetics
      Move HAVE_AV_CONFIG_H definition to the Makefile like in the rest of FFmpeg.
      Use consistent paths for adding values to LDFLAGS.
      Move libpostproc-specific uninstallation command to the libpostproc Makefile.
      We do not create .d files, thus no need to remove them.
      The video hooks do not depend on the libav* libraries.
      Remove unused static function.
      Add some explanatory comments to #endif directives.
      misc spelling fixes
      misc typo fixes
      misc spelling fixes
      Revert videohook dependency change.
      Add proper license header.
      comment grammar fixes
      Clarify what libfaadbin means in the summary output.
      Clarify --enable-libfaadbin option description.
      Add missing comment markers to #endif directive.
      Add missing comment markers to #endif directive.
      Remove obsolete and misleading comment.
      spelling cosmetics
      Fix 'make checkheaders'.
      Fix 'make checkheaders'.
      Remove leading underscores from specifiers, they are reserved.
      Move comment to a slightly better place.
      CONFIG_XVID --> CONFIG_LIBXVID
      URL update, move comment to a better place.
      Add missing license header.
      Group all copyright and author notices together.
      license header consistency cosmetics
      Group all copyright and author notices together.
      license header consistency cosmetics
      UTF-8 conversion
      typos
      spelling/grammar
      Remove mangle.h, it is just a duplicate of parts of libavutil/internal.h.
      grammar error noticed by Mark Cox
      Make vhook build non-recursive.
      ISO8859-1 --> UTF-8
      ISO8859-1 --> UTF-8
      10l: The ppm video hook depends on fork().
      cosmetics: typo
      Add standard licensing header to files that lack it.
      Clarify public domain statement similar to qt-faststart.c.
      license header consistency: Remove redundant "all rights reserved" statement
      cosmetics: tabs --> spaces
      Fix two cosmetic differences between our LGPL copy and the one from the FSF.
      Rename COPYING to COPYING.LGPL, add COPYING.GPL with the text of the GPLv2.
      Clarify license requirements for contributions.
      GXF muxer is no longer GPLed.
      typo
      spelling/grammar/wording
      Add note about vertical alignment to patch guidelines.
      Remove redundant "all rights reserved" statement, OKed by Walken.
      This was rewritten by Michael, switch copyrights to him and license to LGPL.
      typo fixes
      Make video hook infrastructure conditional on video hooks being enabled.
      cosmetics: indentation
      cosmetics: Do not use braces in makefiles, use normal parentheses.
      cosmetics: Use uppercase names for installation directories in makefiles.
      Uppercase variable overlooked in last commit.
      10l: Uppercase some more overlooked makefile variables.
      cosmetics: misc typo fixes
      Ahem, fix typos overlooked in last commit.
      typo
      Identifiers starting with underscores are reserved.
      Remove redundant variable declarations.
      alphabetical order
      Remove unused variable.
      Reduce Theora decoder verbosity.
      Revert unused fixedpoint declaration.
      spelling/grammar
      Bump minor number, NUT muxer was added.
      Make native and libnut NUT muxers mutually exclusive.
      Move msmpeg4 tables from a header file to msmpegdata.c.
      Attempt to better explain how to split patches into individual
      Clarify LGPL and GPL licensing of FFmpeg.
      cosmetics: indentation fix
      spelling/wording cosmetics
      small simplification
      Identifiers starting with underscores are reserved.
      Simplify preprocessor directives.
      cosmetics: alphabetical order
      Add mipsel architecture that differs from mips in endianness.
      Revert previous unneeded mipsel patch (obsoleted by endianness detection fix).
      Rename CONFIG_DARWIN to SYS_DARWIN, it is not configurable (in FFmpeg).
      Rename CONFIG_DARWIN to SYS_DARWIN, it is not configurable (in FFmpeg).
      Replace SYS_DARWIN by the more correct __APPLE_CC__, these preprocessor
      Move Apple gcc AltiVec vector declaration syntax to libavutil.
      Move Apple gcc AltiVec vector declaration syntax to libavutil.
      Merge __MWERKS__ preprocessor directives.
      Replace SYS_DARWIN by the more correct __APPLE_CC__, these preprocessor
      Remove SYS_DARWIN preprocessor directive that is just a duplicate of a
      Fix trivial mixed declarations and code warning caused by a double semicolon.
      Remove HAVE_ALTIVEC preprocessor directives that are only compiled
      whitespace/indentation cosmetics
      prettyprinting cosmetics
      Do not wrongly mark Apple gcc workarounds as being specific to Darwin.
      cosmetics: Insert empty line for readability.
      Change SYS_DARWIN preprocessor checks to __APPLE__, they are specific
      Remove unused definition.
      Simplify preprocessor directives.
      Remove unused variable.
      Split FFmpeg documentation into general documentation and the documentation
      10l: Actually add the .libs dependency to the seek_test binary,
      Use a consistent multiple inclusion guard.
      cosmetics: Sort some lines, whitespace changes.
      cosmetics: Sort some lines, whitespace changes.
      cosmetics: Sort bitstream filters.
      cosmetics: Add some whitespace for better readability.
      Alphabetically sort library dependencies list.
      Sort one more list alphabetically.
      Replace outdated BeOS section with a note that says that it is broken.
      spelling/wording nits in some Doxygen comments
      cosmetics: Add some whitespace for better readability.
      Elaborate on licensing issues: A few files are from libjpeg and the
      cosmetics: Remove useless empty line.
      link fixes
      Pass correct CPU flags to compilation of sparc/simple_idct_vis.o.
      misc wording/grammar/spelling improvements
      spelling/wording/grammar
      warning fix:
      cosmetics: typo
      Replace manual variable setting with the disable function.
      Remove redundant ARCH_SPARC #ifdef, the file is compiled conditional to
      cosmetics: sort()
      cosmetics: typo
      spelling
      Add a proper test for VIS instruction support.
      VIS-specific code should be compiled conditional to HAVE_VIS, not ARCH_SPARC.
      Move VIS CFLAGS settings into configure.
      cosmetics: Move LDCONFIG variable to a more sensible place in config.mak.
      VIS-specific code should be enabled conditional to HAVE_VIS, not ARCH_SPARC.
      cosmetics: Sort CODEC_ID list.
      Replace HAVE_BFIN with ARCH_BFIN. Blackfin is a CPU architecture, not an
      Revert wrong ARCH_BFIN --> HAVE_BFIN change.
      typo
      Mention SPARC optimizations.
      Remove Metrowerks compiler workaround.
      Remove Metrowerks compiler workaround.
      Remove const vector macro indirection that is useless and obfuscating
      Remove pointless HAVE_ALTIVEC #ifdefs from a file that only contains AltiVec
      cosmetics: Fix indentation after last commit.
      cosmetics: Fix AltiVec spelling.
      cosmetics: Fix AltiVec spelling.
      Merge mpegvideo AltiVec code into mpegvideo_altivec.c where it belongs.
      Replace CONFIG_VORBIS_DECODER #ifdef by if (ENABLE_VORBIS_DECODER).
      cosmetics: Fix indentation after last commit.
      100l: Revert accidental commit, restore previous version.
      Fix compilation with AltiVec disabled, MPV_common_init_ppc is part of a file
      Rename MPV_common_init_ppc to MPV_common_init_altivec, the function is
      license header consistency cosmetics.
      warning fixes:
      Disable unused code; fixes several gcc warnings:
      10l: refill2() is not unused, it is used conditionally.
      Consistently use $(FOO) instead of ${FOO} in Makefile variables.
      Use single quotes to get rid of a bunch of backslash escapes.
      cosmetics: sort()
      Consistently place comments in file header.
      Add standard license header.
      Add multiple inclusion guards.
      cosmetics: Move asf_read_close() down to get rid of a forward declaration.
      misc Doxygen spelling/grammar fixes
      comment grammar fix
      Use filename as multiple inclusion guard.
      Fix #endif comment typo.
      Use filename as multiple inclusion guard.
      Add missing multiple inclusion guards.
      Add a comment that explains why this header lacks multiple inclusion guards.
      Add FFMPEG_ prefix to all multiple inclusion guards.
      Add SLIB_UNINSTALL_EXTRA_CMD for MinGW.
      Clarify comment that explains why this header lacks multiple inclusion guards.
      Explain why there are no multiple inclusion guards in these header files.
      Add multiple inclusion guards to config.h.
      Add FFMPEG_ prefix to all multiple inclusion guards.
      1l: Update wrong #endif comment.
      Rename LIB to LIBNAME for consistency.
      Rename LIB to LIBNAME for consistency.
      Simplify  one variable setting.
      cosmetics: Add #endif comment.
      Clarify comment block and rewrite in better English.
      Simplify install-progs prerequisite declaration.
      Further simplify install-progs prerequisites, no need to duplicate PROGS.
      Fix make install when PROGS is empty.
      Fix make install of man pages when PROGS is empty.
      Fix uninstall-vhook, SLIBSUF was being left out.
      Simplify install-libs prerequisite generation.
      Rename optimize variable to the slightly more meaningful optimizations.
      Rename --disable-opts option to --disable-optimizations in order to be less
      Remove unused variable, fixes warning:
      Declare libpostproc prerequisites directly instead of using a layer of
      Clarify comment about gcc -MM.
      Remove unused variables, fixes the following warnings:
      Remove redundant -DHAVE_AV_CONFIG_H from tests/seek_test command.
      Move external library codec registration to a separate section after all
      The function names of [rgb|bgr]1[56]to[UV|Y] had rgb<->bgr flipped.
      Enable debug variable only when debug code is enabled, fixes the warning:
      Remove libvorbis Vorbis decoding support.  Our native decoder is complete
      Add some parentheses to clarify operator precedence, fixes the warnings:
      Quote paths in uninstall commands where necessary.
      Keep -lvorbis in libvorbis linker flags even when only used for encoding.
      Give the libnut muxer a lib prefix in its name so it does not conflict with
      Remove disabled muxer skeleton, blessed by Mans.
      Rename ogg2.[ch] to oggdec.[ch].
      Add some explanatory comments.
      Move external library registrations into a separate block after all native
      Fix make checkheaders by adding missing #include.
      Give Ogg muxer a lib prefix in the name like we do with all other formats
      Disable unused function, fixes the warning:
      Clarify Visual Studio FAQ.
      misc spelling/grammar/wording fixes in comments
      comment typo
      cosmetics: Fix spacial --> spatial typo in function names.
      cosmetics: Fix vissual --> visual typo.
      Add comments to some #endif directives.
      cosmetics: Keep all OBJS entries on a line of their own.
      Remove unused variable, fixes the warning:
      Reorder functions to get rid of ugly forward declarations.
      Put halfpel_interpol under the same #ifdef as its usage, fixes the warning:
      Remove unused functions, fixes the warnings:
      whitespace consistency cosmetics
      main() --> main(void)
      comment spelling/grammar fixes
      spelling
      main() --> main(void)
      cosmetics: Consistently indent with 4 spaces, break some long lists into single
      Keep -L flags before corresponding -l flags.
      Fix out of tree builds after introduction of libavdevice.
      Add missing #includes to fix 'make checkheaders'.
      Remove possibly misleading comments.
      spelling cosmetics
      Comment some #endif directives.
      Fix make checkheaders, #include <inttypes.h> was missing.
      succes --> success typo
      Fix some spelling mistakes.
      Fix some spelling typos
      Misc spelling fixes, prefer American over British English.
      cosmetics: spelling fixes
      Fix make checkheaders.
      cosmetics: comment spelling fixes
      colour --> color in variable names
      Remove unnecessary quotes.
      Add input and output devices to configure output.
      Add --disable-devices to --help output.
      Add --list-indevs and --list-outdevs command line options.
      Fix 'make checkheaders' on AltiVec-enabled systems.
      Add missing declarations for AltiVec functions, fixes the warnings:
      typo/clarification
      cosmetics: comment typo fixes
      Remove unnecessary system include path from pkg-config files.
      dv1394 depends on dv_demuxer, fixes linking with demuxers disabled.
      cosmetics: Sort some more entries.
      Fix "suggest parentheses around assignment used as truth value" warnings.
      comment typo fixes
      comment typo fixes
      Fix Doxygen function parameter name.
      Fix typo in macro name: WARPER8_16_SQ --> WRAPPER8_16_SQ.
      typo: begining --> beginning
      Remove outdated comment.
      Add necessary #include, fixes the warnings:
      Fix make checkheaders.
      Remove mistakenly added multiple inclusion guards.
      cosmetics: sort()
      cosmetics: Reindent.
      Change some files to only include the necessary headers.
      Spell out WB and NB.
      Add a note about the license of libamr.
      Rename two structures, identifiers starting with _[A-Z] are reserved.
      cosmetics: Add another pair of parentheses to aid readability.
      Fix test program compilation, random() needs to be undefined.
      Remove unused variable variance.
      Remove unused variable j.
      Allow compilation of test programs when TEST is defined.
      Remove unused variable i.
      Add compilation rule for tools/trasher.
      Fix illegal identifiers, names starting with _ and uppercase are reserved.
      Fix random typos.
      Add LGPL license header.
      Fix a couple of 'return type defaults to int' and 'control reaches end of
      Add '#undef strcat' to fix compilation.
      Add compilation rule for tools/pktdumper.
      Remove bogus cast.
      Add const qualifiers to unmodified function parameters, fixes the warning:
      Move "Creating config.mak and config.h..." message down to the point where
      Rename illegal identifiers, _ followed by capital is reserved for the system.
      misc spelling/grammar fixes
      Remove some spaces to keep certain people's eyes from hurting.
      Replace puts() by dprintf where appropriate.
      Disallow puts(), av_log() should be used instead.
      Simplify preprocessor expression as suggested by Mans.
      Consistently use TEST as the preprocessor condition to enable test code.
      small grammar fix
      Remove unnecessary forward declaration.
      Remove two unused variables.
      cosmetics: Fix Benoit's ugly formatting.
      Fix warnings in test code:
      Remove unused variables in test code.
      Add #undef random and call the right function to fix test program compilation.
      Put some disabled functions that are only used in the test program in cabac.c
      Fix test program compilation: Add missing #include and update the call
      Add #undef printf and disable non-functional code to make test program compile.
      Update AC-3 FAQ entry.
      Split general problems section into general questions and usage.
      Clarify some FAQ entries.
      Do not group libamr_nb and libamr_wb together as libamr.
      Add a --enable-nonfree command line parameter similar to --enable-gpl.
      Add -pthread to extralibs and not to ldflags so that the dependency gets
      Properly remove tools on clean.
      Merge rules for tools/* into a generic rule.
      Add generic rule for all tests.
      whitespace and sorting cosmetics
      Add $(EXESUF) where necessary.
      typo
      Do not install integer.h, it is not part of the public API.
      Add Makefile rules for test programs.
      Move some lines in preparation for an upcoming commit.
      Move common test program infrastructure to common.mak.
      seektest depends on both codectest and libavtest.
      Pass argc and argv to the main function of the SDL tests.
      Print the name of the regression test that was run after finishing the test.
      Do not print an (L)GPL license statement when nonfree parts have been compiled
      cosmetics: sort ogg_codec_t list
      Remove -Werror= CFLAGS that do not have the desired effect.
      typo
      Make luma_dc_quant_i and luma_dc_quant_p const, fixes a couple of
      Update policy entry about compiler warnings to reflect ml discussions.
      Use HAVE_ALTIVEC preprocessor conditionals like in the rest of FFmpeg.
      Remove pointless #if.
      Do not use ranges with tr, the results are locale-dependent.
      cosmetics: Replace // by /* */ comments.
      Print Theora bitstream version at DEBUG, not at VERBOSE level.
      Disentangle nested preprocessor directives.
      Mark MSVC compiler macros as such.
      Add fallback for DECLARE_ALIGNED and DECLARE_ASM_CONST.
      typo
      typo fix: inited --> initialized
      prettyprinting cosmetics
      Remove check for the availability of the -u diff flag, we require it in
      This shell script is not bash-specific.
      Remove stray #undef.
      Refactor the FOUROF macro using the AVV macro.
      Refactor vcprm and vcii macros by using the AVV macro.
      Explain that libavcodec/i386/idct_mmx.c is GPLed.
      Add -lm to x264 check, it is needed on most systems.
      cosmetics: Sort and prettyprint codec_wav_tags[].
      Fix outdated comment.
      cosmetics: prettyprint
      Improve wording/grammar of Doxygen comments.
      Remove commented-out av_log calls.
      Remove START_TIMER and STOP_TIMER macros.
      cosmetics: spelling/grammar
      Add comments to some #endif directives.
      cosmetics: Demplayerify indentation.
      Add explanatory comment to '#define FLOAT float'.
      Add multiple inclusion guards.
      Add empty line in output to increase readability.
      Set TCP protocol to is_streamed=1 as seeking is not possible.
      cosmetics: Consistently move NAME and FFLIBS to the top of each Makefile.
      cosmetics: Consistently move NAME and FFLIBS to the top of each Makefile.
      cosmetics: Remove useless empty line.
      cosmetics: Remove useless empty line.
      Make ffplay use parse_number_or_die instead of atoi.
      The videohook target does not depend on .libs.
      Install libavfilter if it has been enabled.
      cosmetics: prettyprinting
      Rename pp variable to postproc for consistency.
      Mention that libavdevice/x11grab.c is GPLed.
      cosmetics: Normalize {} placement after for, while, if.
      cosmetics: Fix indentation in a few more places.
      #error in case we do not know how to provide alignment.
      Work around gcc stupidity: With -MM gcc outputs just the basename of files as
      Remove redundant swScaler: output from places where av_log()
      Rename --enable-swscaler --> enable-swscale to be consistent with the
      vhooks really need libav* even if they compile fine without on Linux.
      typo
      misc spelling fixes
      Merge SDL_CFLAGS handling for files that need it.
      Replace the awkward '1>' with the more natural '>'.
      Simplify .depend generation command.
      Clarify comment.
      Mark Y variable in EPILOG macro as av_unused to avoid unused variable warnings.
      Use $^ without header files instead of $(SRCS) to reference the dependencies
      Do not calculate dependencies for header files in the .vhookdepend target.
      Use proper library versions instead of SPPVERSION in Darwin SHFLAGS.
      Move dependency generation commands into configure.
      typos
      wording improvement by the Wanderer
      typo fixes
      Add missing stdint.h #include to headers that use it.
      Add #include "config.h" to headers that need it.
      This file does not use size_t, so no need to #include <sys/types.h>.
      #include stdint.h instead of inttypes.h like all libavutil headers do.
      #include stddef.h instead of sys/types.h, the former is a standard C header.
      Build ffserver_g as well as ffserver, consistent with the other binaries.
      Use correct t printf modifier for pointer differences.
      typos
      cosmetics: Reformat Apple cc check.
      Only check for altivec.h when not using Apple gcc and not adding -faltivec to
      Remove the unused function roundToInt16. It is a duplicate of the same function
      Properly check for AltiVec CFLAGS instead of guessing based on compiler vendor.
      typo
      typo fixes
      cosmetics: #include config.h before all other FFmpeg headers.
      Use $< to simplify a few lines.
      Merge binary generation rules.
      10l: Fix stupid typo, VHOOKDEPEND_CMD --> VHOOK_DEPEND_CMD.
      Return of the living typo II: VHOOKDEPEND_CMD --> VHOOK_DEPEND_CMD
      Escape \1 and \2 in sed expressions written to config.mak to avoid mangling.
      Move functions that are only used when libswscale is not compiled in below
      Add -q to grep invocation where appropriate.
      Simplify MANGLE macro preprocessor condition checks.
      cosmetics: indentation
      Drop unnecessary subshell invocation from BEINCLUDES check.
      Fix previously broken xlc checks, simplifying them in the process.
      Fix previously broken x86_64 check, simplifying it in the process.
      Remove redundant ARCH_POWERPC #ifdef around HAVE_ALTIVEC.
      Remove redundant ARCH_POWERPC #ifdef around HAVE_ALTIVEC.
      Only #include sys/mman.h if configure set HAVE_SYS_MMAN_H.
      british_english--;
      two more typos spotted by Mike
      typo: occured --> occurred
      Include files are now installed in $PREFIX/include.
      cosmetics: Fix indentation to be 4 spaces and consistently place {}.
      cosmetics: Fix typo in variable name, numSkiped --> numSkipped.
      cosmetics: typo fixes
      Fix typo in variable name: blured --> blurred.
      cosmetics: comment typo fixes
      typo fixes: tempBlured --> tempBlurred
      cosmetics: prettyprinting
      Remove unneeded sys/mman.h #include.
      Remove #ifdef HAVE_SYS_MMAN_H around sys/mman.h #include.
      Move declaration of temporary filenames after the OS-specific settings
      Add OS-specific executable suffix to temporary executable name.
      Remove ineffectual --log option.
      Rename --log option to --logfile.
      Introduce --disable-logging option to control logging.  --logfile now just
      cosmetics: indentation
      Rename --disable-strip option to --disable-stripping. This allows
      cosmetics: alphabetical order
      --log=no is now --disable-logging.
      Invert logging variable logic so that it works without setting a default.
      Rename --cross-compile option to --enable-cross-compile.
      cosmetics: Rename targetos variable to target_os.
      Rename --enable-sunmlib option to --enable-mlib.
      Classify mlib as a configurable option, not as a hardware feature.
      Classify mlib as a configurable option, not as a hardware feature.
      cosmetics: Rename 'all' target to 'main'.
      cosmetics: Move around some rules and variable declarations.
      Merge LDFLAGS lines.
      cosmetics: Rename SRCS variable to PROGS_SRCS.
      Use common.mak in the top-level Makefile as well.
      Move CPU detection code after OS-specific settings and logging system
      Remove TAGS target. It is ages old and has never been updated to match
      Fix circular dependency between install-progs and install-libs targets.
      The install target should depend on enabled installation targets.
      Replace ifeq to declare install-progs dependencies by a variable.
      Handle EXTRALIBS and LDFLAGS for libswscale and libavfilter in a simpler way.
      Move INSTALL_PROGS_TARGETS variable to a better place.
      bandaid fix for broken shared library installation
      cosmetics: alphabetical order
      Fix wrong #endif comment.
      Make MMX2 depend on MMX.
      cosmetics: Write some trivial C test programs in a more compact form.
      Use a different filename that ends in .sh for temporary shell scripts.
      cosmetics: alphabetical order for temporary filenames
      Make XVMC compilation depend on HAVE_XVMC, not HAVE_XVMC_ACCEL.
      Remove redundant #ifdef around the complete file.
      Fix a bunch of typos in comments.
      cosmetics: Write some Doxygen comments in a more compact fashion.
      minor typo fixes
      Remove unused variable.
      10l typo, restore forgotten $.
      Fix reference to a file in MPlayer that was renamed.
      Relicense test/example files as LGPL with Michael's permission.
      Add explanatory comments to enum CodecID.
      Simplify CLEANFILES generation.
      Split CLEANFILES variable into CLEANSUFFIXES and LIBSUFFIXES.
      Generate the list of files to clean in subdirectories in common.mak.
      Unset DIRS after including subdir.mak so that the variable is not used
      Remove all dependency files on distclean.
      Clear DIRS variable in common.mak, so it does not have to be cleared in
      Add DIRS variable to the list of variables to be cleared instead of
      spelling/wording/grammar and better phrasing
      We require GNU Make 3.81.
      Print duration with higher precision.
      cosmetics: Fix nonstandard indentation.
      Set correct value for includedir in pkg-config files.
      cosmetics: __asm__ __volatile__ --> asm volatile
      cosmetics: __asm__ __volatile__ --> asm volatile
      cosmetics: Sort some lists alphabetically.
      cosmetics: spelling
      Fix des-test compilation.
      Fix build of cputest test program.
      Rename diracschro_common.[ch] to libdirac_libschro.[ch], external library
      Adjust pkg-config files to include files residing in separate subdirectories.
      Remove unnecessary test program dependency declarations.
      Rename i386/cputest.c --> i386/cpuid.c.
      Use filename as multiple inclusion guard.
      Remove unused avdevice.h header file; add config.h which is used.
      Comment out unused ff_vc1_norm6_spec table.
      Add Dirac parser from SoC; written by Marco Gerards;
      10l: Unbreak libdirac compilation; CONFIG_LIBDIRAC is defined in config.h, so
      cosmetics: Consistently use C-style comments in #endif preprocessor directives.
      Use correct length modifier for pointer diff argument in av_log() call.
      cosmetics: alphabetical order
      cosmetics: sort lines (correctly)
      Remove stray extra parameter from libavutil pkg-config command.
      In the pkg-config file generation, derive the short name from the library
      Merge both pkg-config creation commands into one.
      cosmetics: whitespace/order changes, if/then --> && in pkg-config section
      minor spelling fix
      minor wording fix
      Remove unnecessary parentheses from return calls.
      Remove one more set of useless parentheses from a return call.
      #include stdint.h instead of using a manual typedef for uint8_t.
      prettyprinting cosmetics
      #include required headers directly.
      #include required stdint.h header directly.
      string.h is a system header file, so use <> for it.
      dsputil.h is not a system header, use "" when #including it.
      ipp.h is a system header, use <> when #including it.
      Revert previous mistaken commit.
      Add required stdint.h header.
      Use full path for #includes from another directory.
      Remove redundant #undef statements that were mistakenly committed.
      Use full path for #includes from another directory.
      Remove -I flags pointing to local paths, they are no longer necessary.
      Add missing path to #include statement.
      Add missing path to #include.
      Add missing required header directly.
      ADPCM codec long names by Stefano Sabatini, stefano.sabatini-lala poste it
      cosmetics: Add whitespace around redirection operators.
      cosmetics: Fix typos in ADPCM codec long names.
      cosmetics: Use C-style comment delimiters with #endif directives.
      Add PCM codec long names; by Stefano Sabatini, stefano.sabatini-lala poste it.
      consistent codec name by Stefano Sabatini, stefano.sabatini-lala poste it
      Fix big-endian vs. little-endian typo.
      missing codec long names by Stefano Sabatini, stefano.sabatini-lala poste it
      The MPEG4AAC codec is provided by an external library and will be disabled
      cosmetics: Use C-style comments in #endif directives.
      cosmetics: typo
      consistency cosmetics: indices --> indexes
      consistency cosmetics: indices --> indexes in variable names
      Rename test-server target to servertest for consistency with other test targets.
      Simplify phony target declaration.
      Remove TAGS target cruft.
      Do not remove gmon.out on clean, we never create it.
      cosmetics: typo fixes
      cosmetics: XviD is now called Xvid.
      Call x264 and Xvid libx264 and libxvid in the configure output for consistency.
      Remove unused variable 'prefix' from config.mak.
      cosmetics: spelling typo fixes
      Add command line option to set the value of bindir.
      Add a note about everybody's favorite recurring gcc bug:
      Some (Solaris) gcc versions apparently output a comment as first line
      Remove unnecessary #include.
      10l: Add useless backslashes in AltiVec CFLAGS test.
      Simplify altivec.h test.
      Implement a proper configure test for AltiVec vector declaration syntax.
      cosmetics: Remove useless backslashes.
      cosmetics: Split all path variables into their own list.
      small spelling/wording fixes
      minor spelling/grammar fixes
      Run documentation through ispell, plus some minor fixes.
      Add -Wl,-rpath-link flags for all libraries to LDLATEFLAGS.
      Remove Fabrice's copyright from the Makefiles. They have been entirely
      OBJS should end in .o, not .c.
      Fix typo in the names of variables assigned to OS-specific path variables.
      Add missing #includes to fix 'make checkheaders'.
      Generate pkg-config files in each library subdirectory so that they do
      Add missing entry for PCM_DVD_ENCODER.
      cosmetics: alignment
      Remove unused variable, fixes the warning:
      Remove redundant mm_flags declaration, fixes the warning:
      Replace redundant MAX macro declaration by proper use of FFMAX.
      Each dependency file needs to be updated when one of the files the
      spelling cosmetics
      Place license header at the top of the file for consistency.
      whitespace cosmetics
      spelling/grammar/wording overhaul
      Remove unused variable, fixes the warning:
      Add a note to remind people that new PCM/ADPCM formats need to be added to
      spelling cosmetics
      Add required stdint.h header #include.
      Add correct header #include to fix 'make checkheaders'.
      spelling cosmetics in PCM codec long names
      Add missing dependency declaration for raw Dirac demuxer.
      Add a note to remind people to add new raw formats to the Makefile.
      Alphabetically order AVInputFormat/AVOutputFormat declarations.
      spelling cosmetics
      liba52 is no longer necessary for AC-3 decoding.
      spelling cosmetics and format name fixes
      WMV8 is fully supported now.
      Sort image formats table.
      liba52 is not always used for AC-3 decoding.
      Remove file extension from some formats, it's never mentioned consistently.
      Sort and prettyprint the file formats table.
      Sort and prettyprint video codecs table.
      Sort and prettyprint audio codecs table.
      format name cosmetics, spelling cosmetics
      Remove redundant Macromedia Flash entry in supported formats list.
      List QuickTime container as MOV/QuickTime.
      Ignore case when sorting.
      Remove wrong casts from vector declarations, this would cause errors when
      Add FFMPEG_CONFIGURATION to config.mak as well, it sometimes helps debugging.
      cosmetics: Fix inconsistent indentation.
      Sort lists ignoring case.
      Update info for Sonic, add Dirac to supported formats list.
      Move functions only used for muxing below #ifdef CONFIG_YUV4MPEGPIPE_MUXER,
      Move a function and an #ifdef CONFIG_MUXERS so that functions used only for
      Move one function that is only used for muxing below #ifdef CONFIG_MUXERS.
      Make preprocessor condition more specific and useful by changing
      Move #ifdef CONFIG_DVVIDEO_ENCODER up to cover all functions used for encoding.
      Add parisc* to the list of architectures requiring PIC flags for shared libs.
      Drop non-C99-compliant initialization.
      gifdec.c was removed, purge its entry from the Makefile.
      cosmetics: Fix mxf codec long name.
      Move up swscale_error target so that regression tests error out immediately.
      Replace LDLATEFLAGS hackery by proper LDFLAGS tests.
      cosmetics: Reformat PPC code in libavcodec according to style guidelines.
      cosmetics: Reindent two misplaced braces.
      cosmetics: Make libavcodec/ppc/dsputil_altivec.c conform to style guidelines.
      Move --disable-debug to a better place within configure help output.
      Remove AltiVec vector declaration compiler compatibility macros.
      misc spelling/grammar fixes
      Remove AltiVec vector declaration compiler compatibility macros.
      Fix indentation after last commit.
      cosmetics: Fix indentation after last commit.
      cosmetics: typo/grammar fixes
      cosmetics: Fix two common typos: wont --> will not, lets --> let us.
      cosmetics: typo fixes
      typo fix
      Move #includes, which are only used in the test program, below the #ifdef
      Change return type of main function to int to avoid a warning.
      Add random to list of test programs.
      Remove unused variable.
      Add note about gcc bugs that gcc devs refuse to fix.
      USE_FASTMEMCPY is now called CONFIG_FASTMEMCPY in MPlayer.
      Add missing header #includes.
      Only check for socklen_t if network support is enabled.
      Explain the difference between HAVE_LIST and CONFIG_LIST in a short comment.
      Replace call to deprecated avcodec_build() with avcodec_version().
      misc spelling/wording fixes
      misc spelling/grammar fixes
      Fix small typo.
      Remove unused variable.
      Remove unused variable.
      Check for sys/select.h and poll.h unconditionally, not just if ffserver
      Use '#include <poll.h>' instead of '#include <sys/poll.h>'.
      Remove hackish support for fastmemcpy from MPlayer.
      cosmetics: misc spelling fixes
      cosmetics: Remove redundant #endif comments that are very close to the #ifdef
      Replace generic CONFIG_MUXERS preprocessor conditions around AVOutputFormat
      Replace generic CONFIG_MUXERS preprocessor condition around format-specific
      Surround AVInputFormat declarations with format-specific #ifdefs.
      Surround format-specific functions with matching preprocessor conditionals.
      Mark adx_decode_init() as type int instead of void, the function returns
      Add missing stdint.h header to fix 'make checkheaders'.
      Add missing header includes to fix 'make checkheaders'.
      Mark init_vlcs function as av_cold.
      Untangle mpeg12.c and mdec.c so that mdec.c can be compiled separately.
      cosmetics: comment spelling/grammar fixes
      Rename ff_init_vlcs function to the more descriptive name ff_mpeg12_init_vlcs.
      Fix #include path, headers from other directories need to have
      cosmetics: Fix incorrect indentation.
      Add missing FFMPEG_ prefix to multiple inclusion guard.
      Remove pointless #ifdef CONFIG_VHOOK from the code.
      Replace generic CONFIG_MUXERS preprocessor conditionals by more specific
      Silence a couple of 'defined but not used' warnings by adding an av_unused
      Replace generic CONFIG_DECODERS preprocessor conditionals by more specific
      Replace generic CONFIG_ENCODERS preprocessor conditionals by more specific
      Surround some encoding-specific functions with the appropriate
      Disable encoders by undefining CONFIG_FOO_ENCODER once instead of littering
      license header consistency cosmetics
      Mention that libavcodec/ac3dec.c is GPLed.
      license header consistency cosmetics.
      Relicense from LGPL 2.1 to LGPL 2.1 or later with the authors' permission.
      Relicense as ISC with the authors' permission.
      spelling/wording cosmetics
      misc spelling/wording/grammar fixes
      Fix typos found by ispell.
      Only define x86 register names on x86.
      Remove useless #ifdef around #include.
      cosmetics: Fix indentation.
      wording/spelling
      small wording fixes
      Rename check_func2 to check_func_headers.
      spelling cosmetics
      Add MIPS to list of architectures requiring PIC flags.
      Remove unused variable.
      whitespace cosmetics
      Use better table heading names in the supported formats tables.
      Add support for passing a version string to configure that is appended to
      Replace 'test' invocation by 'test -n', this should be safer.
      Add alltools target to build all the tools/ stuff.
      Move doxygen comments for av_gettime from utils.c to avformat.h.
      spelling/wording/grammar
      cosmetics: Break overly long lines, one grammar fix.
      cosmetics: indentation
      Remove unused function fast_memcpy.
      cosmetics: Consistently use #ifdef in preprocessor directives.
      Uniformly define _XOPEN_SOURCE to 600.
      Globally set LC_ALL=C to prevent locale issues from breaking tests.
      Update copyright years.
      cosmetics: Change comment style for consistency.
      spelling cosmetics
      Remove offset_t typedef and use int64_t directly instead.
      FFmpeg is not made for bats: Fix 22050Hz vs. 22050kHz typo.
      Remove MPlayer-specific MP_WORDSIZE hack.
      Fix 'make checkheaders', based on a patch by Diego Pettenò, flameeyes gmail com.
      Merge variable declaration and export.
      Use HAVE_FAST_64BIT instead of nonstandard __WORDSIZE macro.
      Remove obsolete liba52 wrapper code.
      Sync cosmetics from upstream.
      Add a bunch of const qualifiers to function arguments.
      Revert the merger of LC_ALL declaration and export.
      Replace sed trickery in the gcc dependency generation command by use of
      Update DEPEND_CMD comment to better explain why we postprocess gcc output.
      Remove -fasm CFLAGS check, it is no longer needed.
      cosmetics: Save some backslashes.
      cosmetics: Leave out quotes from libdirac pkg-config call for consistency.
      Add information about a workaround for MinGW gcc versions.
      cosmetics: typo fix
      spelling cosmetics
      prettyprinting cosmetics
      Move declaration of mlp decoder dependency on mlp parser to the Makefile
      cosmetics: Change the order of mlp parser objects to be consistent.
      cosmetics: Uniformize external library comments.
      Add GSM into list of supported audio codecs.
      Mark formats requiring external libs with an 'E' in the format support tables.
      cosmetics: Slightly update MP3 support entry.
      Add subtitle format section.
      Add the function declaration of ff_svq1_packet_checksum to svq1.h and include
      Add 'make checkheaders' to developer guidelines and patch checklist.
      Revert adding controversial 'make checkheaders' item to the development policy.
      Add missing headers to pass 'make checkheaders'.
      asm() --> __asm__() in #warning comment
      Remove duplicated patch submission checklist entry by putting the more
      Add check target to run both tests and checkheaders targets.
      cosmetics: small spelling fixes
      Remove nonsense @return Doxygen comment for void function.
      Initialize variable to silence the warning:
      Fix illegal identifier, __ are reserved for the system.
      Add tests target for libswscale test programs.
      Merge ARCH_BFIN lines.
      Add some const specifiers to function name variables; fixes a bunch of
      Fix test program build: ff_eval was replaced by ff_eval2.
      Add missing #include "libavutil/avstring.h", fixes
      Clarify that we require GNU make 3.81+.
      Remove unused variable.
      Put variable declaration inside an #ifdef to avoid an unused variable warning.
      cosmetics: Reorder processor-specific entries alphabetically.
      cosmetics: Group ARM objects sections together.
      Print just [no] or [yes] in the help output instead of [default=no/yes].
      spelling and wording cosmetics for the help output
      Move zlib/bzlib options to the external library section of the help output and
      cosmetics: Move some options to more sensible places within the help output.
      Add missing options to the help output.
      cosmetics: Group ppc-specific objects together.
      Only the yasm-dependent objects should be enabled by a yasm conditional.
      Revert previous commit: the libavcodec/i386/fft_*.o objects depend on
      Add a Makefile variable for the old scaler that gets enabled when the new
      cosmetics: Sort processor-specific object lines into alphabetical order.
      cosmetics: Add a few more explanatory comments to the object sections.
      Use standard OBJS-$(HAVE_MMX) trick for MMX objects instead
      cosmetics: indentation
      Move FFT-specific MMX optimizations out of the 'ifdef HAVE_MMX' block.
      cosmetics: Consistently place HEADERS before OBJS in all Makefiles.
      cosmetics: Consistently place HEADERS before OBJS in all Makefiles.
      Get rid of last ifdef HAVE_MMX. Instead, all MMX-specific objects are gathered
      Move H.264 tables that are only useful for encoding to h264enc.c.
      cosmetics: Put statements after 'if' on their own line.
      cosmetics: consistent function declarations
      cosmetics: indentation, prettyprinting
      cosmetics: Align table for better readability.
      Use golomb_to_pict_type instead of its duplicate, slice_type_map.
      Move #defines that are mostly used in h264.c out of h264data.h and into h264.h.
      Make non-void replacement functions actually return a value.
      Add missing headers to get 'make checkheaders' closer to passing.
      Disable unused code for now to allow 'make checkheaders' to pass.
      Move encoder-specific quantize_c and related tables to the H.264 encoder.
      cosmetics: Rename aanscales to ff_aanscales, it will soon be externally visible.
      Move aanscales tables to their own file; fixes compilation without encoders.
      Automatically check the dependencies for the AAN DCT tables.
      Only build the imgresample test if the old scaler was enabled.
      Add missing '-test' in name of imgresample test.
      Do not duplicate AAN DCT tables in dct-test.c.
      Remove ineffectual hack that attempts to build ppc/check_altivec.o without
      Handle AltiVec objects in a more elegant fashion, similar to MMX.
      Remove unused declarations.
      Remove commented-out duplicate declarations.
      Remove entry about unsupported QCELP codec, it is now supported.
      Delete unnecessary 'extern' keywords.
      Delete unnecessary 'extern' keywords.
      Move declaration of av_set_program_name and av_program_add_stream_index
      Add necessary header for av_set_program_name and av_program_add_stream_index.
      Do not set COMPILE_C if AltiVec is enabled without runtime CPU detection.
      Drop _t from typedef names for POSIX compatibility.
      Add backwards-compatibility typedef for removed _t names under appropriate
      Add alternative names for sh4 and parisc architectures.
      Add mips64 to the architecture list.
      Add img_convert() to imgconvert.h and #include it from imgresample.c.
      Replace 'typedef struct ustr_t' by 'struct ustr', _t is POSIX namespace.
      whitespace cosmetics
      Make mips64 a separate arch with fast_64bit.
      Remove misleading comment, _t is POSIX-reserved namespace.
      cosmetics: Break some lines for a more aesthetically pleasing result.
      100l: Add missing backslashes.
      Do not invade _t POSIX namespace.
      cosmetics: typo/grammar/wording/whitespace fixes
      Avoid POSIX-reserved _t in identifier names.
      Avoid invasion of POSIX-reserved _t namespace.
      whitespace cosmetics
      Switch xvmc over to a CONFIG_ option.
      Add xvmc configure support.
      Roundup now works from a ffmpeg.org domain as well, so prefer that.
      Remove obsolete and misleading comments.
      Remove pointless malloc.h #include.
      Remove pointless malloc.h #include.
      Do not use full include path for headers in the same directory.
      Use ARCH_X86_32 instead of !ARCH_X86_64, it is more straightforward.
      Remove a bunch of unused variables.
      Rename libavcodec/i386/ --> libavcodec/x86/.
      Fix sample URL.
      Change mplayerhq.hu references to ffmpeg.org where appropriate.
      ASAO Nelly Moser implementation moved to a better place on mphq.
      consistency cosmetics: Rename POWERPC identifiers to PPC.
      consistency cosmetics: Rename POWERPC identifiers to PPC.
      Avoid POSIX-reserved _t namespace.
      Avoid POSIX-reserved _t namespace.
      Avoid POSIX-reserved _t namespace.
      spelling/grammar and whitespace cosmetics
      Add missing 'void' keyword to parameterless function declarations.
      Revert deprecation of AVFrac structure. Attempts to replace AVFrac have turned
      cosmetics: alphabetical order
      Remove inappropriate const qualifier from function argument, fixes the warning:
      alphabetical ordering cosmetics
      Disable SVQ3 checks if SVQ3 support is not compiled in.
      Use <> instead of "" for system headers.
      Add missing headers to allow 'make checkheaders' to pass.
      The SVQ3 decoder uses ff_svq1_packet_checksum from svq1dec.c.
      Only #include svq3.c if the SVQ3 decoder is enabled.
      Add missing --disable-armv6t2 to help output.
      1l: better alphabetical order
      cosmetics: Remove useless quotes.
      Add missing #include to fix 'make checkheaders'.
      Drop unnecessary intreadwrite.h and bswap.h #includes.
      Add necessary, remove unnecessary #includes.
      Fix illegal identifier starting with an underscore.
      cosmetics: indentation
      Add my GPG key.
      Use 'enabled' function instead of raw 'test' calls in arch checks.
      Fix build: Add intreadwrite.h and bswap.h #includes where necessary.
      Use 'enabled' function to simplify x86_64 check in the mingw32 section.
      Split arch variable into arch and subarch.
      Extend 'checkheaders' target to architecture-specific subdirectories.
      Remove pointless #if around header #includes.
      Use CONFIG_MPEGAUDIO_HP directly instead of USE_HIGHPRECISION indirection.
      Remove one more forgotten instance of USE_HIGHPRECISION indirection.
      Replace #ifdef CONFIG_ preprocessor check by #if CONFIG_.
      cosmetics: Remove pointless period after copyright statement non-sentences.
      Restore mistakenly removed period after "DivX, Inc".
      Use "" instead of <> when #including non-system headers.
      Add void keyword to parameterless function declaration.
      Make base64 test program compilable as a standard test program.
      Avoid unused variable warning when compiling DES test program.
      Add required headers to fix 'make checkheaders'.
      Fix undefined preprocessor directives warnings during 'make checkheaders'.
      Directly #include a bunch of indirectly #included headers.
      Remove pointless #if HAVE_ALTIVEC around internal header #include
      Add a check for ppc4xx instructions; rename preprocessor directive accordingly.
      Use cc_default variable in help output for --cc option.
      Remove --enable-static and --disable-shared from help output.
      help output cosmetics
      Add configuration options section to help output.
      Add required headers to fix warnings during 'make checkheaders'.
      Fix #endif comments.
      spelling/grammar cosmetics
      Disable C code when compiling AltiVec code, fixes the warning:
      Declare struct SwsContext before using it, fixes the checkheaders warning:
      Move internal.h #include to the end of the file so that it is after
      Remove redundant av_log2_16bit forward declaration, fixes checkheaders warning:
      Do not #include common.h without necessity.
      SSE depends on x86.
      Add missing architecture-specific options to summary output.
      Directly #include some required headers.
      Move DECLARE_ALIGNED and DECLARE_ASM_CONST to internal.h.
      Drop deprecated av_fifo_write function with the next libavutil version bump.
      Drop _MSC_VER case from macro declaration.
      Drop redundant architecture extension dependencies.
      Add AMD 3DNow! instruction enabling/disabling support in the build system.
      Restore _MSC_VER case in DECLARE_ALIGNED/DECLARE_ASM_CONST macro declaration.
      HAVE_3DNOW --> HAVE_AMD3DNOW to sync with latest configure changes.
      HAVE_3DNOW --> HAVE_AMD3DNOW to sync with latest configure changes.
      Add required limits.h header.
      Disambiguate the macros used to disable fprintf/printf/puts functions.
      C files should #include the header files of the same name.
      10l: Add #undefs for system free/malloc/realloc, which must be used here.
      Reorganize header #includes: Unconditionally #include standard C headers,
      HAVE_3DNOW --> HAVE_AMD3DNOW
      version.h depends on version.sh.
      Remove misleading comment about how to build SHA1 test program.
      Fix filenames in Doxygen comments.
      spelling/grammar/consistency review part I
      Add common.h #include, necessary for av_const.
      spelling/grammar/consistency review part II
      spelling/grammar/consistency review part III
      Rename string.c to avstring.c so that the name of header and C file match.
      Add missing void keyword to parameterless function declarations.
      Add 'void' keyword to parameterless function declaration.
      cosmetics: Move all #includes together at the beginning of the file.
      Add required header mem.h, av_malloc and friends are used.
      Use '#if defined()' for OS-specific preprocessor checks.
      cosmetics: Use 'num' instead of 'nom' as abbreviation for numerator.
      Fix filenames in doxygen comments.
      Use full internal pathname in doxygen @file directives.
      Use full internal pathname in doxygen @file directives.
      Remove H.264 encoder initialization.
      cosmetics: Remove period after copyright statement non-sentence.
      cosmetics: Remove period after copyright statement non-sentence.
      Add av_unused to conditionally used variable, fixes the warning:
      Add necessary #includes to pass 'make checkheaders'.
      cosmetics: Place all subtitle entries together.
      Add XSUB to list of supported subtitle formats.
      Add necessary header for LIBAVUTIL_VERSION_MAJOR, fixes the warning:
      Use directory name as multiple inclusion guard prefix.
      fastdiv is disabled by default, so put --enable-fastdiv into the help output.
      Remove pointless '#if CONFIG_ENCODERS' around forward declarations.
      Remove unused CONFIG_H264_ENCODER #define.
      Remove documentation of non-existing function parameters.
      Remove const qualifier from function argument to eliminate the warning
      cosmetics: comment grammar/spelling fixes
      Remove rendering column from list of supported subtitle formats.
      typo: CODEC_ID_cljr --> CODEC_ID_CLJR
      Remove unused variables.
      Add av_cold attribute to init function.
      Remove unused variables.
      cosmetics: Fix a few typos and use a consistent codec long name.
      cosmetics: typo/grammar fixes in doxygen and other comments
      Remove file name from file headers.
      Remove file name from file headers.
      Remove svn:executable property from source file.
      Drop DECLARE_ALIGNED from extern declarations.  It creates trouble when
      Add #undefs to reenable system functions that are normally forbidden in other
      Add config.h #include for ARCH_X86 definition.
      Revert #undefining system functions, it is not necessary.
      Use AV_GCC_VERSION_AT_LEAST instead of complicated __GNUC__ version check.
      Remove unnecessary emms Assembler instructions.
      Add check whether the compiler/assembler supports 10 or more operands.
      Rename vdpauvideo.c --> vdpau.c and xvmcvideo.c --> xvmc.c to better fit
      Add av_uninit macro to variable to avoid false positive warning:
      Add av_uninit macro to variable to avoid false positive warning:
      Add av_uninit macro to variables to avoid false positive warnings:
      Add av_uninit macro to variable to avoid false positive warning:
      Fix wrong file name in header, noticed by David DeHaven, dave sagetv com.
      Rename xvmc_render.h --> xvmc_pixfmt.h and vdpau.h --> vdpau_pixfmt.h as
      10l: Rename vdpau.h --> vdpau_pixfmt.h in the Makefile as well.
      Remove reference to renamed header file xvmc_render.h.
      cosmetics: format long_name spelling fixes
      Use more descriptive format long_names.
      Use more descriptive format names, similar to the format long_names.
      Add entries that were missing from the supported file formats list.
      Add raw (de)muxers to list of supported file formats.
      Fix typo: vido --> video
      Remove pointless #if around the file, it is already conditionally compiled.
      Add supported network protocols section.
      cosmetics: alphabetical order
      Revert adding a _pixfmt to vdpau/xvmc header names.
      Make xvmc.h an installed header, applications like MPlayer need it.
      Rename xvmc.c --> mpegvideo_xvmc.c as suggested by Ivan.
      Remove redundant CONFIG_XVMC option, CONFIG_MPEG_XVMC_DECODER suffices.
      100l: Remove stray character that slipped into header name.
      Remove commented-out code.
      whitespace cosmetics: Fix indentation depth.
      whitespace cosmetics: consistent (more or less) K&R style
      whitespace cosmetics: another round of formatting consistency fixes
      whitespace cosmetics: vertical alignment
      Fix compilation: Restore mistakenly removed constant definitions.
      Update multiple inclusion guards to match filename.
      Replace all (x == NULL) or (x != NULL) in assert and if conditions by !x and x.
      Give public #defines that are used by MPlayer an AV_ prefix.
      Add avcodec.h #include so that LIBAVCODEC_VERSION_MAJOR is not undefined.
      Fix non-Doxygen comment.
      whitespace cosmetics: vertical alignment
      Replace useless X11 #includes by the one that actually provides the required
      whitespace cosmetics: Restore vertical alignment.
      Fix compilation: XVMC_ --> ff_xvmc prefix for xvmc functions.
      Rename some forward declarations that were missed, fixes the warnings:
      Remove some useless parentheses.
      cosmetics: Consistently format all if/else statements in K&R style
      Drop two more sets of useless parentheses.
      Remove compatibility layer for old names of xvmc definitions.
      whitespace cosmetics: Correctly indent all comments.
      whitespace cosmetics: consistently format all comments
      Restore one set of parentheses to avoid the warning:
      Remove useless comment.
      cosmetics: grammar/spelling/wording fixes in comments
      some more K&R whitespace cosmetics
      Remove one more commented-out line.
      Make one comment slightly clearer and more readable.
      Remove a comment that references a non-existing function.
      whitespace cosmetics: Put some spaces around operators for better readability.
      whitespace cosmetics: Align some comments.
      another (last?) round of K&R whitespace cosmetics
      Add xvmc_internal.h that contains all internal xvmc function declarations.
      Attempt to clarify draw_horiz_band description and fix two random typos.
      cosmetics: K&R style
      cosmetics: Change '8*8' to '64'.
      cosmetics: comment wording/grammar
      Merge some declarations and initializations.
      Clarify ff_xvmc_field_start() comment.
      Doxygenize ff_xvmc_field_start() comment.
      whitespace cosmetics: remove stray space
      Add Doxygen comment similar to ff_xvmc_field_start() to ff_xvmc_field_end().
      Add avcodec.h #include so that LIBAVCODEC_VERSION_MAJOR is not undefined.
      whitespace consistency cosmetics
      Remove completely unsed structure fields mc_type and chroma_format.
      Doxygen comment/explanation for ff_xvmc_init_block().
      Drop struct members display_flags and state with the next major version bump.
      Remove asserts that are immediately followed by a check and error return.
      Remove redundant assert(render) in ff_xvmc_init_block().  The immediately
      Replace assert in ff_xvmc_field_end() by av_log call at level AV_LOG_ERROR.
      10l: Remove duplicated hunk of code.
      K&R whitespace cosmetics
      Rename xvmc_pixfmt_render structure to xvmc_pix_fmt.
      Explain how AV_XVMC_RENDER_MAGIC got its seemingly arbitrary value.
      comment spelling/grammar/wording fixes
      spelling/grammar/wording fixes for Doxygen comments
      cosmetics: typo fix
      Rename AV_XVMC_RENDER_MAGIC constant to AV_XVMC_ID to reflect a similar
      spelling/grammar/wording/whitespace for Doxygen comments
      draw_horiz_band() is really ff_draw_horiz_band().
      spelling/wording fixes for the Doxygen comments
      Doxygen spelling/wording fixes
      cosmetics: Keep lines below 80 characters where sensibly possible.
      cosmetics: one more wording fix
      Give struct members more sensible names:
      whitespace cosmetics
      Make format long_names consistent.
      Doxygen comment wording fix
      Replace two asserts by checks and error messages.
      Add missing object dependencies: eatqi.o depends on mpeg12.o.
      whitespace cosmetics: Place spaces around += for better readability.
      Get rid of some '#if CONFIG_MPEG_XVMC_DECODER' preprocessor checks around if
      whitespace cosmetics: Reindent a few lines after the last commit.
      cosmetics: 'const static' --> 'static const' to avoid warnings of the type
      cosmetics: 'void/int inline' --> 'inline void/int' to avoid warnings of the type
      Add missing #includes to fix 'make checkheaders'.
      spelling/grammar/wording fixes for the Doxygen comments
      Revert converting two asserts into if checks and error messages.
      small Doxygen comment grammar fix
      Replace sizeof(short) by the more robust sizeof(*s->pblocks[i]) in memset and
      mpegvideo_xvmc.o depends on mpegvideo.o.
      Add support for enabling/disabling AMD 3DNow! extended optimizations.
      Handle conditional compilation of CPU-specific FFT code in the Makefile.
      The 3DNow! and SSE FFT optimizations depend on the yasm optimizations in
      prettyprinting whitespace cosmetics
      Directly #include required X11/XvMC header.
      Only enable SSE/3DNOW optimizations when they have been enabled at compilation.
      Mark functions only used within the test programs as static.
      Fix swscale-error vs. swscale_error typo; it is marked as the former in .PHONY.
      Bail out if regtest-flashsv is run but zlib is not enabled.
      Revert adding av_uninit to out_width/out_height variables.
      Fix gcc detection: Not all gcc versions have "(GCC)" in the version string.
      Fix xvmc object dependencies.
      Updates notes about which files are GPL.
      Check if -lrt is required for nanosleep, as is the case on Solaris.
      Mark two variables with av_uninit to avoid false positive gcc warnings.
      Remove GPL version of yuv2rgb.c that has been replaced by an LGPL substitute.
      Update information regarding LGPL libswscale.
      The software scaler is no longer GPLed.
      Relicense AltiVec optimizations as LGPL with the permission of Marc Hoffman
      100l: Revert removal of GPL check for libswscale.
      Rename yuv2rgb2.c --> yuv2rgb.c.
      cosmetics: Use $() instead of ``, the former can be nested more easily.
      Read revision string from the file snapshot_version if available.
      10l: Ignore error output from failed attempt to read snapshot_version.
      spelling/wording/grammar cosmetics
      spelling/wording/grammar cosmetics
      Add release version number.
      Targa encoding is supported.
      Integer-only decoding of AC-3 is not supported.
      Add some missing codecs and update names to match the codec long_name.
      Merge MP2, MP3 and MPEG audio demuxer entries.
      Add some more missing codecs and update names to match the codec long_name.
      Another round of missing codecs and name updates to match the codec long_name.
      Decoding of PAM, PBM, PGM, PGMYUV and PPM image formats is not supported.
      Change a bunch of codec long_names to be more consistent and descriptive.
      xvmc works with MPEG-2 as well as MPEG-1, note this in the codec long_name.
      small spelling/grammar fixes
      cosmetics: Reformat long_names so that "PCM" comes first.
      cosmetics: Reformat long_names so that "ADPCM" comes first.
      cosmetics: Reformat long_names so that "DPCM" comes first.
      Revert previous removal of PAM, PBM, PGM, PGMYUV and PPM from the list of
      cosmetics: spelling/wording fixes
      Mention non-recursive Makefiles in the changelog.
      Create 0.5 release branch.
      Output 0.5 as version string.
      Peg libswscale to the revision corresponding to the moment the branch was cut.

Diego Pettenò (51):
      Fix order of operands in lrintf test to make the test not fail if you have
      Move initialisations and internal symbols in allformats.h,
      make sure NDEBUG is not defined already before defining it
      Disable w53 and w97 cmp methods when snow encoder is disabled
      Remove parameter unused in format string.
      call faacDecClose() with all libfaad versions
      Make filterDelimiters and optionDelimiters two static constant array of
      Make pp_get_mode_by_name_and_quality accept a constant string.
      Make pp_help a constant pointer to constant characters, moves it partially
      Make AVRtpPayloadTypes static and constant
      Remove unused symbol.
      Make v4l.c's video_formats constant and static.
      Mark the tables in g726.c as constant.
      Make the av_class member of PPContext a poiner to constant AVClass.
      Move wmv1_scantable to .rodata section by making it an array of arrays.
      Mark the ff_svq1_frame_size_table as constant.
      Remove clip_table as it is not used anymore.
      Reduce the size of the replaceTable entries.
      Make pp_help a constant array of characters to move it to .rodata.
      Make MMX vectors constants.
      Build both static and shared libs from the same object.
      Make .vhookdep depend on version.h, same as .depend.
      Mark symbols as static, patch by Diego 'Flameeyes' Pettenò, flameeyes gmail com
      Mark symbols as static, patch by Diego 'Flameeyes' Pettenò, flameeyes gmail com
      Mark symbol as static, patch by Diego 'Flameeyes' Pettenò, flameeyes gmail com.
      Fix 'comparison always false' warning.
      Use strcasecmp() instead of re-inventing it.
      fix -uninstalled.pc includedir for out-of-tree builds
      Remove some format string warnings.
      Use enum typers instead of int.
      Remove duplicate "const" qualifier for *matrix.
      Intel C compiler warns on this assignment in this if(), probably
      fix warning reported by Intel C compiler:
      Use __typeof__ instead of GCC-specific typeof keyword.
      Make sure strcasecmp() is declared.
      Stop Sun Studio from issuing a warning during configure when passed
      Mark ffplay internal variables static. Patch by Diego Pettenò
      Mark print_all_lib_versions() static in cmdutils.c . Patch by Diego Pettenò
      Mark list heads static. Patch by Diego Pettenò
      Mark ff_div6 and ff_rem6 static in h264.c. Patch by Diego Pettenò
      Mark the ff_find_best_tables symbol static to msmpeg4. Patch by Diego Pettenò
      Declare ff_pw_53 and ff_pw_18 as assembler constants.
      Invert logic for the single-pass in swScale() functions.
      Mark variation-specific interleaveBytes static.
      Mark dither_2x2_{8,4} static to swscale.c
      Mark some symbols in swscale.c as constant.
      Use CONFIG_* macros in cmdutils.c for consistency.
      Convert asm keyword into __asm__.
      Convert asm keyword into __asm__.
      Include headers for optional libraries only when enabled.
      Rename template included sources from .h to _template.c.

Dieter (9):
      -mdynamic-no-pic for darwin -> 5% speedup patch by (Dieter Shirley <dieters at schemasoft dot com>)
      crop optmization patch by (Dieter Shirley <dieters at schemasoft dot com>)
      practically disabling altivec resampling code (some ppl said its broken) patch by (Dieter Shirley <dieters at schemasoft dot com>)
      make some functions static (patch by Dieter < freebsd at sopwith.solgatos.com >)
      long is better than int for pointer differences
      add missing #includes
      add static keyword to some functions
      add static keyword to some functions
      Use gmake instead of make on *BSD.

Dmitry Antipov (1):
      Add Intel IWMMXT2 link, patch by Dmitry Antipov, dmantipov yandex ru.

Dmitry Baryshkov (2):
      attribute used patch by (mitya at school dot ioffe dot ru (Dmitry Baryshkov))
      attribute_used patch by (mitya at school dot ioffe dot ru (Dmitry Baryshkov))

Dmitry Borisov (1):
      Mpeg start codes patch by ("Dmitry Borisov" <jbors at mail dot ru>)

Dominik Mierzejewski (13):
      warning patch by (Dominik Mierzejewski <dominik at rangers dot eu dot org>)
      warning patch by (Dominik Mierzejewski <dominik at rangers dot eu dot org>)
      Fixes:
      Fixes:
      Fixes:
      Fixes:
      Fixes:
      Use proper PRI?64 macros for %ll? format to fix printf format warnings:
      Remove redundant Requires: in libswscale.pc, approved by Diego.
      add a link to VIS instructions whitepaper
      Enable CMOV usage on x86_64 arch. All x86_64 CPUs have it.
      Remove duplicated MM_* macros for CPU capabilities from dsputil.h.
      Fix typo in comments.

Dominique Leuenberger (1):
      Fix a compiler warning:

Doug Larrick (1):
      Enable decoding of more mpeg-ts streams for pcHDTV patch by (Doug Larrick <doug at ties dot org>)

Drew Hess (3):
      avcodec const correctness patch by (Drew Hess <dhess at ilm dot com>)
      recommit of
      fixes for PIC code on x86-64 patch by (Drew Hess <drew.hess gmail com>)

Dujardin Bernard (4):
      FAQ clarification concerning the configuration of FFmpeg for usage with liba52.
      Doxygenize some comments
      move doxy comments from rational.c to rational.h and add some new comments
      Doxygen documentation for all functions, patch by Dujardin Bernard,

Earl Levine (1):
      Fix segfault on Mac OS X, patch by Earl Levine elevine**@**gmail**.**com.

Eddie Pang (2):
      10l: Correctly use preprocessor conditionals.
      Fix MSVC identification, patch by Eddie Pang.

Edouard Gomez (7):
      Adapt the patch to more recent FFmpeg habits
      RFC 1 stage: Improve maintainability of the cursoir painting code
      RFC 2 stage: First feedback from review comments
      RFC 3&4 stage: FFmpeg style aint't my style
      RFC 5: Mouse cursor painting reduc and more cleanups
      RFC 6: Overall cleanups again.
      Feedback from Karl H. Beckers about mouse cursor

Eduardo Bragatto (1):
      Initialize correctly optind.

Edward Hervey (1):
      DV video encoder/decoder doesn't set keyframe and picture type properties patch by (Edward Hervey:bilboed,gmail com)

Eli Friedman (4):
      Escape 124 (RPL) decoder
      ARMovie/RPL demuxer
      Remove flexible array member from Escape 124
      Add support for fourcc "SP54".

Emiliano Parasassi (3):
      disabling all decoders patch by (Emiliano Parasassi, millallo : tiscali it)
      avoid the registration of DECODERS if only CONFIG_DECODERS is not defined.
      Fix compilation with --disable-decoders.

Eric Lasota (3):
      Convert RoQ decoder to use YUV 4:4:4 unpacked macroblocks
      Rename cb field
      Cygwin compatibility workaround

Eric Paul (1):
      check udp_read_packet return value not to access uninitialized memory

Eric Sesterhenn (1):
      remove unused varibles patch by (Eric Sesterhenn <SnakeByte at gmx dot de>)

Erik Auerswald (1):
      10l in rgb2rgb_template.c

Erik Hovland (6):
      Remove unnecessary header #includes.
      fix feed_filename test, field is char [1024], patch by Erik Hovland, erik at hovland dot org
      One-bit bitfields should probably be unsigned. Unfortunately, it is left
      Fix a mem leak in av_find_stream_info().
      Fix a mem leak in vc1_decode_frame().
      Check the allocated pointer instead of the given pointer as intended

Erik Slagter (3):
      This trivial patch adds "avc1" as fourcc for h264 in mp4, which makes it
      Increase maximum number of B-frames, necessary for H.264 encoding.
      pass noise reduction option to x264

Erwan Ducroquet (1):
      Add examples in documentation showing how to avoid to decode audio and

Evgeniy Stepanov (5):
      export raw SSA tracks with their own codec ID (patch by Evgeniy Stepanov)
      Add support for Matroska attachments.
      Add 'disposition' bitfield to AVStream and use it for both muxing and demuxing
      matroskadec: move buffer decoding code to a separate function
      matroskadec: decode track private data if needed

Fabien Cosse (1):
      Add fourcc "SEDG" for MPEG-4 files produced by Samsung VP-M110 Miniket

Fabrice Bellard (612):
      Initial revision
      test
      removing old files
      Initial revision
      merge
      fixed unsigned pb
      fixed config for direct mplayer build compatibility
      added CONFIG_AC3, CONFIG_MPGLIB, CONFIG_DECODERS and CONFIG_ENCODERS (Arpi: don't forget to put CONFIG_DECODERS in mplayer)
      fixed cpuid macro to allow PIC compiling
      suppressed no longer needed emms()
      added emms_c() macro which should can used in c code in both mmx/non mmx cases
      added skip macroblock optimization (big perf win on black regions for example)
      suppressed unused stuff
      added --disable-mp3lib
      added benchmark option
      added mpegvideo_mmx.c - added HAVE_AV_CONFIG_H
      added HAVE_AV_CONFIG_H
      comments
      use block[] in structure to have it aligned on 8 bytes for mmx optimizations
      use block[] in structure to have it aligned on 8 bytes for mmx optimizations - dct_unquantize is always a function pointer - added specialized dct_unquantize_h263
      renamed - use of s->dct_unquantize function pointer - SHOULD add faster h263 mmx specific unquantization stuff
      added cpuid.s
      added external cpuid code to solve -fPIC and gcc unclear constraints problems
      added MJPEG decoder, renamed mjpegenc.c to mjpeg.c
      added MJPEG decoder + comments
      types fix
      added correct component id handling
      added 422P and 444P formats (need to patch ffmpeg.c so that it is handled in all the program)
      added 422P and 444P support - fixed block parsing error
      use default huffman tables if none given
      added idct mmx code
      updated mmx macros
      added idct mmx
      updated dct-test to test IDCTs too
      added block permutation functions
      added idct reference code
      permuted coefs in normal IDCT to avoid having different cases there
      fixed mjpeg matrix quant encoding and decoding
      changed quant matrix order (should fix mmx mpeg decoding bug)
      updated scans
      fixed block permutation in encoder (not optimal - should move it in forward DCT code)
      fixed scan protos
      added mmx idct
      updated
      fixed avi RIFF header size
      added forgotten emms() - fix various segmentation faults when using mjpeg
      fixed mpeg1 last block bug (mb stuffing code was not included in vlc table...)
      fixed mpeg1 first block bug (pb with black picture optimisation for B frames)
      return dummy quality to avoid bug in -sameq case - forgot emms in error case
      fixed mb_intra_vlc decoding table size
      updated
      use MAKE variable
      add deinterlacing support - add new image converter support
      added PPM read/write support - added pgmyuvpipe format - added PPM pipe read/write support
      new img_convert API - added deinterlacing filter
      export for imgconvert usage
      fix msmpeg4 seek
      added 422P, 444P support - added deinterlace support - added xxx to RGB24 convertion
      added interlaced MJPEG support
      fixed mpeg2 qscale decoding
      picture utils
      removed incomplete seek patch
      fixed mpeg2 interlaced DCT
      fixed last zero mv for field - fixed mismatch handling for intra coefs
      fixed mpeg2 non intra dequant - fixed MPEG1 and 2 matrix parsing
      moved matrix init away from MPV_common_init()
      win32 fixes
      arm specific code
      added arm support - added --disable-grab
      win32 fixes
      arm optimizations
      better vol header parsing for mpeg4
      win32 fixes
      updated
      added win32 vc6 project files
      added FAQ - you are free to add entries - the web page faq is generated from this file
      added jpeg image read/write
      jpeg hack
      updated
      changed jpegenc.c to jpeg.c
      added raw mjpeg format - added jpeg image format
      added static
      added raw mjpeg
      fixed wav riff tag size
      renamed log2 to av_log2
      log2 to av_log2 - added integer version define
      added draw_horiz_band test
      fix got_picture output
      copy avctx
      log2 to av_log2
      changed opendivx to mpeg4
      distclean fix
      suppressed nasm dependancy - rewrote forward DCT and motion estimation code
      removed
      updated
      added win32 cross compile support
      suppressed nasm stuff
      motion test
      win32 fixes
      suppressed mpglib
      added completely new mpeg audio decoder (integer only, free format support, lsf/mpeg25 support, high/low precision support, simpler code)
      added mpeg audio decoder tables
      merged code and tables between encoder and decoder
      added get_bits_count()
      fixed raw read for eof
      updated
      non 420P grab fix - suppressed unused 'sys/poll.h' header - do not put error if audio or video stream not found (no longer needs to use -an or -vn when it is obvious) - display error message when image filename does not contain '%d'
      added image question
      added patch info (I prefer only diff -u) and clear info about my patch copyright policy
      image number handling
      restore audio settings - use all the frame buffers provided by the driver
      fixed image number syntax problems
      version change
      fixed symetric quantization (better quality!)
      mlib merge
      corrected mpeg audio encoding overflows - now it should give correct quality even for very high volumes
      copyright change
      oops - forgot grabbing case for audio/video input auto detect
      fixed stereo grab
      suppressed mpglib option
      complete handling of pcm formats - hex dump option
      wmv format is asf format
      added pcm codecs
      suppressed mpglib
      added pcm formats
      added new ffplay program
      unified grab & convertion loops (should be easier to achieve correct synchronization now - but still work to do!) - added ffplay support
      update
      changed av_open_input_file() prototype
      correct pcm audio format handling
      changed audio and video grab interface (simpler now)
      fixed pcm bitrate
      fixed prototype change
      updated for my short term goals
      tried to avoid gcc 2.95.2 bug by puting explicit register constraints - added comment about rounding bug in some functions (need to correct or suppress them for regression tests)
      added proper memory handling functions - fixed include paths
      improved mpeg1/2 decoding speed
      put all integer init code to compute n^(4/3) - memory alloc and header fixes
      removed debug function
      proper memory handling functions
      added VPATH support for multiple dir compilation
      removed useless header includes - use av memory functions
      removed not maintained vc stuff
      removed unmaintained vc stuff
      added dummy crc output format for easy codec testing
      added VPATH support
      fixed pcm read
      use av memory handling functions
      added VPATH support - added regression targets
      added bigendian support - added cross compilation support - added generation of objects in another directory for simultaneous compilations (needs testing) - simplified redundant code
      use av memory functions
      added first version of regression tests - simply type 'make test' to test every codec and some mux/demuxes - added simple synthetic audio and video bitstreams generators so that no big streams needs to be used
      update
      oops - removed obsolete license draft
      added libav regression tests
      added regression test mention
      update
      added avcodec_set_bit_exact() for testing
      better av_freep()
      mpeg1/2 identifier - fixed frame rate for some bad mpeg1 streams
      removed unused code
      added first version of MPEG/DVB transport stream demux
      split mux/demux related structures - added file probing support - improve media file reading API
      suppressed mpeg demux mess - use now dynamic stream creation api
      converted to new API
      began to simplify code - use modified API for stream reading
      use new api (NOT TESTED) - suppressed strlcpy
      added libavtest target
      removed unused stuff - added dsputil_set_bit_exact() support for easier testing
      fixed memory free
      fixed one overflow problem
      fixed incorrect global gain - slightly increased precision of n^(4/3) table
      updated
      add more info about new configure features
      license/copyright change - use consistent library names
      license/copyright change
      more config options
      license/copyright change - fixed memory leaks
      license/copyright change - fixed header mess
      update
      fixed memory free (no free necessary for private data) - added better support for automatic testing
      fixed double free - fixed jpg image sequence automatic selection
      fixed double free - fixed probe function
      license/copyright change
      moved in libavcodec
      removed
      better logic for bswap
      fixed incorrect buffer handling
      consistent include usage
      license/copyright change
      reactivated swf/mjpeg/jpg formats testing
      added license
      removed invalid buffer handling
      fixed mangle issue
      bsd fixes
      use sys/soundcard.h for OSS audio
      don't use inttypes.h - use inline function for bswap_64()
      fixed layer1/2 overflow if very loud sound - fixed broken free format decoding to pass all mpeg audio standard decoding tests (please avoid patching the parser without having all test streams available - contact me if necessary)
      mpegaudio decoder is in high precision mode by default (low precision is only interesting for slow/old CPUs)
      added missing externs
      added missing externs (we were lucky that commons are used)
      avoid name clash with libjpeg - added missing externs
      removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
      use a local dir for tests
      use ff_idct_put()
      use ff_idct_put/add()
      added inlined put/add functions
      added ff_idct_put/add
      fixed case where abs_level > MAX_LEVEL
      fixed multiple allocation bug
      optimized ticker_tick() if num is big - please test it (should reduce CPU usage - this code is temporary before new PTS handling scheme)
      do not export internal functions - use stdint.h
      added installlib target to install static libraries and includes
      put memory functions in a separate file so that the user can redefinite them without modifying the library
      put codec registering in another file so that the user can install the codecs he wants
      memory functions are exported
      move unaligned access macros to dsputil.h - added unaligned 32 bit store
      fixed one mpeg decoder memory leak
      moved code to mem.c and allcodecs.c
      reverted to inttypes.h since it gives problems on some unixes
      use #ifdef
      fixed VOB parsing bug
      added raw tcp support
      added rtp support (not activated yet)
      added rtsp support (not activated yet)
      added rtp protocol
      added local port option
      added rtp port range support
      better UDP support - added preliminary multicast support (untested)
      api change for packet based protocols (such as UDP and RTP)
      added url_get_max_packet_size() support - added URL_RDWR flag
      added dynamic buffer support - added url_fprintf() and url_fgetc()
      http protocol now uses tcp: protocol (simpler)
      suppressed packet size dependancy
      added compilation of RTP/RTSP stuff
      added url_split()
      added RTP/RTSP stuff
      more consistent naming
      separated init code from utils to reduce static library size
      moved simple c utils
      renamed gettime to av_gettime()
      exported av_probe_input_format()
      added installlib target
      renamed gettime() av_gettime() - same for register_all()
      added include for ffserver dynamically loaded modules
      added proposed API for parse_only mode to extract compressed frames from compressed streams
      build stripped and unstripped ffmpeg - ffserver exports all its symbols for modules - added installlib target for development static libraries installation
      added RTSP and RTP server support - added daemon mode as default (use -d for foreground mode) - added module support
      added 'NoDaemon' command - added RTP/RTSP stream example
      suppressed unrelated logs
      first version of IDCT248 for DV decoding support
      added IDCT248 testing
      in fact IDCT248 needs to be normalized as I suspected
      first working version of DV video decoder
      added DV support
      simplified parsing - added MMX idct support
      fixed MPEG2 pts parsing - ignore unknown stream type
      minimum support for YUV411P (new combined scaler/converter will handle that better...)
      fixed NTSC end of line bug
      added raw DV demux (audio is currently not handled)
      support buffers containing more than one frame
      added DV support
      resample fix by Petros Tsantoulis
      added two pass support (same syntax as mencoder)
      added mpeg in avi support
      added two pass info
      use C99 syntax for field init
      added stream copy capability
      corrected pixel format display
      activated 'raw stream copy' feature (use -acodec copy or -vcodec copy)
      removed unused variables
      better MPEG4 support (at least can view some mpeg4 examples)
      cross compiling support
      raw picture support in mux
      added null output format for speed testing
      removed invalid header include
      removed warnings
      removed obsolete file
      suppressed vcd flag - use new mpeg mux format instead
      better framework for specific mpeg mux options such as vcd - added (incomplete) mpeg2 vob muxer for dvd creation
      AV synchronisation API
      use consistant PTS handling - fixed MPEG2 Pack header generation (clock_ref value is still slightly inaccurate)
      use PTS api - suppressed redundant fourcc tags (don't know why it was duplicated)
      the new output PTS handling is now generic
      use new PTS api - corrected AV sync for transcoding - factorized print_report() code
      changed write_packet API - should work as before
      single frame output fix
      put/get portable for IEEE double - renamed put_native_string/get_native_string
      new texinfo documentation - HTML version also included
      added Michael's page
      updated - feel free to correct/improve
      new generic FFT/MDCT code for audio codecs
      preparing integration of new AC3 decoder
      added wma decoder
      added new fft/mdct code - added fft-test program
      unified wav header parsing - added extradata support
      fixed block size selection
      updated
      update
      avoid name clash
      avoid name clash - fixed again block size selection
      added lrintf for non ISOC libcs (fixme: find a better test)
      use lrintf to avoid double conversion
      fixed nb_block_sizes detection - fixed codec_id test (avctx->codec_id does not need to be initialized)
      updated FAQ - converted it to texinfo
      added faq.html
      fixed multicast connect
      do not read probe data if format is specified - match_ext() must be exported
      added multicast SDP/RTP demux for multicast streams - added support for MPEG4 video decoding in SDP/RTP
      now raw mpeg4 video decoding is fully supported
      factorized All In Wonder code (not tested) - Added AIW auto detect (not tested) - mute audio when leaving (maybe an option would be needed ?) - removed all global state
      a/v sync support: added correct pts handling
      better than nothing patch for correct audio output with ffplay
      fixed output pts computation in case of pcm audio (fixes ffplay status display)
      '-' can be used for standard input
      doc update
      ffserver documentation update
      added CREDITS file (tell me if I forgot someone)
      update
      update
      update
      pkt->pts is already set, please fix audio if you modify this code - supressed unused PKT_FLAG_DROPPED_FRAME
      suppressed PKT_FLAG_DROPPED_FRAME
      supressed incorrect pts init (when the pts is not set, ffmpeg.c presumes the nominal frame rate is used
      added handling of URL redirectors (needed for RTSP streaming client)
      match SDP based on content instead of extension
      export match_ext()
      clean up of redirector code - first stage of RTP multicast support - added SDP redirector output for multicast
      update
      RTP multicast begins to work in MPEG1 - simplified stream bandwidth computation (no need to recompute it at each request)
      added example to hack with SDP/multicast streaming
      minimum handling for mpeg4 video multicast from avi files
      use codec_id so that the codec does not need to be opened
      experimental mpeg4 video support (currently no slice/timestamp handling)
      renamed libav to libavformat
      fixed multidir compile
      fixed 'file:' in URLs
      added ffserver disabling for easier porting
      updated with my current projets - add your own TODO if you wish
      added release tar target
      updated html docs
      update
      oops : better error resilience - should fix most wma decoding problems
      allow easy use of GCC vector builtins on x86 (has someone a better idea?)
      added define for builtins use - inverse fix by Romain Dolbeau
      compile fixes
      use av_malloc instead of malloc
      enabled SSE fft on x86
      align fixes for sse/altivec
      enabled SSE fft (first code using GCC 3.2 vector builtins on x86 - please tell me if you have better ideas for intrinsic naming or Makefile support)
      removed redundant includes
      fft altivec by Romain Dolbeau - simplified patch, test it on PPC with fft-test and wma decoding
      generic M_PI define
      suppressed redundant math.h - suppressed M_PI redefinition
      suppressed pix_norm_altivec
      tests link all libavcodec
      suppressed getopt.h
      altivec patches by Romain Dolbeau
      new image formats - suppressed ill defined formats (endianness, unless specified, is always the one of the CPU) - added avcodec_get_pix_fmt_name()
      moved avcodec_get_chroma_sub_sample() to imgconvert.c
      simplified code (need automatic testing) - added primitive new format support.
      changed BGRA32 to RGBA32. XXX: clarify expected behaviour on big endian cpu
      build update - avpicture_fill() returns the image size
      moved avpicture_xxx to imgconvert
      added missing formats in all functions - added monoblack, monowhite and gray8 support for most conversions
      added simple still image format support to simplify image and imagepipe video formats
      extracted yuv4mpeg from img.c (untested)
      added url_get_filename()
      added still image support
      still image support - increase audio buffer size - do hex dump even if packet is discarded
      added av_set_parameters()
      more formats
      jpeg detection fix - fixed imagepipe output
      fixed imlib2.c pixel format (imlib2 seems to use CPU endianness in RGBA32 as libavcodec)
      added gray<->RGB functions - fixed rgb<->yuv functions for non RGB24 case
      fixed dependancy generation (I want the case where SRC_PATH is not . to work too - very useful if you have different gcc versions to test at the same time...) - fixed -msse stuff
      fixed -msse stuff
      dependency handling is consistant with libavcodec
      removed unused rule
      dependency handling is consistant with libavcodec
      removed warnings
      video_device fixes
      grab device is in AVFormatParameter (at least better than global variable) - currently unused video output stuff for ffplay
      grab device is in AVFormatParameter (at least better than global variable)
      dv1394 channel selection by Max Krasnyansky - modified channel number to be in base 10 by default
      use av_malloc()
      undefine malloc macros
      use av_realloc()
      use av_malloc() functions - added av_strdup and av_realloc()
      removed warning
      use av_strdup()
      removed unneeded mallocs
      DV audio decoder by Roman Shaposhnick
      Audio support for DV1394 by Max Krasnyansky
      win32: added vsnprintf
      win32: rint() does not seem to be defined with mingw32-gcc 2.95 - do you have a better solution ?
      make consistent decisions for dlopen/dlfcn and vhook
      added DVD LPCM decoding support
      added stripped down libpng
      added basic png handling - do not remake dependencies automatically
      added basic png handling
      fixed dependencies
      dependencies are only build with make depend
      Try this fix
      added animated GIF decoder (pts and various disposal handling are missing)
      added jpeg image encoder and decoder (new YUV handling routines and mjpeg codec fixes are necessary to go further)
      fixed copy
      handle odd image sizes when using subsampled chroma (useful for JPEG images)
      avoid too many false detections
      added automatic GIF/animated GIF probing - added GIF as an image format too - added interlaced gif support
      added primitive image GIF encoder based on animated gif encoder
      added GIF image format (both read and write)
      zero copy packet handling for DV1394 by Max Krasnyansky
      removing unused stuff
      yes, PNG can be simpler :-)
      DV packet copy fix by Roman Shaposhnick
      added paletted 8 bit format support
      added 8 bit palette support
      added 8 bit palette support for non animated GIF
      raw ac3 auto detects parameters
      avoid false URL protocol detection when using ':' in filenames
      added DTG active format information
      removed useless copyright policy - added xawtv link
      up to date license terms
      update from texi
      added new netpbm pam format support (needed for alpha plane support)
      rgba32 convert
      rgba32 support
      added progressive image support
      added progressive PNG support (both reading and writing)
      new YUV formats for correct JPEG YUV support - added automatic pixel format conversion guessing (minimize loss and size)
      fixed static init
      YUV formats/gray formats are correctly defined - added format loss information - preliminary JPEG YUV formats support
      accurate YUV to RGB and RGB to YUV conversions - added comments
      almost exhaustive image conversion support
      added all missing UV conversions
      use correct YUV format for JPEG
      loss fixes (thanks to Daniel Serpell) - shrink22 fix
      fixed PIX_FMT_YUV422 conversions
      replaced progressive by interleaved
      transparency support
      added img_get_alpha_info()
      rgb555 alpha extraction fix
      update
      autogeneration of html files
      fixed SRC_PATH usage
      use explicit image name
      SDL probe - ffplay build option
      added ffplay utility
      use custom packet allocation only for DV
      better pause when displaying audio samples
      disable ffplay if no SDL
      works with old gcc versions too
      dynamic array functions
      mpeg ts demux uses DVB SI info and generates correct PTS info - added prototype mpeg ts mux
      win32 compile fix for ffplay by gbazin at altern dot org
      fixed segfault if sequence header has not been found before slice decoding
      cosmetic change in resync code - added PAT scanning code if no SDT is found (in the futur it would be interesting to give an API to change channel - also useful for DV input number or TV grab tunning
      fixed RTP/TCP client support
      fixed NTP generation for mpeg
      fix fmt == NULL case
      fixed test.conf path
      put b- prefix to libav tests to avoid incorrect ffserver test
      added fulltest target
      warnings are errors
      added primitive aborting system
      removed warnings
      suppressed RTSP abort hack - added PLAY/PAUSE
      support aborting in TCP
      added correct RTSP aborting - added option to force TCP usage in RTSP playback - added experimental RTSP pause support
      added RTP/TCP protocol support
      mention the two asf formats
      incorrect comment
      long awaited simple libavformat output API example
      mention libavformat API examples
      removed unnecessary codec
      suppressed memory leaks - added image conversion support if the codec does not support YUV420P - added raw output (any comments are welcome to simplify the API)
      fine A-V synchronisation support with either audio or video as master clock source - MPEG2 A-V synchronisation fixes - fixed format selection - fixed error reporting - audio and video component selection
      added stream start time and duration API - AV_NOPTS_VALUE is no longer zero
      add av_new_stream() usage
      update duration and start_time - add av_new_stream() usage
      update duration and start_time
      exports mpegps_demux
      update duration and start_time
      removed MAXINT64
      initial duration/start_time generic support - displays stream duration and average total bitrate when using an input file
      added get_audio_frame()
      no default bit rate if decoding
      support for older real audio files (<= version 3)
      repeat_pict was never returned
      update
      added ffplay documentation - added automatic man page generation for ffmpeg, ffserver and ffplay
      updated supported codec/format
      when changing version, please keep my changes (win32 compile fix)
      win32 compile fix
      removed deprecated win32 option (use --enable-mingw32 and --cross-prefix)
      explain Windows Native and Cross build - linked to DLL build page (unsupported yet in the makefiles)
      added all options
      more sections in help - copyright fixes
      export int64_t_C as it is needed by libavformat.h
      man page installation support
      spell fixes
      exclude Duration field too
      disabled file based tests
      use new loop option
      added loop option
      suppressed frame number modulus hack - added loop_input hack which I find easier to understand
      accepts pgmyuv extension
      ac3 test
      up to date regression tests
      update
      update version
      memmove fixes (Jon Burgess)
      removed unused variables
      removed unused variable
      avoid warning
      avoid warning
      removed unused variable
      spelling fix
      fixed audio frame buffering problem (should correct problems on some streams) - faster synthesis filter - prototype 'parse_only' support
      version update
      ctype.h is a common header
      fixed endianness dependancies (untested)
      removed ctype.h header
      removed os_support.h as most of libavcodec is os independant - most of the time only libavformat needs OS dependant features
      removed os_support.h
      simpler strptime - added os_support.[ch] - moved localtime_r to os_support.c
      AC3 note
      removed os_support.h
      changed my mind about CODEC_ID_MPEGVIDEO - CODEC_ID_MP3LAME is absolete
      CODEC_ID_MP3LAME is obsolete
      header fixes - removed MPEG-4 b frame limtation in help
      removed ctype.h
      removed strptime test
      removed localtime_r.h
      ID3 parsing and generation in MP3 format
      update sub_id in mpegaudio decoding (might need same method as MPEG2VIDEO too ?)
      show stream info with -stats
      mpegaudio decoder no longer forgets to decoder last audio frame
      remove warnings
      removed warnings
      use const data - began to make code more portable
      removed warnings
      64 bit pts for writing - more const usage
      removed invalid sequence end code
      example using video encoder latency
      simplified version handling - set win32 install path
      windows installer
      simplified version handling
      doc update
      update
      fixed windows build
      correct DLL windows install
      generate correct PTS in transport stream - use mpeg2 encoder by default
      better image sequence support (Charles Yates) - suppressed unused non overlay support
      avoid initializing pts for not decoded streams
      doc update
      use VC++ lib tool if available
      remove unused AMREXTRALIBS
      update
      clean target removes output_example
      updated
      added fake codec CODEC_ID_MPEG2TS of type CODEC_TYPE_DATA (needed for simpler handling of raw transport streams in ffserver and RTP - better solutions are welcomed)
      mpeg TS demux API (called from RTP layer) - raw mpeg TS reader (uses fake codec MPEG2TS)
      make AVFMT_NOHEADER flag dynamic - added av_open_input_stream()
      added MPEG2TS support in RTP, SDP and RTSP - replaced fake RTP demux by a specific API
      comment fix
      exported mpa_decode_header for new parser API
      new audio/video parser API
      also parse extradata for MPEG4
      initial av_read_frame() and av_seek_frame() support
      initial seek support - more generic play/pause support
      use parsers
      seek support for PCM formats
      added parsing
      fixed double free
      seeking support - fixed some ADPCM decoding cases
      seek support
      initial seek support
      frame rate should be completely disabled in asf (closer now) - disabled seek
      use av_read_frame() - added initial seek support (rename -start to -ss (mplayer...)) - added '-dump' option to dump input packets
      initial seek support
      simpler bandwidth allocation for RTSP streaming - use av_read_frame() - initial support for raw MPEG2 transport stream streaming using RTSP
      fixed memory leak
      fixed low delay decoding
      copy packet without omitting the padding
      fixed potential problem if aspect_ratio_info == 0 for MPEG stream - fixed aspect ratio problem if CODEC_ID_MPEG2VIDEO is used to decode an MPEG1 stream (which is the recommended codec id for mpeg video)
      -interlace option - better consistency in help
      fixed interlaced encoding
      interlaced DCT support for MPEG2 encoding
      generate MPEG elementary streams in all cases - added interlaced MPEG2 test
      update -interlace help
      update
      fill codec info
      fixed top_field_first support when encoding
      init picture structure to avoid uninitialized fields
      added End Of File handling to return last picture for MPEG1/2/4
      moved packet output to a separate function - added the frame buffered by the decoder at EOF, if any
      make av_read_frame() return buffered frames at EOF
      MPEG1/2 decoding update (buffered frames are decoded now)
      pts and dts support in parser API
      fixed incorrect PTS/DTS logic in MPEG video case (caused rare PTS glitches if start codes were between two PES packets)
      better and simpler logic for MPEG muxing - fixed rare MPEG muxing PTS generation bug (stuffing is added in such rare cases) - fixed AC3 payload size generation - generate correct AC3 frame header (need spec checking)
      primitive LPCM generator
      harcoded DTS computation for mpeg
      subtitle codec type support
      DVD subtitle parsing - show mpeg component IDs by default
      optional ISO 639 3 letter language field
      buggy PCR output (just to keep HW decoders happy) - audio and subtitle language support - 'title' stream field sets the service name - fixed TS packet stuffing code - support for subtitle PES packets
      subtitle option
      support for adding new components (syntax: ffmpeg -i input_file [output options] output_file [[component output options] {-newaudio|-newvideo|-newsubtitle}]... - added subtitle support - added '-alang' option
      Use the ffmpeg mdct function, patch by Fabrice Bellard. Thread: [FFmpeg-devel] NellyMoser transform bug, 10/25/2007 12:24 PM

Falk Hüffner (49):
      Minor warning cleanup.
      Revert previous bogus warning fix.
      Reintroduce lost idctSparseCol for Alpha. Sorry for adding even more
      * Fixed constraints
      Ugly hack to make the assembler accept MVI instructions.
      Implement put_pixels_clamped and add_pixels_clamped in Assembler. This
      Add profiling support.
      Assembly version of put_pixels. This is currently the function that
      Activate ALT_BITSTREAM_READER by default on Alpha, since it seems to
      Use updated motion compensation routines.
      Update and activate dct_unquantize_h263_mvi. Thanks to Måns Rullgård
      Make dct_unquantize_h263 work on systems without MVI extension.
      * Detect "retrict" keyword
      * Move "restrict" fallback to common.h
      Implement clear_blocks_axp.
      Remove some cruft.
      Tweak Alpha compiler flags.
      Warning fix.
      * Improve xy2 routines slightly
      Remove support for variable BSIZE and INCR, as sub_pixels_* is no
      Force inlining on get_vlc2.
      MVI optimizations for motion estimation.
      Adapt to new 8/16 table scheme.
      Due to version renumbering, Alpha builtins aren't supported in thegcc
      Fix for h % 8 == 4 in put_pixels_axp_asm.
      Synthesize pixels16 functions from pixels functions.
      * Remove -finline-limit=8000 (not supported by gcc 2.95)
      Minor warning fix.
      An easy way to speed up encoding by 6%.
      Add Alpha assembly for pix_abs16x16. Optimized for pca56, no large win
      Fix for source dir != build dir. Patch provided by Måns Rullgård.
      Some nifty stuff for upcoming patches.
      Make compilable again after DCT cleanup. It gives wrong results,
      * Fix dct_unquantize_h263_axp for changed block_last_index indexing.
      Move Alpha optimized IDCT to own file. Based on a patch by Måns
      Kludge around compilation failure on Alpha.
      Implement sad8x8 and sad16x16 with pix_abs.
      sse16_c is totally fucked up (unaligned loads, LONG_MAX is undefined,
      Warning fixes.
      Warning and compatibility fixes.
      Use asms instead of builtins when compiling for generic Alpha. Less
      Fix compilation with ccc
      tiny warning fix
      Update coding rules.
      warning patrol
      Don't include config.h from assembly, because the mplayer developers
      -freduce-all-givs has been removed from gcc
      Communicate proper aliasing to gcc (needed for 4.1).
      Change license header to LGPL for consistency.

Felix Bünemann (10):
      fix rgb/bgr for 15/16bpp
      Move ignores in subdirs to their respective .cvsignore files.
      mangle for win32 in postproc
      fix mangling with runtime cpu detection
      add mangling
      add mangling
      gcc 3.1 warning fix (patch by Felix Buenemann <atmosfear at users.sourceforge.net>)
      in libavcodec/Makefile there is in line 110: "distclean: clean<binary char>"
      warning fix patch by atmos
      cygwin patch by (Felix Buenemann <atmosfear at users dot sourceforge dot net>)

Felix von Leitner (2):
      add/diff_bytes bugfix patch by (Felix von Leitner <felix-ffmpeg at fefe dot de>)
      sse16 & pix_norm1 optimization patch by (Felix von Leitner <felix-ffmpeg at fefe dot de>) (with some modifications)

Flavio Pimentel Duarte (2):
      Error path fix: close the video device on failure
      Fix v4l grabbing with some webcams, and simplify the code.

Florian Echtler (1):
      Parse 'dats' as data.

Fortin Denis (1):
      modify msmpeg4 functions to generate VC1 compatible block layer bitstream

Francesco Cosoleto (4):
      Substitute the not anymore supported "me" option name with "me_method"
      Document the directpred auto value.
      Add a separate 'testclean' rule to clean regression test data without
      Add the ability to set the timestamp to the current time by specifying 'now'

Francisco Javier Cabello Torres (2):
      via c3 detection patch by (Francisco Javier Cabello Torres <fjcabello at visual-tools dot com>)
      via c3 fix patch by (Francisco Javier Cabello Torres <fjcabello at visual-tools dot com>)

Francois Oligny-Lemieux (2):
      harden h264 decoding to prevent some crashes when input data is corrupted.
      always decode extradata when of non-avc stream (like RTSP)

Frantisek Dvorak (2):
      Build process: libpostproc patch by (Franti?ek Dvo?ák <valtri at users dot sourceforge dot net>)
      use sdl-config with $cross_prefix

François Revol (88):
      * BeOS patch by François Revol <revol at free.fr>
      beos/mov/adpcm patch by François Revol <revol at free dot fr>
      YUV410P to YUV420P patch by François Revol <revol at free dot fr>
      nanosleep patch by François Revol <revol at free dot fr>
      adpcm encoding patch by François Revol <revol at free dot fr>
      patch by François Revol <revol at free dot fr>
      added perm inheritance from ffmpeg_g (it looks like 'strip' in BeOS doesn't keep them, though the Linux one does !?)
      added BeOS net_server support (R5 network stack), basically the same
      added MACE (Macintosh Audio Compression/Expansion) 3:1 & 6:1 support
      added BeOS net_server support (R5 network stack), basically the same
      gcc optimization on BeOS (patch by François Revol <revol at free dot fr>)
      BeOS Audio ouput patch by (François Revol <revol at free dot fr>)
      MIN/MAX sys/param.h patch by (François Revol <revol at free dot fr>)
      ringbuffer patch by (François Revol <revol at free dot fr>)
      Bump up the priority on the decoding thread.
      Avoid "unused variable 'c'" warning.
      Hey, not everyone has usleep() ! (temporary fix, we should check for this in configure)
      That should be enough for now :)
      redid usleep() fix for BeOS, more consistant with the MINGW32 one. removed a TAB :P
      Fix for odd strf tag in Stargate SG-1 - 3x18 - Shades of Grey.avi
      Fix probing for some files (ew.mov, april.mov)
      removed redundant 4CC
      Fix against C integrists breaks :P (hope I didn't break anything else)
      fix for using ffplay without specifying the audio device
      fix for legacy audio
      Looks like this one was forgotten in the INT -> int_t move
      temporary fix: BeOS only has 'which' as an alias to a shell func :^)
      prettier debug output
      Cleaner support for usleep()
      Cleaner support for usleep() (with the correct header :)
      more fix for make -j x
      /me should really learn doxygen :)
      fix build issue with --enable-shared & --enable-vorbis
      debug stuff only for DEBUG :)
      cleanup; unused pipe() stuff removed.
      experimental BeOS audio input support. (needs unreleased library)
      the media node now won't connect itself to the default audio input with -ad wait:
      Fix for buggy (?) g++ (bailling out for beosaudio.cpp)
      avoid detecting mpeg stream when other formats embed mp3 tracks.
      Add latency check, the Media Kit shoul repport it, but this seems broken.
      cleanup; BSoundRecorder moved to Experimental namespace
      BeOS fix: NOT every ld likes undefined syms, include C++ objs, and link to libs needed for audio in.
      OSX and BeOS networking fix (socklen_t)
      64 bit pts for writing - more const usage (don't forget me !)
      First implementation of nsv demuxer.
      BeOS threading support. changed some "if FOO_THREAD||BAR_THREAD" to a more generic "if HAVE_THREADS"
      cvsignore config.log
      get debug code to compile again.
      Revert the fixed-size-sample patch as it brokes  and others
      AAC fix by Thomas Raivio <tjraivio AT cc DOT hut DOT fi>
      probe for 'junk' and 'pict', changed the scores.
      fix the patch to allow Sony PSP MP4 creation
      don't crash on null codec; by matthieu castet <castet DOT matthieu AT free DOT fr>
      fix segfault by matthieu castet <castet DOT matthieu AT free DOT fr>
      added a lock/unlock_lib pair to allow extern progs to serialize access to lavc.
      Stereo MACE fix by Burkhard Plaum <plaum AT ipf DOT uni-stuttgart DOT de>
      use av_free() instead of free() where it's meant to.
      get fish to compile (time() forbidden), pstrcat not declared from ppm.c
      Add 3IV2 4CC by "Rob Coenen" <rob AT massive-interactive DOT nl>
      pktdumper debugging utility for demuxers. This tool dumps each AVPacket demuxed to a stamped file.
      s/\t/    /g for indentation fairies
      Patch for "uuid" atom followed after atom "ftyp" by "QuickTime <ffmpeg AT gmail DOT com>"
      "don't quit" option to attach a debugger /check mem leaks
      change my ç from latin-1 to UTF-8. UTF-8 ownz!
      fix for systems not defining PRIu64, not everyone knows C99.
      Add some more PRIfooNN crap so it builds again here. Why can't ppl add defaults when they use such things ?
      fix parsing version 1 mdhd atom (64 bit) as suggested by b-o-n-d AT gmx DOT net (unstested); added code to map language codes to stream->language[]
      finally found what those >138 codes were... crappy compressed 5bit ascii. this gets them correctly, and adds setting track lang in movenc too.
      this fixes compilation with gcc 2.95 which seems to not like array casting... special case inline of check_block which makes the array and passes it
      :%s/check_block_2p/check_block_inter/g
      fix debug code: do not try to take st->codec if st is NULL. This makes ffplay give me pics from FreeBoxTV \o/
      The long awaited BeOS cleanup, phase 1
      The long awaited BeOS cleanup, phase 1.1: removed useless ifdef, changed check for net_server to IP_MULTICAST_TTL.
      an alloc func failing should give ENOMEM anyway.
      BeOS specifics: check for Haiku & Zeta (for the future); ignore ldconfig, it just runs here :) ; comment about network checks.
      Remove unused ENODATA define
      This fixes error handling for BeOS, removing the need for some ifdefs.
      This is no longer necessary for BeOS now that we don't play nasty tricks with error codes.
      poll() emulation for BeOS, and maybe MinGW.
      MinGW patch by <koyama AT infocity DOT co DOT jp>, reworked. Removed placeholder for winsock.
      Make sure we don't try to emulate poll() when it's there if --disable-ffserver... 10l
      ffserver uses dlopen(), so make sure it links to libdl if needed.
      BeOS only: we don't have SA_RESTART. #define to 0 and emit a warning. I don't plan on hosting a streaming tv anyway but it gets ffserver building.
      Some more BeOS cleanup: check for arpa/inet.h; declare the prototype for inet_aton if not found; remove barpainet.h as it's not longer needed.
      Suppress possible error from ulimit not having -c or being missing.
      Fix building with C++ objects, variable names were inconsistent.
      Fix non-debug build.
      Fix build.

Fred (1):
      1) Add MMX deinterlace code.

Fred Rothganger (2):
      rawvideo patch by (Fred Rothganger <rothgang at uiuc dot edu>)
      rational.h not installed patch by (Fred Rothganger <rothgang at uiuc dot edu>)

Fredrik Orderud (2):
      Put double-quotes around the install path in the "install-headers" section
      Define EMULATE_INTTYPES for all windows-systems, except Cygwin and MinGW

Gabriel Forté (1):
      fix memory leak (patch by Gabriel Forté  gforte _at_ wyplay _dot_ com)

Gabucino (2):
      postproc/ dir now respects --with-extraincdir
      many small typo and grammar fixes

Gael Chardon (4):
      Log file and Windows patch by ("Gael Chardon" <gael-announcements+ffmpeg at 4now dot net>)
      int64_t_C patch by (Gael Chardon <gael-announcements+ffmpeg at 4now dot net>)
      Seeking in .mov/.mp4 files
      Timestamp patch by Gael Chardon <gael DOT ffmpeg AT 4now DOT net>

Garrick Meeker (1):
      flash video (flv) support patch by (Garrick Meeker <gmeeker at theoryllc dot com>)

Gert Vervoort (3):
      more msmpeg4 fourccs patch by Gert Vervoort <Gert.Vervoort at wxs.nl>
      10l (obmc_scratchpad not always initialize)
      Fix setting default value of scan_offset option for target svcd.

Giacomo Comes (1):
      Fix DTS support compilation issue. Patch by Giacomo Comes < comes AH naic POIS edu >

Giancarlo Formicuccia (2):
      *** fix for read()
      - Gracefully handle the case where not all the streams are requested/wanted

Gianluigi Tiesi (4):
      Disable encoders patch by (Gianluigi Tiesi <mplayer netfarm it>)
      Place proper #ifdef around muxer-specific code.
      Fix compilation with --disable-muxers.
      Support shared pthreads-win32.

Gildas Bazin (13):
      movenc patch by (Gildas Bazin <gbazin at altern dot org>)
      memleak patch by (Gildas Bazin <gbazin at altern dot org>)
      avio patch by (Gildas Bazin <gbazin at altern dot org>)
      AAC support in mov, correct aspect ratio support and user data support patch by (Gildas Bazin <gbazin at altern dot org>)
      avsync patch by (Gildas Bazin <gbazin at altern dot org>)
      fft_*() renamed into ff_fft_*() patch by (Gildas Bazin <gbazin at altern dot org>)
      AAC in mov/mp4 patch by (Gildas Bazin <gbazin at altern dot org>)
      small typo patch by (Gildas Bazin <gbazin at altern dot org>)
      IWMMXT configure support + runtime selection patch by (Gildas Bazin, gbazin : altern org)
      Properly #ifdef muxer-specific parts.
      Add --disable-protocols option to configure to disable I/O protocol from
      WinCE cross-compilation support
      Disable protocols by default on WinCE, not MinGW.

Glenn A. Serre (2):
      variable initialization (because of a gcc warning)
      Use getCachedContext instead getContext to support resolution change. Patch by Glenn A. Serre

Glenn Maynard (3):
      localtime_r patch by (Glenn Maynard <g_sf at zewt dot org>)
      Debug/optimization patch by (Glenn Maynard <g_sf at zewt dot org>)
      Trivial error reporting patch by (Glenn Maynard <g_sf at zewt dot org>)

Glyn Kennington (1):
      SWF sanity check patch by (Glyn Kennington <glyn dot kennington at ox dot compsoc dot net>)

Godwin Stewart (1):
      update -t option documentation

Graham Booker (5):
      Do not add -fomit-frame-pointer to CFLAGS when optimizations are disabled.
      work around issues with the old version of Gnu Assembler shipped on
      Altivec version of h264_(h|v)_loop_filter_luma
      Use a faster way to compute 255-val: Instead of creating a vector of
      part 2/2 of fixing Altivec-accelerated H264 luma inloop filter

Gregory Montoir (6):
      fix SHOW_SBITS for ALT_BITSTREAM_READER_LE, patch by Gregory Montoir %cyx A users P sourceforge P net%
      Remove av_cold attribute on decode_frame and add it to decode_end
      Remove inaccurate computation of pts for audio packets and rely on
      Remove context fields which can be used as simple local variables.
      Motion Pixels Video Decoder.
      Motion Pixels MVI Demuxer.

Guillaume LecrefECERF (1):
      url_write is part of the public API so it may not be under ifdef.

Guillaume Poirier (89):
      GCC-4 fix for AMD-64
      10l. Previous GCC4 commit broke compilation with gcc-3.4 and maybe others ia-32
      GCC4 fix
      Fixes GCC4 fix by using "g" instead of "mp" as some compilers misscompile
      Reverts GCC-4.0 "fixe" which broke GCC-3.3 and maybe others
      Re-enables the GCC-4 fix for AMD-64 only. Patch by cartman and poirierg
      revert useless patch that was supposed to fix DTS support (but the fix belongs to MPlayer, not FFmpeg)
      Typo fix
      Make DV(B|D) subtitles (en|de)coders be conditonnaly compiled depending on the configure system
      add some h.264 fourccs to codec_bmp_tags[]
      Add the rest of missing Reg_* macros to support both AMD-64 style regs and IA32 regs.
      Colin offered to maintain Amiga PPC port, Romain has only OSX
      Add 3dnow intrinsic support for AMD64 machines: -march=athlon means IA32 for GCC, so use -m3dnow instead.
      Add an actual 3dnow intrinsinc so the configure script doesn't just check for the existence of the 3dnow header
      fix compilation of 3dnowext instinsincs, both on ia32 and AMD64.
      Woops, this file was supposed to be added in r5556: "Move initialisations and internal symbols in allformats.h"
      Add/improve PPC64 support. On such machines, "uname -m" returns "ppc64,
      Add #define REGc
      Mention MacIntel support
      Compile fix (for GCC<3.3) when encoders are disabled
      Add missing file from commit r6122 (AVISynth support)
      Use libswscale in fish vhook.
      MONster is no longuer shipped with CHUD, as it's been superseeded by Shark
      Update checksum to match latest ASF patch
      Move TRANSPOSE8 macro to dsputil_altivec.h.
      Add pentium-m to the list of tuning targets
      adds some useful comments after some of the #else, #elseif,
      rename --cpu (and related variables) to --arch as this option always
      rename --tune to --cpu and make --cpu pass the apropriate -march=xx option to the compiler.
      10l: Revert bogus use of GCC's -march=xx:
      Add x86 conditionnal mov (CMOV) "detection" (based on the cpu passed by --cpu=xx)
      fix harmless typo / consistency fix
      Protect code that uses CMOV instructions with HAVE_CMOV,
      Fix licence header (aka <pedantic mode='on' /> )
      move licence header fixes
      Last fix of licence header
      Add base64 encoding and decoding routines
      woops, fix missplaced doxy comment
      Fix wrong params name in PUT_UTF8 doxy comments
      Add doxy comments for macro GET_UTF8
      Add useful Altivec types, taken from with permission from x264 authors
      Fix declaration and code thingie
      Add IDCT8 routine in Altivec. Patch by yours truely with Linux fixes by Luca Barbato
      GCC 3.3 compile fix
      Doxygen comments about variables described by Michael here:
      Original X11 device demuxer patch from Clemens Fruhwirth
      Adapt the patch to more recent FFmpeg habits
      RFC 1 stage: Improve maintainability of the cursoir painting code
      RFC 2 stage: First feedback from review comments
      RFC 3&4 stage: FFmpeg style aint't my style
      RFC 5: Mouse cursor painting reduc and more cleanups
      RFC 6: Overall cleanups again.
      Feedback from Karl H. Beckers about mouse cursor
      revert broken commit of x11 device demuxer patch
      Original X11 device demuxer patch from Clemens Fruhwirth
      remove trailing whitespace left over
      Add SSSE3 (Core2 aka Conroe/Merom/Woodcrester new instructions) detection
      add missing bit from commit r7423
      links to ARM optimization manuals
      remove useless cast that makes code unreadable
      use more portable PRIu64
      add doxygen docs for avpicture_fill()
      fix 2 mistakes in doxy comments, spotted by Michael
      Add doxy comments, based on Loren's explanations posted here:
      use macro Use DECLARE_ALIGNED_16 to align stack-allocated variables
      minor GCC3 compile fix
      one step closer to restoring GCC-3.3 support: kills 12 errors, just 2 to go
      cosmetics
      List myself in CREDITS as I did have some of code committed,
      this file credits authors of non-trivial code, which I don't qualify for
      some samples aren't decoded correctly such as
      convert h264_deblock_q1 to an inline function.
      re-enable use of h264_v_loop_filter_luma_altivec and h264_h_loop_filter_luma_altivec,
      restore GCC3 support
      add a comment to indicate which #endif belong to which #define
      use shorter types vec_"type" instead of the too long vector "type"
      remove alignment correction of the destination pointers in luma_16x6
      Fix links to ARM documentation
      mention slice-based parallel H.264 decoding in changelog
      fix indentation that was messed up by r11628
      10l: fix always false test: Binary & has lower precedence than ==
      fix broken indentation
      arg 10000l. Fix wrong fix committed in r12141
      armv4l/float_arm_vfp.o must be added to OBJS-$(HAVE_ARMVFP) list since
      convert DOS-style carriage return to Unix-style
      compilation fix with GCC 4.0.1 on MacOSX tiger, broken by the removal of AVV() macro
      add Altivec implementation of clear_block
      add AltiVec implementation of int32_to_float_fmul_scalar
      Disable usage of ff_h264_idct_add_altivec since AltiVec versions of h264_idct_add16,

Gustavo Sverzut Barbieri (4):
      Introduce a mechanism to add object-specific linker flags.
      fprintf/perror --> av_log
      Use fixed point in color conversion.
      warning fix from a patch by Gustavo Sverzut Barbieri, barbieri gmail com

Gwenole Beauchesne (19):
      cosmetics: Add empty line for consistent formatting.
      Use a constant instead of a magic number.
      Make ff_h263_resync() return the bit position of the resync_marker or
      Record B-fraction index from frame header and two fields from entry point structure
      Add fields to H264Context and SPS for upcoming VA API support.
      Move initialization of {luma,chroma}_weight_flag[] up,
      More approved hunks for VAAPI & our new and cleaner hwaccel API.
      Add ff_h263_find_resync_marker() to find the bit position of the next resync_marker, if any.
      Rename buf_vdpau to buf_start
      Approved hunk from the AVHWaccel patch by Gwenole Beauchesne.
      Add missing prototype for ff_is_hwaccel_pix_fmt().
      Add ff_hwaccel_pixfmt_list_420[] definition.
      Add VA API pixel formats.
      Return early if a HW accelerator is used.
      Remove superfluous call to mpeg_decode_slice() for hwaccel, this
      Add HW acceleration hooks for MPEG-4 / H.263 decoding.
      Add hardware acceleration hooks to H.264.
      Use ff_hwaccel_pixfmt_list_420[] in H.263 decoder.
      Make ff_print_debug_info() a no-op when using HW acceleration.

Gábor Kovács (1):
      Just noticed there is a memory leak in h264.c with the usage of rbsp_buffer.

Hans Zandbelt (2):
      IPv6 support patch by ("Hans Zandbelt" <Hans.Zandbelt <at> telin {dot} nl>)
      udp ipv6  localhost resolving patch by ("Hans Zandbelt": Hans Zandbelt, telin nl)

Harald Karnet (1):
      Correct the parameter order for the call to "av_clip".

Harikrishnan Varma (1):
      make the delta magnitude a user specifyable parameter:

Hauke Duden (3):
      improved VCD support patch by ("Hauke Duden" <H.NS.Duden at gmx dot net>)
      fixed buffering for low bitrates patch by (Hauke Duden <H.NS.Duden at gmx dot net>)
      mpeg SVCD compatibility, SCR fixes, standard compliance

Heikki Lindholm (1):
      make mpeg decoder handle mid-stream changes in resolution

Heliodoro Tammaro (1):
      patch by Heliodoro Tammaro <helio at interactives dot org>

Henning Haaland Kulander (1):
      VideoBufferSize option patch by ("Henning Haaland Kulander" <hennikul at ifi dot uio dot no>)

Henrik Gulbrandsen (2):
      Correct ASF duration.
      set pkt duration, only when dts == pts though,

Henry Mason (2):
      new vbv calculation patch by (Henry Mason <talus25 at speakeasy dot net>) with slight modification by me
      yuv4mpeg output support patch by (Henry Mason <talus25 at speakeasy dot net>)

Herve W (1):
      * Allocating 16 bytes more for the MEMALIGN_HACK is enough. There's no

Hetz Ben Hamo (6):
      A newer Changelog
      small fix
      No more HTML in Changelog
      Typo fixes (ok, mplayer team, enough laughing ;)
      ..
      Changelog update

I. Po (1):
      Fix memleak, patch by  I. Po % yyymmmm # gmail O com %

Ian Braithwaite (8):
      Don't output the first two frames, since they don't contain valid audio.
      Simplify gain block handling.
      Replace custom modified discrete cosine transform with ffmpeg's own.
      Combine the window and overlap loops,
      subband_coef_noise -> subband_coef_sign
      Small simplifications to subband coefficient handling and use av_random().
      Optimize by building the mdct window and multipying/adding at the same time.
      WMA decoder improvement, output closer to the dmo binary.

Ian Caulfield (11):
      subs.diff fixes a couple of minor bugs in my DVB subtitle decoder, and also fixes a few
      subtitles support for ffplay patch by (Ian Caulfield, imc25 : cam ac uk)
      Allows the AC3 parser to read the frame size and codec parameters from E-AC3 streams,
      fix indentation, patch by Ian Caulfield % ian P caulfield A gmail P com %
      Fix a bug in the DVD subtitle decoder where subtitles with odd heights would not
      create colorspace.h and use it where appropriate
      Rename "palette" to "colormap" when it's referring to the mapping from a 2-bit
      Add HD-DVD subpicture decoding.
      Reindent two lines.
      Update the dvdsub parser to be able to parse HD-DVD subtitle packets.
      MLP/TrueHD parser

Igor Mozolevsky (1):
      pentium4m is a valid CPU name, add it to the list.

Isaac Richards (1):
      - Looks a tiny bit harder in mpegps_probe() for a valid start code.  This is

Ivan Kalvachev (83):
      DMV support patch by ("Ivan Kalvachev" <ivan at cacad dot com>)
      Mpeg2 16x8 Patch by ("Ivan Kalvachev" <ivan at cacad dot com>)
      initial XvMC support
      initial XvMC support, define common packet for decoder <-> libXvMC communication
      initial XvMC support
      compile xvmc only when requested
      make MPV_frame_start fail on get_buffer failer
      Add GNU header, remove X includes, add more check for xvmc render packet
      XvMC speedup by removing one memcpy and doing MB packing
      mpegvideo is back
      silence few warnings
      use get_format
      initial chroma_format changes,xvmc tweaks and codec_cap
      fix typo
      forgot default chroma_format value
      add vlc for cbp=0 that is valid in 422,444
      silence warnings
      move MPV_common_init after parsing stream parameters, matrix rebuild
      mpeg12 fix 10l of aspect calucalations, return gop perser to print debug info, and make mpeg1_seq_header hot overwrite height/width
      use pan_scan to remove some weight, proper pan_scan offset reading
      cola test
      mpeg2 chroma422/444 support, may be slower, may be faster for other codecs
      better cbp handling and small speedup in mpeg12
      reinitialize on aspect change,
      few bistream fixes and verbosity tweaks
      whitespace cosmetics
      parse and save hrd_fullness and range_map
      Mark all 12 block unused when there are no coded blocks
      Use native bswap32 instruction when __CPU__ is x86_64 instead of generic 386 code.
      -s option have different defaults for ffmpeg and ffserver cases.
      get rid of unused entries in the middle of xvmc_render struct,
      Identify the loop_filter flag as such and remove the unused variables in the context.
      Inport the original intiger IDCT from MPEG2 reference decoder.
      Integrate reference mpeg IDCT into dsputil.
      remove file because of lincense issues
      revert commit 9603
      Remove inline from function that is exported.
      On bitstream error is_intra_more_likely() could use dsp.sad[0]() that will segfault when xvmc is in use.
      Describe the purpose of magic in xvmc structure.
      describe idct variable in xvmc struct
      explain chroma_format field in xvmc struct
      remove assertion that use external definitions
      Rename xvmc flags to follow ffmpeg AV_ notion.
      vertically align xvmc struct member comments
      doxygen style comments
      clarify xvmc struct fields usage distribution (internal communication)
      use ff_xvmc for function names instead of XVMC_
      remove some useless ()
      remove obsolete comment
      get rid of the branch prediction
      Clarify TODO comment
      Document that ff_draw_horiz_band() could be used as hwaccell callback.
      Clarify some assert usage.
      Cosmetic 64 instead 8*8
      More clarification for draw_horiz_band() in hwaccell mode.
      Give nice message for failed assert.
      Return error instead of triggering assert.
      Another attempt to clarify draw_horiz_band() usage.
      Check ff_xvmc_field_start() result in all cases.
      Merge preprocessor if CONFIG_MPEG_XVMC_DECODER and code if( &&)
      Cosmetics
      Deprecate some unused xvmc structure fields.
      Move the state field in the xvmc struct to the extensions section.
      Detailed description of the interesting fields in xvmc struct.
      Remove obsolete display_flags member from xvmc struct
      Give a little better name to the magic(_id) field in xvmc struct
      Revert 17308.
      A little better wording of doxygen comments.
      Rename xvmc_render_state to something with pixfmt in its name.
      Proper condition in xvmc decode mb.
      Check all critical xvmc struct fields in ff_xvmc_field_start()
      Almost cosmetics.
      Cosmetics. Vertical align.
      Kill the magic field in xvmc structure.
      Document all functions in mpegvideo_xvmc.
      Rename unique_id to the way michaelni likes it. (xvmc_id)
      Describe all fields in xvmc struct.
      Improvements in xvmc structure fields descriptions.
      Minor clarification of xvmc field description.
      Thoroughly check all fields set by the application in xvmc struct.
      Exit with error if MPV_common_init() is called with PIX_FMT_NONE.
      Add more paranoid checks of xvmc token.
      Change the type of pblocks from pointers to short array into

Ivan Schreter (26):
      Replace hard-coded SEI type constants with symbolic names
      Fix a const related warning
      Add SEI recovery point frame counter to H264Context and use it when
      Parse initial_cpb_removal_delay_length.
      Parse dpb_output_delay in SEI picture structure.
      Parse cbp_removal_delay in SEI picture structure.
      Store CPB count in the context.
      Parse buffering period (H.264, D.1.1).
      Add key_frame to AVCodecParserContext, used in libavformat.
      Silence two warnings:
      Add convergence_duration to AVCodecParserContext.
      Use last decoded SPS as current SPS in order to parse picture timing SEI
      Make the following H264 functions available to the parser:
      Parse NAL units in H264 parser.
      Set context variable key_frame in H264 parser.
      Set context variable picture_structure in H264 parser.
      Use context variable repeat_pict for frame duration computation and
      Change duration computation to use time_base instead of TB/2.
      Correct time_base of H.264 and repeat_pict.
      Correct time_base and repeat_pict for MPEG2 video.
      Add timestamp computation if values are exported by decoder.
      Export timestamp parameter from H.264.
      Initialize H264 parser context correctly if the optional SEI messages
      Correctly reset SEI variables.
      Improve frame rate guessing for streams with two fields per frame.
      Change TS seeking so it returns position/timestamp of a key frame.

Ivan Wong (1):
      ffplay crashes patch by (Ivan Wong: email, ivanwong info)

Ivo van Poorten (64):
      cast pointers to make compiler happy
      add missing declaration for av_md5_sum
      compile md5 and include in libavutil.a
      Fix segmentation fault for gray16le to gray conversion.
      Add support for grayscale images with arbitrary maxvals.
      cosmetics after last commit
      Remove code duplication in sws_rgb2rgb_init.
      cosmetics after last commit; remove superfluous braces
      Replace huge switch statement by a lookup table.
      cosmetics, expand all tabs to four spaces.
      New implementation of rgb32tobgr32
      slightly faster rgb32tobgr32; avoid one add and one cmp
      skip MMX code in rgb32to15 if the size of the input is smaller than the
      skip MMX code in rgb32to16 if the size of the input is smaller than the
      skip MMX code in rgb24tobgr24 if the size of the input is smaller than the
      test widths all down to one. this tests whether the scalers correctly switch
      fix red_15mask and green_15mask. the previous values were wrong and caused
      preserve alpha channel when converting rgb32 to bgr32 in plain C
      add V.Flash PTX decoder
      add AV_[RW][BL]64 support
      generic S3TC DXT1 and DXT3 decoding functions
      Renderware TXD decoder
      Renderware TXD demuxer
      add doxygen comments
      Renderware TXD decoder and demuxer
      add V.Flash PTX image format
      add Renderware TXD format
      add myself for txd and s3tc files
      autodetect ptx files
      Fix rgb15ToUV. Correct order of components and use one shift less.
      remove the need of one extra shift in rgb16ToUV
      cosmetics; one test per line for all isXXX(x) macros
      ffmpeg -sws_flags command line option to set swscaler flags
      -sws_flags command line option to set swscaler flags
      -sws_flags command line option to set swscaler flags
      typo ;)
      remove useless casts
      remove useless assignment
      remove useless assignment
      PC Paintbrush PCX image decoder
      remove useless cast
      Sun Rasterfile decoder
      Sort properly
      remove forgotten debug statement
      add PCX and Sun Rasterfile decoders
      add myself as maintainer for pcx.c and sunrast.c
      TechnoTrend PVA Demuxer
      TechnoTrend PVA Demuxer
      use one shift less in get_pts()
      remove unnecessary shift
      proper type and initialization of pva_pts
      change audio codec id to mp2
      validate streamid before use
      do not return an error, but warn and recover when encountering an audio packet
      proper check for valid pva_pts
      cosmetics, superfluous space
      Move parsing of MPEG-PES timestamp to mpeg.h (as an inline function) so it
      use ff_parse_pes_pts for parsing an MPEG-PES timestamp
      remove unnecessary check
      cosmetics
      Implement read_timestamp and enable seeking in PVA files.
      Linux Media Labs MPEG-4 (LMLM4) demuxer
      LMLM4 demuxer
      add some missing formats: lmlm4, pva, pcx and sun rasterfile.

Jacob Meuser (5):
      -DPIC patch by (Jacob Meuser // jakemsr jakemsr com)
      OpenBSD support patch by (Jacob Meuser // jakemsr jakemsr com)
      amd64 as canonical system type patch by (Jacob Meuser <jakemsr jakemsr com )
      ldconfig as config.mak variable patch by (Jacob Meuser: jakemsr, jakemsr com)
      Make shared lib name handling consistent with the other Makefiles.

Jai Menon (15):
      8SVX decoder
      IFF demuxer
      Add IFF demuxer and 8SVX audio decoder to changelog and documentation.
      IFF stereo support
      fix verbatim mode decoding. patch by Jai Menon (realityman gmx net).
      alacenc: last few hunks approved by michael
      alacenc: compare against ALAC_MAX_LPC_ORDER instead of MAX_LPC_ORDER
      Update changelog and general.texi to reflect
      alacenc : perform decorrelation only for stereo samples
      fix issue 616 on roundup : decoding of short flac files
      vorbis_enc : set coded_frame->pts value to running sample count.
      Fix an overflow in the predictor.
      Avoid _t in identifier names, _t is reserved by POSIX.
      Use libopenjpeg for jpeg2k decoding.
      Check sampling index validity for AAC decoding

James A. Morrison (2):
      whitespace cleanup patch by (James A. Morrison <ja2morri>@<csclub>dot<uwaterloo>point<ca>)
      Build libpostproc out of source directory patch by (James A. Morrison >ja2morri csclub.uwaterloo ca>)

James Klicman (1):
      AltiVec optimized fdct patch by (James Klicman <james at klicman dot org>)

James Morrison (2):
      vis detection patch by (James Morrison <ja2morri at csclub dot uwaterloo dot ca>)
      License change and cpu detection patch by (James Morrison <ja2morri at csclub dot uwaterloo dot ca>)

James Stembridge (1):
      set pix_fmt in rv10 patch by (James Stembridge <jstembridge at users dot sourceforge dot net>)

Jan Knutar (4):
      NetBSD build fix: A forgotten $extraflags caused -lm to fall off.
      Unconditionally use -lossaudio on NetBSD.
      Add missing #include, patch by Jan Knutar, jknutar nic fi.
      VIS OBJS should end in .o, not .c; patch by Jan Knutar, jknutar nic fi.

Jan Kratochvil (1):
      AMR-WB for 3GPP patch by (Jan Kratochvil - kratochvil) / bugreport #1187101

Janne Grunau (1):
      free thread context

Jason Garrett-Glaser (19):
      Use a state machine to reduce branching logic in decode_cabac_residual.
      H264: Fix decoding with CABAC/delta_qp/PCM macroblocks.
      Improve documentation about -bt option
      Small h264 decoding speed optimization: while->do/while
      Fix incorrect printing of brainfart cropping error in some cases in progressive mode.
      x264 has removed the b-rdo and bime options, and instead integrated
      Replace i by trailing_ones, part of a patch by dark shikari.
      Simplify chroma AC in CABAC residual decoding.
      Move filter_luma_intra into dsputil for later addition of asm.
      Add automatic prefix handling to yasm functions.  Does nothing now, but will
      Port x264 deblocking code to libavcodec.
      H.264 loopfilter speed tweaks
      Fix compilation without optimization under 64-bit with x264 deblock asm enabled.
      Add x264 SSE2 iDCT functions to H.264 decoder.
      50l: Reenable error resilience for H264 (Patch by Dark Shikari).
      Faster VC-1 C loopfilter using lots of xor magic
      Fix 10L in r16670 (broke deblocking code)
      Remove (incorrect) filenames from x264 asm files, add descriptions.
      Convert x264 asm files to proper unix line breaks

Jason Millard (1):
      allow to enable or disable every bitstream filters individually

Jason Tackaberry (1):
      -fPIC support for libswscale

Javier Garcia (1):
      Allow RGB Raw encoding

Jay Freeman (1):
      fix mjpega bitfilter when dht, dqt, sof0 are not ordered as expected, patch by Jay Freeman, saurik at saurik com

Jean Delvare (2):
      Add support for the PIX_FMT_RGB555 and PIX_FMT_RGB565 pixel formats in
      Add support for the RGB32 pixel format in video4linux2 devices

Jeff Downs (47):
      Avoid a crash when H264 MMC operation causes the short term reference
      h264/PAFF preparation: use DELAYED_PIC_REF to mark non-refs frames as held for delayed output
      Cosmetic preparations for h264/PAFF implementation.
      Rename MMCO stuff to prepare for h264/PAFF implementation.
      Partial PAFF implementation at macroblock level.
      Modify unreference_pic implementation with PAFF in mind.
      Re-indent unreference_pic.
      Further modularize short reference list management for upcoming PAFF implementation.
      Rename variable to make sense in both field and frame contexts (support of PAFF implementation).
      Fix h->curr_pic_num for field pictures. Necessary for proper PAFF support.
      Fix Picture.key_frame setting to be compatible with frame and field contexts. Part of PAFF implementation.
      Reorganize long reference management to minimize code duplication in upcoming PAFF implementation.
      Support functions and changes to default reference list creation for PAFF.
      Reindent fill_default_ref_list after changes for PAFF
      Support function and changes to reference picture reordering for PAFF.
      Modifies macroblock addressing and current macroblock y-position for field decoding.
      Set Picture.poc for fields and field pairs. Part of PAFF implementation.
      Edge emulation for fields. Part of PAFF implementation
      Augment mmcodecoding process to work properly with fields. Part of PAFF implementation.
      Cosmetic renaming variable so that it makes more sense for forthcoming PAFF mmco patches.
      Augment MMCO execution to work with both fields and frames. Part of PAFF
      Re-indent after PAFF MMCO implementation patch.
      Manage Picture buffers for fields as well as frames. Pair complementary fields into one MPV Picture.
      fix incorrect check for abs_diff_pic_num overflow
      Interlaced pictures (not just MBAFF) + spatial direct mode is not
      Simplify tests for field macroblocks
      Fix chroma mv offsets for PAFF in a way that is compatible with MBAFF by
      Move setting of key frame flag for intra-only codecs, ignoring key frame
      Correct assignment of interlaced_frame; was being set on output frames,
      Fix poc for field pictures. Prior to this, the poc of the second field
      Fix deblocking filter for field pictures by properly looking for first
      Call mpegvideo flush routine on h264 flush. Needed in particular
      When new streams are detected, look for old filters and close them out.
      Initialize function pointers used by error resilience code before any
      Fix non_zero_count_cache for deblocking in field pictures.
      Make h264 decoder conform to requirements of mpegvideo's MPV_frame_start.
      Actually return with an error condition if we're being asked to deal with too
      Ensure that our total reference frame count does not exceed the SPS
      Fix brain-dead parts of r11216, specifically:
      Fix memory leak in h264 decoder. allocated_edge_emu_buffer was being allocated
      Move decode_significance_x86() and decode_significance_8x8_x86() to
      Revert 12838 to redo it the right way (use svn copy to create new
      Redo r12838, this time using svn copy to create h264_i386.h from cabac.h.
      Cosmetics:
      Fix reference to spec to be more specific and correct.
      And.... Fix another spec reference to be more specific and correct.
      Initialize outputed_poc to INT_MIN at decoder initialization.

Jeff Muizelaar (3):
      simpler delta decreasing algorithm patch by (Jeff Muizelaar <jrmuizel at student dot cs dot uwaterloo dot ca>)
      * simplification patch by Jeffrey Ryan Muizelaar
      fix some type mismatches patch by (Jeff Muizelaar <muizelaar rogers com>)

Jens Rosenboom (1):
      The v4l2 driver can silently change the pixel format in the

Jeremy Kolb (2):
      Store license value in a variable for later use.
      Add version information to DLLs.

Jeremy Mordkoff (1):
      Fix typo.

Jindřich Makovička (15):
      support for negative strides
      Fix upside-down picture for BGR24 images (fixes pig-loco-rgb.avi)
      support for both orderings of the slices (top->down / bottom->up)
      check for negative strides before memcpy
      malloc padding to avoid reading past the malloc()ed area.
      new qscale type to distinguish H264
      - fix insufficient code length for exp_vlc
      Kill a warning with MSVC
      Proper fix for r8963
      always use the whole buffer for reading w/ packetized sources to avoid packet truncation
      Allow the UDP socket buffer size to be adjusted using a
      check for interrupt when receiving from socket
      Get rid of MSG_DONTWAIT using a more standard way to use a socket
      bind to the multicast address if possible
      Fix minor memory leak.

Joakim Plate (11):
      fix ac3 and dts detection (patch by Joakim Plate <joakim.plate at ecce.se>)
      Fix A/V (de)sync with discontinuous NSV streams,
      add support for all framerates specified by the standard
      Fix a crash when probing img2 format with a NULL filename.
      timestamps generation improvement when parsing avi
      fix emulated inet_aton so that it fails for invalid addresses
      Check return of url_fseek in mpeg timestamp reading code.
      matroskadec: set cur_dts after seek
      aspect stored in nuv files must be converted from display- to sample-aspect.
      Fix handling of old nuv files where an aspect of 1 actually means 4/3
      Allow AC-3 parser to modify codec_id.

Joel Yliluoma (1):
      16-bit and 15-bit rgb/bgr patch by (Joel Yliluoma <joel dot yliluoma at w-create dot com>) (note, rare formats disabled)

Joey Parrish (1):
      removed -W -Wall -g and other gcc-specific warning/debug options

Johan Bilien (1):
      rtp_callback: send number of mb patch by (Johan Bilien {jobi via.ecp fr)

Johannes Carlsson (2):
      * missing part of the 3gp patch by Johannes Carlsson
      amr typefix patch by (Johannes Carlsson <joca at rixmail dot se>)

Johannes Feigl (1):
      looks better

John Dalgliesh (7):
      os x md5sum patch by (John Dalgliesh <johnd at defyne dot org>)
      Support for Mac OS X Intel, part 3: binary integer constants:
      Add support for Mac OS X Intel part 2: Assembler macros in fdct_mmx.c
      Support for MacIntel, take xx: '/nop' illegal for old versions of GAS
      Support for MacIntel, last part: balign directives
      It's not possible to portably echo a backslash, use printf instead.
      force 16-bytes alignement of block array, as needed by the implementation of ff_dct_sse2

John Donaghy (1):
      DVR-MS probe audio format

John Ryland (1):
      fixing mp3 detection patch by (John Ryland <jryland at trolltech dot com>)

John Schmiederer (1):
      Use tkhd matrix for proper display in mov.

JonY (4):
      Create .lib files in the subdirs.
      Use pkg-config for libschroedinger libs, not only for CFLAGS.
      Properly use av_log rather than fprintf/printf.
      Undefine the macro which defines fprintf as please_use_av_log so that

Josh Coalson (2):
      Add support for FLAC's new RICE2 entropy coding method. Patch by Josh Coalson.
      Add decoding support for variable block size FLAC files using the latest FLAC

Josh Varner (1):
      removing --disable-risky patch by (Josh Varner <jlvarner gmail com>)

Juanjo (59):
      Added CODEC_ID_H263P to fourcc tables.
      Added support to Unrestricted Motion Vectors (UMV)
      Moved some H.263+ variables to MpegEncContext to be thread-safe.
      Added Juergen Keil fix on quant_tab[4] type.
      Restore the main Makefile to have debug flags.
      - Bug fix on inter MCBPC table for inter+q.
      - Fixed a bug on H.263 MV prediction for MB on GOBs limits.
      - Bug fix on MV prediction for MPEG4 caused by new H.263 GOB code.
      - Bug fixed on H.263 decoder initialization.
      - Added support to Inter4V+Q MBs to H.263 decoder.
      - Preliminary RTP friendly mode for H.263.
      - Fixes on RTP and GOB headers for H.263.
      - Now the ME is done for the entire picture when enconding, the
      - More work on preliminary bit rate control, just to be able to get an
      - Minor changes on bitrate control
      - H.263+ decoder support for Advanded INTRA Coding (buggy)
      - Bug fixes in H.263+ Advanced INTRA Coding decoder.
      - Bug fix on RTYPE (rounding type) not being honoured by H.263+ decoder.
      - Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
      - Fix on AVI and WAV headers based on Nikolai Zhubr patch.
      - Added MP3 encoding through libmp3lame contributed by  Lennert Buytenhek.
      - Added MP3 encoding through libmp3lame contributed by Lennert Buytenhek.
      - Added MP3 encoding through libmp3lame contributed by Lennert Buytenhek.
      - Added MP3 encoding through libmp3lame contributed by Lennert Buytenhek.
      - Bug fix on output sample rate for lame MP3 encoding.
      - Bug fix MPEG-2 decoder to handle "repeat_first_field" (Telecine)
      - Fixes on MPEG demuxer to handle VOBs without System Header.
      - Added video coding statistics for ffmpeg. Needs more work.
      - Fix pts calculation on mpeg mux (A/V sync) - Thanks to Lennert Buytenhek
      - Fix memory leak and others bugs for ppmpipe. Thanks to Rudolf Opalla.
      - Added force_pts to av_write_packet() to be able to force PTS, this helps
      - Commented out PIX_FMT_ANY by now. It seems that not all codecs are setting
      - PIX_FMT_ANY = -1 to avoid breaking any stuff that could be using it.
      - Fixed AC3 decoding for 5:1 AC3 streams. Now when calling av_audio_decode for
      - New switch to activate 4MV in ffmpeg.
      - ME setting moved to AVCodecContext/MpegEncContext, no longer a global.
      - ME method compatibility with legacy apps.
      - Versions defined for libav/libavcodec.
      - Changed FOURCC for H263.
      - Advanced Intra Coding (AIC) support for H.263+ encoder, just DC by now.
      - H263 PAR support by Alex Beregszaszi.
      - MPEG-4 B frames coding option for ffmpeg.c
      - Segfault fixed when liba52 dynamic library isn't found.
      - Segfault fixed when mpeg audio decoder returns a negative data_size.
      - Changed default qmin = 3 as ffmpeg.c
      - Updating docs....
      - Warning fix.
      - HQ option for ffserver. Thanks to D Richard Felker III.
      - Bug fix in map stream parser
      - pict_type exported to AVCodecContext
      - Fixes for MPEG-PS which have more than 1 system header, now we don't take
      - repeat_pict added to AVCodecContext to signal if the decoder must repeat
      - ffmpeg relies on libavcodec to repeat frames from 3:2 sources.
      - VCD MPEG-1 compliant stream support (set AVF_FLAG_VCD)
      - Video CD option for ffmpeg
      - repeat_pict meaning changed, now it signals the extra delay for the
      - Added r_frame_rate in AVStream to store the real frame rate.
      - Now ffmpeg handles MPEG-2 streams with pull down, it gets the real frame
      - Bug fix for AVStream->r_frame_rate not being initialized for live video

Julian Scheid (1):
      The current_sample field is meant to be the sample number, not an offset into

Justin Ruggles (358):
      While adding stereo rematrixing, I came across something that needs to
      configure libgsm patch by (Justin Ruggles: jruggle, earthlink net)
      vbr audio encode patch by (Justin Ruggles: jruggle, earthlink net)
      fixed quality / vbr encoding patch by (Justin Ruggles, jruggle, earthlink net)
      Implement audio cutoff frequency to the vorbis encoder.
      AC3: support encoding fractional frame sizes
      Raw flac muxer, patch by Justin Ruggles (jruggle earthlink net). Can be
      CODEC_CAP_SMALL_LAST_FRAME patch by Justin Ruggles jruggle earthlink net
      flacenc - rice param search patch by (Justin Ruggles  jruggle earthlink net
      flac-lpc patch by (Justin Ruggles  jruggle earthlink net)
      porder patch by (Justin Ruggles jruggle  a t earthlink d ot net)
      allows user-settable block size and fixes related typo
      Add lpc order search.  This creates new compression levels 6 to 12.
      Modify preset compression levels to utilize log search.
      cosmetic change to adjust alignment.
      Fix 2 bit allocation bugs.  One fix enables using a higher bandwidth code.  The other fixes an issue with floorcod=7.
      add read_probe function to raw ac3 demuxer
      fix parsing of RealAudio AC-3/DolbyNet
      cosmetics. remove space before semi-colon.
      simplify AC-3 bit allocation
      Create ac3.c which will be used for AC-3 common code.
      Remove common code from AC-3 encoder and utilize ac3.c.
      remove redundancy in AC-3 parser by using common tables from ac3tab.h
      move AC-3 header parsing to ac3.c
      the AC-3 parser needs ac3.o
      hardcode ff_ac3_frame_sizes table
      improve ac3_probe by counting consecutive frames
      split ac3_parametric_bit_allocation into 3 separate functions
      fix handling of special case for lowest snroffset. regressions are unaffected.
      add myself as AC-3 maintainer
      utilize multi-stage AC-3 bit allocation. speeds up encoding by 25-30%
      simplify
      cosmetics. align and condense.
      Update doc to show FLAC encoding
      enable multichannel output in dca decoder.
      use downmixing coefficients in dca decoder.
      cosmetics: indention after last commit
      AC-3 decoder, soc revision 9, Jun 14 03:52:02 2006 UTC by cloud9
      AC-3 decoder, soc revision 26, Jul 5 04:55:15 2006 UTC by cloud9
      AC-3 decoder, soc revision 31, Jul 14 23:53:28 2006 UTC by cloud9
      AC-3 decoder, soc revision 32, Jul 17 09:37:32 2006 UTC by cloud9
      AC-3 decoder, soc revision 33, Jul 17 12:13:40 2006 UTC by cloud9
      AC-3 decoder, soc revision 34, Jul 18 14:27:54 2006 UTC by cloud9
      AC-3 decoder, soc revision 38, Aug 7 00:03:00 2006 UTC by cloud9
      AC-3 decoder, soc revision 40, Aug 9 00:10:14 2006 UTC by cloud9
      AC-3 decoder, soc revision 48, Aug 16 11:27:49 2006 UTC by cloud9
      AC-3 decoder, soc revision 52, Aug 16 22:45:07 2006 UTC by cloud9
      AC-3 decoder, soc revision 53, Aug 17 08:53:44 2006 UTC by cloud9
      AC-3 decoder, soc revision 54, Aug 17 09:13:08 2006 UTC by cloud9
      AC-3 decoder, soc revision 55, Aug 17 12:25:24 2006 UTC by cloud9
      AC-3 decoder, soc revision 56, Aug 17 14:21:58 2006 UTC by cloud9
      AC-3 decoder, soc revision 57, Aug 19 12:44:38 2006 UTC by cloud9
      AC-3 decoder, soc revision 58, Aug 19 14:20:45 2006 UTC by cloud9
      AC-3 decoder, soc revision 69, Aug 31 07:12:56 2006 UTC by cloud9
      AC-3 decoder, soc revision 70, Aug 31 18:04:23 2006 UTC by banan
      AC-3 decoder, soc revision 71, Sep 4 09:47:12 2006 UTC by cloud9
      AC-3 decoder, soc revision 72, Sep 7 04:20:00 2006 UTC by cloud9
      AC-3 decoder, soc revision 109, Sep 22 22:42:54 2006 UTC by banan
      AC-3 decoder, soc revision 402, Jul 14 13:45:14 2007 UTC by jbr
      make ac3dec.c GPL until violations are fixed
      revise license header for simplicity and consistency
      10l to me. I accidentally removed a word.
      use random number generator from libavutil
      move some common values to ac3.h and utilize them
      ABS -> FFABS
      remove a debugging function and variable, and remove the commented-out code which uses them.
      no need for the decoder to have the license as a prefix
      use shared ac3 bit allocation function
      fix the 256-point MDCT
      remove portions of bit allocation which are no longer needed since they are shared with the encoder
      cosmetics: indentation
      new implementation of exponent decoding
      revise comment for exponent decoding function
      update license header. add copyright for myself. correct and complete the statement about liba52.
      don't need to define ALT_BITSTREAM_READER
      get rid of an ugly macro
      don't need to synchronize in the decoder. the parser does that already.
      Use shared AC-3 header parsing. Move some initializations to block parsing function.
      cosmetics: indentation
      change some bit masks to arrays
      use consistent data type for exponents. needs to be int8_t to pass to bit allocation.
      use enum value instead of numerical value for acmod
      set bit allocation parameters directly instead of copying
      only use sized data types when necessary
      simplify snr offset and bit allocation
      remove broken downmixing. will add new implementation later.
      BLOCK_SIZE is not used properly, just use numbers instead to avoid confusion
      change 2 fields from context to local variables
      simplify stereo rematrixing
      align array used in mdct. fixes segfault.
      split out coupling coefficient reconstruction into a separate function
      merge decoding of coupling channel coefficients and regular channel coefficients
      simplify coefficient decoding
      dither zero-bit mantissas by default. remove dithering only if it's explicitly turned off.
      cosmetics: indentation
      use floating-point mantissa dequantization tables
      use table for converting dynamic range codes to scale factors
      use array instead of bitmask
      change AC3DecodeContext variable to local variable
      cosmetics: use braces when it makes the code easier to read.
      make exponent decoding consistent (prepare for merging coupling and lfe code with normal channel code).
      move some variables to a more local scope.
      cosmetics: use more appropriate names for iterator variables.
      there are only 4 rematrixing bands
      consolidate decoding of lfe and coupling channels with decoding of full-bandwidth channels.
      cosmetics: indentation
      use dsputil for float to signed 16-bit sample conversion
      add support for downmixing to stereo or mono
      increase the maximum AC3 probe buffer size and reduce the threshold for 50% reliability score to 2 frames.  Fixes out.ac3 from [FFmpeg-devel] AC3 Format Detection unreliable.
      additional tweaks to AC3 probe function. give a higher score to a single frame
      10l to me. Revert recent changes to ac3_probe() which made misdetection as AC3 too probable.
      add dialogue normalization
      use array for dynamic range scaling factors
      add pointer to the parent context for use with av_log()
      cosmetics: blank lines, remove some comments
      enable the native AC-3 decoder
      comments modification: add, clean up, Doxygenize, and reword
      cosmetics: vertical alignment
      fix decoding of DolbyNet AC3
      cosmetics: rename variables. tbl->tab for consistency with other AC3 tables
      fix random dithering of zero-bit mantissas
      better AC3 header error reporting
      skip incomplete frames
      dnet audio needs avparser to work with the lavc ac3 decoder.
      cosmetics: rename ac3 tables
      cosmetics: rename ac3 bit allocation variables
      cosmetics: vertical align after last commit
      cosmetics: rename common ac3 variables
      cosmetics: rename ac3 decoder variables
      downmix before the IMDCT if no block switching is used
      Revert commit made in revision 11228. I'm getting some strange results in the
      Skip dialog normalization. It shouldn't be used by the decoder, and there is
      Add option for user to scale the amount of dynamic range compression which is
      fix typo
      Do not give advice for which bitrate to choose.
      use bps instead of kbps for bitrate
      use cutoff frequency to adjust bandwidth in ac3 encoder
      remove unused context variable
      cosmetics: rename GetBitContext gb to gbc
      cosmetics: rename all AC3DecodeContext variables from ctx to s
      cosmetics: vertical realignment after last commit
      do not need to cast a (void *)
      cosmetics: rename sampling_rate to sample_rate
      pass an AC3DecodeContext to ac3_downmix() instead of multiple arguments
      simplify
      add crc check to ac3 decoder
      remove workaround which is no longer needed due to
      cosmetics: indentation after last commit
      only check ac3 crc if AVCodecContext.error_resilience > 0
      cosmetics: indentation and add a TODO comment
      adjust output channels based on AVCodecContext.request_channels in AC3 parser
      do not set channels before checking request_channels
      check for request_channels at codec init
      cosmetics: indentation
      use request_channels in liba52 decoder
      do not set channels before checking request_channels
      remove unneeded variables from AC3HeaderInfo
      move E-AC3 header parsing to ff_ac3_parse_header()
      cosmetics: indentation after last commit
      add downmixing support to libfaad decoder
      check for request_channels in dca init function
      fix playing of multichannel audio for codecs without downmixing
      fix phase flag processing for the case when coupling coordinates are reused
      make sure that the last coupling band stops at the end of the coupling range
      remove unneeded variable from AC3DecodeContext
      split out init of downmix coeffs into a separate function that can be called
      used defined name for testing error resilience level
      Use 24-bit fixed-point transform coefficients until just before MDCT. This
      pre-calculate downmix coefficient sums. 2% faster 5.1-to-stereo downmixing.
      use scale factor instead of sum for downmix level adjustment. avoids division
      downmix before imdct unless different size transforms are used. about 20%
      make input buffer const, as it should be. fixes a warning.
      Pass AACAC3ParseContext to sync() instead of individual arguments. Patch by
      additional protection from segmentation faults and memory access errors by
      cosmetics: indentation after last commit
      only allocate context input buffer if AVCodecContext.error_reslience is greater than 0.
      fix Ogg/FLAC header size
      Use 15-bit LPC precision by default. This generally gives better compression
      remove redundant context variable
      remove unused variable, min_framesize
      fix encoding of flac private data
      change function parameters for metadata_streaminfo()
      change function parameters for dump_headers()
      split out some decoder context params to a shared macro
      share streaminfo parsing function
      move call to init_get_bits inside conditional
      include avcodec.h
      fix reading of samples-per-frame
      add check for invalid channel mode when coupling is used
      add checks for missing information in the first block
      10l: fix error in commit r13382
      cosmetics: indent
      add error concealment
      cosmetics: indent
      set audio parameters if frame header is ok, even if crc fails
      safer solution for setting audio parameters
      parse audio in ogm demuxer.
      zero the upper frequencies of the correct coefficients
      move header error logging to after CRC check
      use enum value for CRC error
      get_transform_coeffs() never returns an error, so make the function and its children return void.
      new dba values requires running last 2 bit allocation stages
      10l: cosmetics
      cosmetics: remove unneeded braces
      only calculate number of exponent groups when exponents are not reused.
      set lfe params at frame init instead of in every block
      simplify
      cosmetics: add a comment, remove a comment
      move some AC3HeaderInfo struct fields to the Coded elements group.
      prevent crash when dba params are invalid
      return error on frame sync error
      get the number of blocks from the ac3 parser and use in the ac3 decoder.
      get substreamid from ac3 parser
      read frame type from header info into decode context
      skip unsupported frame types and substream id's
      cosmetics: indent
      set default mix levels regardless of bitstream id
      pass bap table to ff_ac3_bit_alloc_calc_bap()
      add more gain levels and adjust mix level tables accordingly.
      move mix level tables from parser to decoder. have parser read bitstream value instead of using an index to a table in the decoder.
      move the decode context and some macro constants to a new file, ac3dec.h
      cosmetics: reorder decode context fields into logical groups
      cosmetics: vertical align
      use macro constants for array sizes in decode context
      cosmetics: vertical align
      cosmetics: add and align references to spec variable names
      store cpl_in_use for all blocks in decode context
      store exp_strategy for all blocks in decode context
      split up header parsing function
      add fields needed for E-AC-3 decoding to decode context
      add ac3dec_data.[c,h]
      move rematrixing band table to ac3dec_data.c
      cosmetic: use hex value instead of decimal to show relation to 0x7FFFFF
      read 'coupling in use' parameter from bitstream into decoder context first,
      rearrange dimensions of pre_mantissa array
      use MAX_BLOCKS for array size instead of 6
      cosmetics: vertical alignment
      remove a comment which is no longer true
      cosmetics: make all references to AC-3 capitalized and hyphenated
      use the same table for ungrouping exponents and bap=2 mantissas
      Hardcode the table for ungrouping 3 values in 5 bits. It will be reused by
      Read bitstream parameters for E-AC-3 streams before returning an error.
      remove unneeded local variable
      cosmetics: change function name and comments to refer to block decoding
      cosmetics: use a better function name than uncouple_channels()
      Add a generic function to lavc to log messages about missing features.
      use limited range of lpc orders when quantizing coefficients
      cosmetics: adjust line breaks and vertical alignment
      use range of lpc orders in ORDER_METHOD_EST
      update my email address to one which does not depend on my service provider
      fix and simplify frame size check and reencoding in verbatim mode
      use LFG instead of Mersenne Twister for AC-3 PRNG
      remove some unneeded header includes
      rename AC3_MAX_FRAME_SIZE to AC3_FRAME_BUFFER_SIZE and increase the size
      get data size based on the actual data array instead of coding-in the calculation
      commit the OKed parts of the E-AC-3 decoder
      cosmetics: indentation
      add some approved parts of the E-AC-3 decoder
      cosmetics: indentation
      add more OKed parts of the E-AC-3 decoder
      fix decoding of coupling leak info for normal AC-3
      change eac3dec.c license to LGPL
      cosmetics: rename some functions from *get_* to *decode_*
      1000l to me. committed to wrong repo. revert last commit.
      restore accidently deleted file
      right shift the values by 8 in GAQ remap "B" table. data now fits in int8_t.
      cosmetics: rename some functions from *get_* to *decode_*
      add more OKed parts of the E-AC-3 decoder: structural changes to support AHT mode mantissa decoding
      copy default coupling band structure from table to decoding context (used for E-AC-3)
      add definitions and documentation for shared functions for E-AC-3 decoding
      do not read exponent strategies from each block for E-AC-3 since they are in the header
      cosmetics: indent after last commit
      cosmetics: rename a table
      remove unneeded variable, dither_all
      cosmetics: indentation
      cosmetics: change "get" to "decode"
      cosmetics: change another "get" to "decode"
      cosmetics: update comment about dithering
      commit more OKed parts of the E-AC-3 decoder
      commit more OKed parts of the E-AC-3 decoder
      commit more OKed parts of the E-AC-3 decoder
      cosmetics: indentation
      merge declaration and init. variable is not used outside the loop.
      skip converter exponent strategy for all channels at once
      use correct table name
      simplify code and comment regarding determination whether or not AHT is used.
      turn on E-AC-3 decoding support and update the Changelog
      add E-AC-3 to supported codec list
      Dnet AC-3 encoding is integer-only
      cosmetics: capitalization
      the FLAC encoder has an integer-only mode available (compression_level <= 2)
      fix sorting of audio codecs
      create a separate codec_id for E-AC-3
      check ENABLE_*_DEMUXER instead of CONFIG_*_DEMUXER
      change back to using CONFIG_*_DEMUXER for ac3_eac3_probe(), but use it
      add a separate line in Makefile for E-AC-3 decoder and only compile it when
      Return error when trying to decode non-grayscale 16-bit PNM images.
      add error message for unsupported codecs in WAVE.
      write number of samples in FLAC extradata.
      update FLAC regression test due to expected change after r15324
      write MD5 checksum to FLAC header. update regression tests.
      write actual min and max frame size to FLAC header. update regression test checksum.
      cosmetics: fix indentation after r15321
      cosmetics: rename two variables
      simplify decoding of coupling frequency range
      add a line break in error message text
      add a generic function to decode banding structure for coupling, spectral
      add support for spectral extension
      cosmetics: indentation
      silence a gcc warning about using an uninitialized variable
      cosmetics: vertical alignment
      revert r15812 (E-AC-3 Spectral Extension) pending further review
      ac3dec: skip decoding and use error concealment for remaining blocks in
      ac3dec: use uint8_t for band sizes
      silence gcc warning: ‘n_bands’ may be used uninitialized
      ac3dec: detect out-of-range exponents
      cosmetics: reindent after last commit
      cosmetics: vertical alignment
      ac3dec: unroll inner loop in exponent decoding. 46% faster decode_exponents().
      ac3dec: use skip_bits_long() for block start bits
      ac3dec: add missing feature log message for block start info
      ac3dec: simplify exponent bounds checking
      ac3: detect dba errors and prevent writing past end of array
      cosmetics: alignment after last commit
      flacdec: add support for 3 additional sample rate codes
      rename flac.c to flacdec.c
      add myself as a maintainer of flacdec.c
      flacdec: remove some unused code
      flacdec: remove unneeded debugging statements
      flacdec: remove another unneeded debugging statement
      flacdec: change several AV_LOG_DEBUG to AV_LOG_ERROR
      cosmetics: indentation
      flacdec: cosmetics: use consistent coding style (K&R)
      flacdec: cosmetics: indentation
      flacdec: cosmetics: some line wrapping at 80 chars
      define FLAC metadata types in flac.h
      move FLAC_STREAMINFO_SIZE to flac.h
      use function and definitions from libavcodec/flac.h in oggparseflac.c
      flacdec: get total number of samples from STREAMINFO
      flacdec: add support for SAMPLE_FMT_32
      cosmetics: add some documentation to FLACContext
      cosmetics: change comment type and vertical alignment
      flacdec: set default sample_fmt before parsing extradata
      use av_malloc() in vorbis_comment()
      add myself as maintainer for lpc.[ch]
      Set the correct number of samples for E-AC-3 in the AC3 parser.
      Separate the raw FLAC muxer from raw.c to its own file, flacenc.c.
      Use a shared function to validate FLAC extradata.
      cosmetics: line wrap and indentation after last commit.
      Add support for full header extradata to raw FLAC muxer.
      cosmetics: add a comment in flac_write_header().
      Share the function to write a raw FLAC header and use it in the Matroska
      Separate the raw FLAC demuxer from raw.c and put in a new file,
      cosmetics: indentation
      add some missing codecs and update some names to match AVCodec.long_name
      Add some missing codecs and update names to match the codec long_name
      Add some more missing codecs and update names to match the codec long_name.
      AC-3 encoding is integer-only

Jürgen Keil (4):
      Remove a few obsolete header includes (fixes a compile error on solaris)
      16x{8,16} {avg,put}_pixels support patch by (Juergen Keil <jk at tools dot de>)
      segfault fix patch by (Juergen Keil <jk at tools dot de>)
      mlib fix patch by (Juergen Keil <jk at tools dot de>)

KO Myung-Hun (2):
      Use check_lib2 instead of check_lib in SDL test.  This is helpful in
      Check param in sws_getCachedContext().

Kamil Nowosad (4):
      Fix a bug in av_find_opt(). Because some of the AVOption structures have field
      FF_CODER_TYPE_RAW / RLE / DEFLATE extracted from a patch by Kamil Nowosad
      s/invert/photometric_interpretation/ patch by Kamil Nowosad k.nowosad  students mimuw edu pl
      YUV support patch by (Kamil Nowosad  k.nowosad  students mimuw edu pl)

Kareila (1):
      fixing bitrate display if -acodec copy is used, patch by (Kareila <cokewench at yahoo dot com>)

Karl H. Beckers (1):
      Add PIX_FMT_PAL8 to isPacked().

Keiji Costantini (6):
      cosmetics: Break some overly long lines.
      cosmetics: Break some overly long lines.
      cosmetics: Break an overly long line.
      Unroll codepath
      Force inlining of MPV_motion_internal.
      Remove in_asm_used_var_warning_killer()

Ken Mandelberg (1):
      Fixed the brokeness and also bgr32 should work now. Patch by Ken Mandelberg <km at mathcs.emory.edu>

Kenan Gillet (20):
      Rename acelp_math.[ch] to celp_math.[ch] to prepare for QCELP decoder merge.
      Split off celp_filters.[ch] from acelp_filters.[ch] for the QCELP decoder.
      OKed parts of QCELP data tables
      More OKed parts of QCELP data tables
      Add ff_dot_productf() to celp_math.{c,h}
      Add a LPC filter
      Fix numbers in Doxygen comments, patch by Kenan Gillet, kenan.gillet gmail com.
      OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      Fix typo in table value.
      More OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      More OKed parts of the QCELP decoder
      Commit last ok'ed parts of QCELP decoder and enable it.
      Avoid the 'Claimed bitrate and buffer size mismatch' warning storm.
      Split ff_log_missing_feature into ff_log_missing_feature
      Change sign in ff_celp_lp_synthesis_filterf(). This makes this function

Kenneth Aafløy (1):
      avformat/av_seek_frame_binary with growing file patch by (Kenneth Aafløy: kenneth, aafloy net)

Kevin Baragona (3):
      Spelling errors patch by (Kevin Baragona | kevinmb500 gawab com)
      More spelling errors patch by (Kevin Baragona (kevinmb500 gawab com>)
      Even More spelling errors. patch by (Kevin Baragona <kevinmb500 gawab com)

Kevin Kuphal (2):
      ASF extended header parsing patch by ("Kevin Kuphal" <kevin dot kuphal at sitecity dot net>)
      seeking in single stream asf patch by ("Kevin Kuphal" <kevin dot kuphal at sitecity dot net>)

Kohn Emil Dan (1):
      The reader ignores the size of the ASF data object and keeps on

Konstantin Andreyev (3):
      split asf patch by (Konstantin Andreyev <kandreyev at bcsii dot com>)
      asf patch by (Konstantin Andreyev <kandreyev at bcsii dot com>)
      forgot to commit from patch by (Konstantin Andreyev <kandreyev at bcsii dot com>)

Kostya Shishkov (445):
      When playing
      handle AVERROR_IO and AVERROR_NOMEM patch by (Kostya <cannonball bw-team com>)
      VMD playing error patch by (Kostya (cannonball bw-team com>)
      support some more color modes; patch by Kostya
      security fixes
      Two tables in truemotion1data.h were slightly different from VpVision
      ZMBV 15-/16-/32-bit decoding. 24-bit mode is disabled because it's not
      Register myself as a maintainer
      16-bit DPCM and stereo support for Sierra VMD
      More correct demuxing and timestamp setting fot Sierra VMD
      Smacker demuxer and decoder.
      Don't use get_vlc2() when tree is one symbol. This fixes audio decoding
      Fix demuxing of AVIs with pad streams.
      KMVC (used in Worms games) decoder. Works fine with samples from MPHQ
      Now MPlayer should understand Smacker audio and video codecs.
      Proper PTS generation for Smacker audio
      Palette support for newer KMVC
      Rename VC9 to VC1 as there is no such thing as VC9
      VC-1 decoder with I-frames support and partial P-frames decoding
      VC1 Intra blocks in P-frames use different zigzag table than I-frames.
      Added loop filtering as ersatz for overlap filter (improves picture quality for coarse quantization).
      This should make P-frames decoding work on x86 (by avoiding get_bits(0))
      Some optimization and fixes - mostly reworked MC and bitplanes.
      typo
      Slightly reworked MC code, it gives better picture.
      Use MpegEncContext->mbskip_table instead of custom bitplane.
      Remove unused variable (and it wasn't free'd too)
      0xFFFF l of cola. Now P-frames are decoded almost without distortions.
      Adjust AC prediction if (some) predictors are not available.
      4-MV decoding support
      vc1_decode_intra_block() simplifications and corrections
      4-MV mode final fixes (now it works for non-exotic modes)
      Correctly detect when use hpel or qpel mode
      Optimizations suggested by Michael Niedermayer
      Correctly choose global transform mode, MV mode and fix bitplane decoding
      Implement intensity compensation
      Fix one case of choosing MV for chroma
      Don't read out-of-bounds values.
      Enable coefficient scaling
      Improve chroma MC: correct case for FASTUVMC=1, use slower but correct /2, and always use halfpel MC.
      Use s->first_slice_line in checks instead of s->mb_y
      Use overlap filter instead of loop filter.
      Fix mquant selection for ALL_MBS mode
      MVRANGE may occur in all frames and RESPIC in all but B-frames
      Check if 4-MV MB should not be decoded.
      Rounding control for motion compensation.
      Correct inverse quantization for P-frames
      Don't pass block[6][64] parameter to decode_p_mb as we always use s->block
      Set correctly quantizer and transform mode when parsing frame header.
      Set DC = 0 for skipped MB as it interferes DC prediction
      1000l: scale DC predictors instead of predicted value.
      Don't do non-uniform quantization on zero coeffs.
      Replace code for clipping MV vectors (which is wrong to use here) with clipping source coords.
      Correct edge MC for chroma
      Some fixes for decode_p_mb()
      Some B-frames support (parsing and decoding only, no motion compesation is done)
      Add rounding for overlap filters
      Support grayscale decoding.
      Support range reduced frames
      Set MpegEncContext->mspel flag (here it indicates that bicubic MC will be use)
      New functions in DSPContext for VC-1 decoding
      Use functions from DSPContext in vc1.c
      Use bicubic MC (should also remove those ringing artifacts) when needed
      Fix overflows in bicubic interpolation.
      Forgot to update LIBAVCODEC_VERSION_INT
      Some AltiVec optimizations for VC-1
      Don't try to decode P-frames from old WMV3 variant until their format is figured
      Looks like to be a good time to re-enable WMV3 decoder
      100l: intensity compensation must be also done in 4-MV mode
      Bicubic interpolation requires two additional pixels for block
      Correct MC for B-frames and some improvements (not 100% complete though)
      Disable B-frames decoding until I return and finish their proper decoding.
      Motion compensation for luma always use halfpel precision.
      Spelling
      New qpel MC functions conforming to VC-1 standard.
      New function for chroma MC (will be used in VC-1)
      Replace usage of put_*_vc1_qpel_pixels_tab with put_*_h264_chroma_pixels_tab
      Drop put_vc1_qpel_pixels_tab as they won't be needed anymore.
      AltiVec version of put_no_rnd_h264_chroma_pixels_tab[0] (slightly changed version of put_h264_chroma_pixels_tab[0])
      VC-1 Advanced Profile support (progressive only, tested on WVC1 samples)
      WVC1 is now decodeable with VC-1 decoder.
      VMware Video decoder (fourcc: VMnc)
      Handle raw blocks correctly (both updating pointer and storing to memory)
      Try to handle all chunks, previous scheme was not correct.
      Cursor drawing support
      Tests for overreading input data
      Handle the rest of known blocks (by skipping them)
      Proper support for B/BI frames
      [Cosmetics] Remove some done TODOs/FIXMEs from comments
      [Cosmetics] Remove if(profile < PROFILE_ADVANCED) from decode_sequence_header()
      Don't print header bits by default
      Support for BI-frames in Advanced Profile
      Drop unneeded checks before av_free() and change to av_freep() where it's more suitable.
      Disable B-frames decoding until they produce correct picture
      2989l: Set avctx->has_b_frames value in header and don't change it
      Fix edge emulation for bicubic MC in B frames
      Use h264_chroma funcs in B-frames MC too
      Set motion vectors used in B-frames to zero by default
      Swap back and forward motion vectors to achieve correct picture
      Zero number of current scan after decoding one picture.
      Reorder MV order in B-frames so no swapping in vc1_b_mc() is needed
      For B-frames if the second reference frame signals intensity compensation
      Finally enable B-frame decoding as all known samples work perfect
      WavPack lossless audio decoder
      bump lavc version for WavPack decoder
      free memory before return
      Handle case of get_bits(0)
      1l: correct argument order in avcodec_check_dimensions
      10l: remove alien seek function from demuxer
      Targa image decoder
      TIFF decoder
      Move block size check to decoder
      Make PNG produce correct 8-bit pictures
      Ignore blocks with no samples and flags (but usually with MD5 sum)
      handle limit>32 for set_ur_golomb_jpegls()
      Register .tga and .tiff image extensions
      16-bit grayscale support
      PNG 16-bit gray decoding support
      PGM 16-bit gray support
      JPEG-LS 16-bit gray support
      16-bit grayscale support
      10l predictor should not skip first line
      LZW decoder as separate module plus TIFF LZW support
      IMC decoder
      Init vlc tables as static
      Use bpp from header in error message
      RGB32 support in Targa
      Fraps v2 and v4 support
      Update comment
      Correct support for Fraps v4 (and Huffman tree for < 256 symbols)
      Limit output sample to 0..maxval, fixes decoding of T16E0.JLS
      WavPack support in Matroska
      Check for symbol count
      Bump version for Fraps v[24] support
      Output proper 16-bit sound (fixes decoding on PPC)
      Simplify VMD demuxer (but it still does not work right)
      Correctly detect 4-component images
      Warn about JPEG in TIFF
      Rows per strip may be >= height
      8-bit images support for TIFF
      Print error message for unsupported mode (RGB planar,CMYK,YCrCb)
      Decoder can decode many ZLib-packed strips
      Use table for determining type sizes
      Update error message
      Give error about planar RGB
      Support for 16-bit grayscale
      Overlap filtering was done in reverse order
      Fix comments
      Optimize overlapping
      Divide first audio buffer chunk into atomary bufffers.
      Identation
      Some progressive JPEG support
      Decode non-interleaved scans
      Use halfpq in coefficient scaling
      100l: Initialize dc_scale with current quantizer for adv I frames
      B-frames support for adv. profile
      Correct rounding values in overlap filtering
      Correct information in header
      VQA v1 support
      Better VQA demuxer
      100l FastUVMC rounding was incorrect.
      10e6l FastUVMC was done right but in the wrong place
      Some VQA v1 files don't have audio stream
      Do not touch pkt->pts
      ZMBV encoder
      Register myself as maintainer of zmbvenc.c too
      1e6l forgot to add zmbvenc.c
      Now ZMBV encoder compiles
      Deobfuscate expression
      Musepack SV7 decoding support
      Enable forward seek in Musepack demuxer
      Make MPC demuxer deal with ID3 tags at the beginning
      Decode previous 32 frames to avoid seeking artifacts in MPC
      Fail on VC-1 interlaced streams
      Off-by-one error fix
      Add debug messages for AP headers
      Some TIFFs do not set rows per strip for single strip.
      Set aspect ratio if present (for AP only)
      Coded residual in WavPack may be > 0xFFFF
      Seeking support in WavPack
      Fix BI-frames decoding for Adv. profile
      Check for out-of-bounds access
      General approach to parsing chunks in VC-1 AP
      0x00000103 is not an escape marker
      VC-1 parser
      Set codec dimensions if provided by extradata
      These messages are just for debug purposes
      Set frame rate from extradata
      Save frame header info in context variables
      table fix
      Set codec dimensions in extradata parsing
      Wrong flag tested
      B-frames could not be determined from broken_link/closed_entry, use fixed value
      Progressive frames disguised as interlaced are supported
      Raw VC-1 demuxer
      Intensity compensation for B-frames in AP was missing
      10l: forgot break statement
      Use AVRandom
      100l: forgot to reset LS mode on new frame
      DCA decoder
      WMV3 Complex Profile initial support
      Looks like I'll work on it this year too
      Correctly ME border blocks
      DXA demuxer and decoder
      Proper handling of motion vectors for adv. profile
      indentation
      Zero MV vectors for P frame intra blocks
      Correctly handle data_size on decoding
      Smacker Audio stereo fix
      1000l to myself - get correct quantization for blocks 1 and 2
      Remove redundant assign
      One more FASTTX-related hack
      FASTTX==0 means additional 16 bits in extradata
      vc1_split should be static
      This codec is PAL8 so make it output PAL8 too
      Set C predictor to zero if unavailable (should fix B-frame border artifacts)
      Pullback should be performed on scaled motion vectors in B-frames
      1000l to myself as used VLC indexes were totally wrong
      Support for newer MacOSX AIFF
      Register myself as DCA maintainer
      1000l to myself -  DCA decoder should not be registered in video codecs section
      Move some declarations into header
      Typo
      100l to myself. Do not include stuff unneeded by parser
      Move variables to vc1data.h
      Make vc1_parser.c compilable without special defines
      Move VC-1 data to .c file
      Add prefix to some variables for consistency
      Use correct dequantizer value
      Add ff_ prefix to (externally) visible variables
      Do proper frame swapping so VMD video decoder works again
      Typo
      this prediction direction adjustment is wrong
      Drop redundant assignment and variable
      Rename flag for consistency with the next commit
      Take ring frame into account when demuxing.
      Move H.264 intra prediction functions into their own context
      Set correct length for last slice in frame subpacket
      Monkey Audio decoder
      Add checks on input/output buffers size for some audio decoders
      Check unp_size for possible overflows too
      In case when any quantizer may occur, HALFPQ should be zero
      HALFPQ should be used only for when PQUANT is selected
      Enable error resilience in WMV3/VC-1 decoder
      Register myself as APE maintainer
      Do not decode more data than output buffer may hold
      Guard against output buffer overflows
      Looks like this APE decoder support versions starting from 3.95
      Fix decoding of VMDs representing sprites (Last Dynasty, Woodruff).
      Do not try to decode more data than output buffer may hold
      Remove check for input buffer size as it does not guarantee that
      WMV3 mode with FASTTX=0 seems to use standard IDCT instead of own 8x8 transform
      Demux full frames instead of sliced for RealVideo.
      Make RV10/20 decoder work with new frame format
      indentation after last commit
      New RM format deserves small version bump
      Beam Software SIFF demuxer and video decoder
      Remove incorrect and unneeded code
      Use simple IDCT for now
      Revert get_num() to old behaviour in order to handle
      Split Musepack decoder into SV7 decoder and synth core
      Seek regression for RM demuxer fix
      Make get_v() available to the other demuxers
      Add ff_ prefix to get_v()
      Musepack SV8 demuxer and decoder
      X8 is supported now, so warning is incorrect
      MP3 encoding is not integer-only (it's up to LAME)
      Data tables for future RV30/40 decoder
      TIFF may omit compression tag.
      RV30/40 common VLC tables
      Fix regression tests after r11127 and 11129
      Fix muxer so that generated files are playable by realplayer, other open
      RealVideo 3 decoder functions
      Rename function to reflect its action
      RV40 decoder specific functions
      Update tables for RV40 decoder
      Register myself for future RV30/40 decoder
      Replace table with its H.263 counterpart
      Add decoder flags
      Intra types will be stored in int8_t array
      Check decoded dimensions for validity
      Use existing function for VLC reading
      Rename some tables
      Switch VC-1 decoder to output decoded residual immediately.
      Use IDCT functions when FASTTX=0
      Update Altivec variant of vc1_inv_trans_8x4
      Reindent after last commit
      Add 4x4 IDCT
      Enable 4x4 IDCT for FASTTX=0 mode in WMV3
      Benjamin expressed the wish to be DCA maintainer too
      Make fill_rectangle() available for other decoders
      RV30/40 decoding core
      RV40 decoder should use availability cache
      Correct spatial prediction mode in RV30/40 for vertical left direction
      Direct blocks should use motion vectors from the second reference frame
      Make decode210() common function.
      RV30 thirdpel motion compensation support
      Fractional parts of motion vectors should be accounted separately too
      pqindex=0 is forbidden
      Save coded block patterns for future loop filtering.
      Save quantizer values as well
      Factor out constant addition
      simplify 4-point transform part a bit
      SMPTE 421 Annex L format demuxer
      IMC decoder always operates on 64-byte blocks
      Do not modify input data
      IMA ADPCM encoder - QT variant
      Add step variable (for future v5 decoding)
      Fraps v5 decoding support
      Correctly handle empty frames
      indentation
      DCA frame size is constant
      Correct company name
      Missing tags in APE is normal situation, so don't print an error.
      Skip blocks in B-frames reuse motion vectors from next reference frame.
      Calculate motion vector information based on PTS provided in slice header
      Move MLP/TrueHD from video to audio codecs section.
      Cosmetics: align
      Add several vector functions used by Monkey's Audio decoder to dsputil
      Altivec implementation of APE vector functions
      Monkey's Audio decoder vector functions work on input with length
      SSE2 optimizations for Monkey's Audio decoder vector functions
      In-loop deblocking filter for VC-1 decoder
      SMPTE 421M Annex L (aka .rcv) muxer
      R and B components are stored as a differences to G component in Fraps v5.
      Simplify colour conversion in Fraps as suggested by Michael
      Okayed parts of AAC encoder
      Remove not OKed parts from AAC encoder
      Add approved chunks to AAC encoder
      Synchronize AAC encoder with renamings in aac.h
      More okayed chunks of AAC encoder
      Add okayed parts for AAC encoder
      Add okayed chunks of AAC encoder psychoacoustic model
      Do not pass unnecessary arguments to AAC encoder functions
      Simple lowpass filter implementation.
      okayed chunks of AAC encoder
      Add generic IIR filter interface with Butterworth lowpass filter implementation
      While APE changed container format in 3.98, frequency tables for range coding
      Make it possible to disable loop filter in VC-1.
      Factorize out code used for MS RLE format decoding in different decoders.
      Add fourcc for MS RLE4
      ZMBV encoder forgot to fill coded frame pointer
      add newline at the end of message
      Add myself to BMP format maintainers (with permission from Mans on IRC)
      Calculate line size correctly for bit depths < 8
      Add support for 1-bit, 4-bit, 8-bit and some 16-bit raw BMP
      Add RLE4 and RLE8 decoding support for BMP
      Cosmetics: reindent after last commit
      Correct wrong lower limit and condition used in APE decoder
      Register RV-[34] block specific feature for long future loop filter
      Mark some of RV3/4 block types as having separate DC subblock
      RV3 and RV4 decoders set some deblocking coefs for each macroblock,
      34l: free RV3/4 context variables at the end
      Choose RV3/4 DC quantizer correctly
      Fix chroma motion compensation for RV30
      Detect split RV30 slices
      Correct motion vector scaling in B-frames for RV3/4
      Introduce RV3-specific motion vector prediction.
      Calculating an additional MV-based deblocking pattern is the same
      Drop redundant check in RV3/4 deblock coefficients calculation
      Split RV3/4 deblock pattern into horizontal and vertical parts
      Weak deblock filter function for future RV40 loop filter
      421l: postproc flag may be present in any VC-1 AP frame type
      Add channel layout to several audio decoders I maintain
      Use wider margins for edge emulation in RV30/40
      Eliminate direct use of s->current_picture in RV30/40 decoder
      33l Fix applying residue condition in RV30/40 decoder
      Strong filtering function for future RV40 loop filter
      Add known BMP header sizes.
      Give more meaningful message on BMP header parsing error
      Some BMP files have file size declared in the header equal to headers size
      Fix RV40 chroma MC
      Invoke future RV30/40 loop filter for already decoded rows instead of
      RV40 loop filter
      Add support for parsing and decoding DCA-HD streams.
      RV40 MC functions
      enable RV40 decoder
      Remove unused RV40 spatial intra prediction function
      Update ff_rv34_decode_frame() arguments definition
      Check RV30/40 slice offsets to be inside buffer.
      34l: forgot to change one case of chroma MC for RV40
      Fix reading out of buffer during RV30/40 deblock mask calculation
      Fraps version 3 seems to be the same as Fraps v5
      AASC can contain raw data in addition to MS-RLE
      Print out some errors found during MS RLE decoding
      Make decoding run of pixels in MS RLE saner.
      Do not attempt to decode RV30/40 B-frames without anchors.
      4l: TIFF stores short strings inside tag, do not interpret it is as an offset
      Unfortunately RV30 luma thirdpel MC functions cannot be substituted with
      Generate correct deblock pattern for RV30
      Merge deblocking pattern with CBP for RV30/40 loop filtering
      use precalculated mb_pos where available
      add const qualifier to some pointers for input data
      cosmetics: remove one unneeded space
      a table for upcoming RV30 loop filtering
      RV30 loop filter
      RV30 decoder
      Register TIFF tags for CCITT Group 3 and 4 compression options
      Calculate line size variable correctly for lower bitdepths and use it for raw data copying
      Add monochrome TIFF support
      Some TIFFs declare -1 for the single strip height
      Decode TIFF image only after all tags have been decoded
      352l: correct calculating number of bits for storing macroblock offset in RV3/4
      Latest Coktel Vision VMDs contained Indeo 3, add demuxer support for it
      Fix a typo that made VMD demuxer always assume Indeo 3 as video codec.
      It turned out that RV30 uses motion vectors for forward motion B-frame
      RV30 decoder passes possible frame sizes in extradata and selects
      30l: when reading mb_offset in RV30 slice header its size
      Cinepak strip and chunk sizes are 24-bit, not 16-bit as it was thought earlier.
      Cinepak strip ID is a single byte
      421l: fix loop filter for P-frames in VC-1
      Print VC-1 aspect ratio in debug
      VC-1 display dimensions should affect coded dimensions only (I think)
      WavPack hybrid mode support
      VMD first chunk of audio is not coded as many separate chunks of block_align
      cosmetics: reindent after last commit
      Empty audio chunks in VMD are silent, not skipped
      yet another attempt on fixing synchronization in VMD
      VC-1 postproc field is 2 bits wide while decoder read only single bit.
      Remove the block with always false condition from vc1_parse_frame_header()
      Fix AAC data smuggling:
      Fix Doxygen comments for APE decoder.
      Fix Doxygen comments for RV3/4 decoder.
      Fix Doxygen comments for VC-1 decoder.
      Silence useless compiler warning when passing AVFrame* instead of AVPicture*
      Silence two pointer assignment compiler warnings in rv34.c
      Factorize stream reading in TM2 decoder
      Make TM2 decoder byteswap input into separate buffer instead of doing it in-place.
      Make WNV1 decoder use temporary buffer for bit-reversed input
      Silence compiler warnings for XL decoder
      Add decoder flush after seeking for RV3/4 decoders
      Move RV3/4 frame type check up so 1-slice unanchored B-frames are checked, too.
      Give better name to Inverse_Table_6_9
      Prefix visible YUV2RGB functions with sws_
      Initialize s->gob_index for Intel variant of H.263 too
      In RV3/4 nonzero MB offset for the first slice also indicates error
      Enable dropping frames for RV3/4
      Some AltiVec functions in SwScaler produce different output than their
      Make hScale_altivec_real() trim its output like other implementations do
      Make BMP decoder try to decode files with incorrect filesize field value
      New LGPLed YUV2RGB table generator for SwScaler
      Update regressions for YUV2RGB table generator change in SwScaler
      Better parsing of i263 picture header
      Since motion vectors for intra blocks are always zero, do not try to retrieve them.
      cosmetics: reindent after last commit

Kristian Amlie (1):
      replace the auto array "duration_error" with a malloced array.

Kurosu (1):
      av_malloc vs av_mallocz patch by (Kurosu <kurosu inforezo org>)

Kurtnoise (3):
      Surround '#include <sys/select>' by HAVE_SYS_SELECT_H.
      actually write dac3 atom, patch by Kurtnoise, kurtnoise at free dot fr
      matroska: add dirac support (patch by Kurtnoise  kurtnoise _at_ free _dot_ fr)

Lars Täuber (3):
      Nitpicks patch by Lars Täuber.
      LPCM 24 bits support, patch by Lars Täuber, lars.taeuber gmx net
      support dvd pcm 20/24 bits, patch by Lars Täuber, lars.taeuber gmx net

Laurent Aimar (8):
      H264 fourcc patch by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      h264 - progressive I frame CABAC support patch by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      progressive P frame CABAC support patch by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      h264 hurry up fix and a tiny cabac clean patch by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      Fix an invalid free in the AVI muxer when an unsupported audio track is added.
      Prevent a division by 0 in the g726 decoder when the configured samplerate is 0.
      mlp: initialize all CRC tables in a common function.
      Check extradata is large enough.

Laurent Desnogues (2):
      Mark constant structure member as const to avoid some warnings.
      Add command line support for '-idct simplearmv6'.

Lennert Buytenhek (4):
      be less verbose patch by (Lennert Buytenhek <buytenh at math dot leidenuniv dot nl>)
      dont discard first frame patch by (Lennert Buytenhek <buytenh at math dot leidenuniv dot nl>)
      unmap video buffers on close patch by (Lennert Buytenhek <buytenh at math dot leidenuniv dot nl>)
      dont needlessly gettimeofday in video grabber patch by (Lennert Buytenhek <buytenh at math dot leidenuniv dot nl>)

Leon van Stuivenberg (10):
      handle DRI/RST markers patch by Leon van Stuivenberg <leonvs at iae dot nl>
      ps2 idct patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
      ps2 idct bugfix patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
      ps2 optimizations update patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
      dsputil mmi cleanup patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
      clear_blocks_mmi() optimization by BroadQ (and patch by Leon van Stuivenberg <leonvs at iae dot nl>)
      new versions; put_pix's about 20% faster, get_pix 3.5x faster. patch by (Leon van Stuivenberg <leonvs at iae dot nl>)
      tcp select() check and enables pressing 'q' when reading/(writing) from
      decode latency patch by (Leon van Stuivenberg <l dot vanstuivenberg at chello dot nl>)
      support url_read which reads less then requested patch by (Leon van Stuivenberg <l dot vanstuivenberg at chello dot nl>)

Likai Liu (1):
      use vec_splat_u32 to initialise a v16ui (patch by Likai Liu < liulk _at_ cs.bu.edu >)

Limin Wang (16):
      add support for more pixel format (yuv422p, yuv444p, etc instead of yuv420 only.)
      fix frame size option, patch by Limin Wang, lance lmwang gmail com
      remove unused variables
      Remove unnecessary the_end label.
      As avcodec_decode_audio() is deprecated, replace it with avcodec_decode_audio2().
      remove unused code, patch by Limin Wang <lance lmwang at gmail com>
      cosmetics: Fix indentation, patch by Limin Wang, lance.lmwang gmail com.
      Explain a few more options that were recently added.
      add fps stats, patch by Limin Wang <lance lmwang at gmail com>
      Fix H.264 8x8 transform selection
      Allow setting v4l2 input and video standard.
      add 'all' and 'none' options for partitions config
      add a check before find_frame_end call
      Allow avoid setting the video standard and input when capturing v4l2 video.
      fix imlib2 memory leaks
      Set data_size to zero when DCA header parse failed

Loic Le Loarer (6):
      add multi slice support
      H.264 b ref pic list order and long term pictures patch by (Loic Le Loarer <loic.le-loarer polytechnique org>)
      H.264 multiple slice support in CABAC patch by (Loic (lll+ffmpeg m4x org)
      INTRA PCM macroblocks support patch by (Loic )lll+ffmpeg m4x org)
      remove goto and reindent patch by (Loic Le Loarer <lll+ffmpeg m4x org>)
      MBAFF I slice no deblocking patch by (Loic >>lll+ffmpeg m4x org<<)

Lord (1):
      Workaround adjustments for DivX 6.1 and later

Loren Merritt (282):
      shape-adaptive diamond + prediction crash patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      h264 deblocking crash patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      Always write VOL control parameters, needed by old xvid patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      10l (scene change pict_type was wrong after coded_frame fix) patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      (non)interlaced huffyuv patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      simplify getsymbol patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      h.264 ref list reordering bugfix patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      H.264 CABAC + B-frames patch by (Loren Merritt <lorenm at u dot washington dot edu>)
      Fixes a segfault when flushing delayed B-frames on the 2nd pass. patch by (Loren Merritt <lorenm.u at washington@edu>)
      Fixed typo which caused incorrect motion prediction in B-frames. patch by (Loren Merritt <lorenm ta u tod washington tod edu>)
      new option: context_model
      split ffhuffyuv into 2 codecs:
      fix interlacing check in v1 huffyuv.
      vismv for h264 displayed incorrect motion vectors.
      implement B_DIRECT and B_SKIP macroblock types.
      Sort B-frames into display order.
      reduce stutter if we learn too late that the stream contains B-frames.
      H.264 weighted prediction.
      10l: scratchpad could be allocated before its size was known.
      Fix a crash when multiple metadata packets are present.
      10l in implicit_weight_table
      In-loop deblocking for B-frames.
      10l (deallocated memory again)
      vismb: display 4x4 partition boundaries in H.264
      vismv didn't display 8x8 mvs in H.264
      oops, other half of the vismv patch
      H.264: decode arbitrary frame orders and allow B-frames as references.
      10l (fix reordering when B-frames are not present)
      remove an assert that was breaking compilation on some gccs.
      Set keyframe flag only on IDR-frames (needed for reordering across I-frames).
      10l (compile fix)
      slice-based idwt (10% faster decoding)
      allow 16 refs with B-frames
      display deblocking strength in FF_DEBUG_PICT_INFO
      allow 2pass ratecontrol. also fixes psnr displayed by mencoder.
      deblocking filter doesn't need to call fill_caches again. 1.4% faster decoding.
      fix ref comparison in B-frame deblocker ('unused' and 'unavailable' were treated as different)
      MMX for H.264 deblocking filter
      clear the DPB after seeking.
      fix storage of motion vectors for frames with more than 4096 macroblocks.
      H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
      fix crash on negative POC
      memory leak
      decode h264 end-of-slice flag
      MMX for H.264 iDCT (adapted from x264)
      10l (symbol mangling)
      sort H.264 mmx dsp functions into their own file
      decode H.264 with 8x8 transform.
      typo in cabac tables
      H.264 lossless mode
      fix emulated edge motion compensation when resolution%16!=0
      6% faster decode_cabac_residual
      sse2 16x16 sum squared diff (306=>268 cycles on a K8)
      h264 mmx weighted prediction. up to 3% overall speedup.
      decode custom quant matrices.
      h264 custom quant matrices
      10l: cqm weren't initialized in svq3.
      faster h264_chroma_mc8_mmx, added h264_chroma_mc4_mmx.
      fix a crash on dropped frames
      25% faster deblocking, 4-6% overall.
      tidier cabac_intra4x4_pred_mode
      fix a crash on seeking to an invalid frame
      use sse16_sse2() in nsse
      recognize some more nal types
      fix some potential arithmetic overflows in pred_direct_motion() and
      fix direct_8x8_inference_flag with 4x4 mvs.
      direct mvs in B_8x8 mbs were set too early.
      CQM: fix fallback to JVT scaling lists
      fix asymmetric CQMs with pre-transposed DCT.
      10l: bidir_refine didn't save the new mvs. also improve speed.
      tweak inlines.
      tweak h264_biweight
      faster iterative_me: avoid duplicate mc of neighboring blocks.
      obmc-aware 4mv
      detect when iterative_me fails to converge, and stop searching
      revert 1.83 because it's not needed
      fix convergence of snow iterative_me
      fix crash on height < 16 pixels.
      15% faster put_h264_qpel16_hv_lowpass_mmx2
      11% faster put_h264_qpel16_v_lowpass_mmx2
      18% faster put_h264_qpel16_mc[13]2_mmx2
      10l (x86_64)
      slightly faster loopfilter
      10l in 1.12
      h264: special case dc-only idct. ~1% faster overall
      fix some crashes when missing frames
      fix some crashes on negative nalsize.
      don't arbitrarily limit delta_qp
      minor simplifications in cabac_mb_type
      fix a memory leak on seeking.
      4% faster h264_qpel_mc
      keyframe flag was wrong if SEIs followed the keyframe
      remove some useless instructions
      gcc2.95 workaround
      h264_idct_add only needs mmx1
      h264_idct8_add_mmx
      obey vqscale in 2nd pass
      seed iterative_me with mvs from conventional search.
      seed iterative_me with mvs from conventional search.
      1.5x faster write_back_motion, 1-3% overall
      added mmx implementation of h264_chroma_mc2
      10l
      prefetch pixels for future motion compensation. 2-5% faster h264.
      prefetch pixels for future motion compensation. 4-10% faster mpeg1/2/4 decoding (on an athlon-xp).
      simplified and slightly faster h264_chroma_mc8_mmx
      10l in 1.199 (write_back_motion with B-pyramid + spatial)
      2x faster ff_mpeg4_set_direct_mv
      mmx implementation of 3-point GMC. (5x faster than C)
      gmc_mmx tweaks
      10l in h264 vismb/vismv (broke on 2006-03-09)
      memory leak.
      don't force asserts in release builds. 2% faster h264.
      warn on interlaced streams
      reinit quant matrices if pps is overwritten by another of the same pps_id.
      fix a crash on x86_64
      Snow 1pass ratecontrol
      Snow multiple reference frames
      h264: faster fill_rectangle()
      h264: MBAFF interlaced decoding
      Snow: 10l. 1pass ratecontrol failed to set snow's internal quant.
      Snow: fix subband weighting in wavelet cmp functions. use 32x32 cmp in iterative motion estimation.
      Snow: cosmetics
      Snow: scale predicted mv based on which reference frame the neighbors used.
      Snow: cosmetics
      Snow: don't try to encode lossless with 9/7 wavelet, because it isn't lossless.
      adpcm_ima & adpcm_yamaha: improved quantization
      ADPCM: trellis quantization
      h264: accept rbsp_trailing longer than 1 byte
      update regression tests for r5449 (adpcm)
      slightly faster h264 fill_caches()
      sync to x264 r537
      10l in r5768 (broke mbaff)
      int16_t is faster than int_fast16_t for division.
      sse & sse2 implementations of vorbis channel coupling.
      another div -> fastdiv, another 2% faster vorbis.
      10l, vorbis_inverse_coupling_sse() was really 3dnow
      3dnow2 implementation of imdct.
      slightly faster ff_imdct_calc_3dn2() on amd64. (gcc added a bunch of useless movsxd)
      gcc 2.95 and 3.4.x on x86 32bit without fomit-frame-pointer can't even find 5 registers for asm input.
      emms -> femms
      r5954 broke fft on cpus with 3dnow but without mm3dnow.h
      sse and 3dnow implementations of float->int conversion and mdct windowing.
      misc tweaks in vorbis_residue_decode().
      vorbis cosmetics: mdct0,mdct1 => mdct[2]
      vorbis simd tweaks
      ff_fft_calc_3dn/3dn2/sse: convert intrinsics to inline asm.
      use ff_fft_calc_3dn2 instead of sse on a K8.
      tweak cabac. 0.5% faster h264.
      mmx implementation of deblocking strength decision.
      slightly faster deblock
      fix snow 1pass rc: the qlog used and the one written to the bitstream weren't always the same.
      snow cosmetics: merge the sliced and non-sliced versions of add_yblock
      asm implementation of mid_pred.
      sse implementation of imdct.
      tweak ff_imdct_calc_3dn2
      cosmetics (indentation)
      simplify asm mid_pred, as suggested by michaelni.
      remove a division from STOP_TIMER
      approximate qpel functions: sacrifice some quality for some decoding speed. enabled on B-frames with -lavdopts fast.
      10l in r6412
      simplify write_back_motion for skip mbs
      don't try to inline cabac functions. gcc ignored the hint anyway, and forcing it would make h264 slower.
      more prefetches, 2% faster h264
      10l in r7174 caused all options to be treated as filenames.
      simplify in fill_caches
      huffyuv rgb encoder
      oops, encoder was under ifdef decoders.
      mmx 16-bit ssd. 2.3x faster svq1 encoding.
      move some stuff out of an inner loop. 12% faster svq1 encoding.
      doxy comment tweaks
      faster and simpler vp6 bilinear mc
      simplify
      simplify
      cosmetics: remove duplicate transpose macro
      cosmetics: remove code duplication in hadamard8_diff_mmx
      sse2 version of fdct_col.
      tweak mmx2 sad.
      tweak mmx2 sad.
      sse2 version of fullpel sad.
      10l, r8991 broke mmx1 sad
      fix a crash with svq1 + sse2.
      sse2 & ssse3 versions of hadamard. unroll and inline diff_pixels.
      factor sum_abs_dctelem out of dct_sad, and simd it.
      configure check for whether we can compile ssse3
      sse2 & ssse3 versions of dct_quantize.
      r3938 broke 2pass huffyuv (not that anyone uses it)
      cosmetics: macrofy duplicate code in huffyuv
      change brute force search to min-heap. 3.6x faster generate_len_table, 8% faster ffvhuff encoding.
      merge huffman tables so that we read 2 symbols at a time. 30% faster huffyuv decoding.
      add init_vlc_sparse(). faster than init_vlc() if there are lots of holes in the tables.
      use sparse huffman tables. 1.5% faster huffyuv decoding.
      cosmetics (reduce nesting)
      merge huffman tables for rgb too
      AC-3 decoder, soc revision 148, Sep 28 18:38:27 2006 UTC by banan
      prevent huffyuv from generating codewords of length 32. (regression in r9069)
      simd gmc wasn't impossible.
      optimize encode_residual_lpc()
      optimize decode_subframe_lpc()
      oops, potential overflow on really large blocks
      unroll encode_residual_lpc(). speedup varies between 1.2x and 1.8x depending on lpc order.
      oops, revert unrelated change
      2.5x faster compute_autocorr()
      gcc isn't smart enough to factor out duplicate stores
      replace brute force find_optimal_param() with a closed-form solution.
      was computing one more autocorrelation coefficient that was actually used
      sse2 version of compute_autocorr().
      encode_residual_fixed(): replace FIR with finite differences.
      unroll finite differences to avoid swapping registers.
      replace FIR with finite differences.
      20% faster lpc, 6% overall flac decoding
      simplify lpc
      div -> mul
      remove a branch from an inner loop.
      mbaff spatial direct
      fix mbaff diagonal neighbor mv
      "swap first two elements of L1 if L0 and L1 are identical" applies after splitting the lists into fields
      remove some movq in ff_h264_idct8_add_mmx. 225->217 cycles on core2.
      ff_h264_idct8_add_sse2.
      use mmx2/3dnow avg functions in avg_qpel*_mc00
      avg_pixels4_mmx2
      reduce code duplication
      clean up an ugliness introduced in r11826. this syntax will require fewer changes when adding future sse2 code.
      add qpel rounder once during hv rather than twice during hv and whatever it's averaged with
      ssse3 h264 motion compensation.
      constant was excessively aligned
      fix aliasing warnings. simpler too.
      put loop counter in a register if possible. makes some of the qpel functions 3% faster.
      sse2 h264 motion compensation. not new code, just separate out the cases that didn't need ssse3.
      use fewer registers in apply_welch_window_sse2
      30% faster ff_vorbis_floor1_render_list, 3% faster overall
      asm argument that might be in memory needs a size
      simd and unroll png_filter_row
      pseudo-simd add_bytes and diff_bytes
      indent
      cleaner way of writing packed-byte constants.
      remove a memcpy
      png filters
      simplify
      decode mng color decorrelation
      more aspect ratios added in one of the ammendments
      h264 chroma mc ssse3
      Fix H.264 interframe decoding when compiling with icc. Patch by Loren
      rename me=tesa
      simplify vorbis windowing
      exploit mdct symmetry
      10l, float_to_int16_interleave_sse/3dnow wrote the wrong samples
      unroll another common case in vorbis_residue_decode
      vorbis: reduce memory by reusing buffers
      float_to_int16_interleave: change src to an array of pointers instead of assuming it's contiguous.
      r14205 broke vorbis float_to_int on ppc
      yasm buildsystem
      import yasm macros from x264
      split-radix FFT
      indent
      optimize imdct_half:
      remove a float->double->float conversion.
      remove mdct tmp buffer
      mdct wrapper function to match fft
      cosmetics in dsp init
      use imdct_half in ac3
      use float_to_int16_interleave in ac3
      align the audio decoding buffer, since some codecs write to it with simd
      optimize ac3_downmix.
      oops, non-normalized downmix to mono contaminated add_bias for non-simd float-to-int.
      don't mark the delayed samples for upmixing if they haven't been downmixed
      simd downmix
      simd int->float
      special case 6 channel version of float_to_int16_interleave
      gcc chokes on xmm constraints, so pessimize int32_to_float_fmul_scalar_sse a little
      gcc chokes on the 7 registers needed for float_to_int16_interleave6 (even inside HAVE_7REGS), so write it in yasm
      missing prototype
      file which should have been added in r14749
      document some dsp alignments
      document some dsp alignments
      fix progressive jpeg:
      clear_block mmx
      1.6x faster pred*_vertical_add_c
      flatten an array, since gcc fails at optimizing multidimensional arrays
      fix an overflow in vc1 intra overlap filter
      re-enable mid_pred asm on x86_64. (broke in r16681)
      ff_add_hfyu_median_prediction_mmx2
      Add my GPG fingerprint

Loïc Le Loarer (9):
      - correct several errors on the deblocking accross slice boundaries.
      compile with TRACE define patch by (Loic <lll+ffmpeg m4x org>)
      correct long term picture management patch by (Loic <lll+ffmpeg m4x org>)
      H.264 multiplce instance in reference list patch by (Loic <lll+ffmpeg m4x org )
      The cvs version 1.103 of h264.c brokes 13 conformance streams, this
      This is the second patch for MBAFF support, this adds the deblocking
      CABAC support for MBAFF I frames patch by (Loïc Le Loarer | lll+ffmpeg m4x org)
      CABAC support for MBAFF I frames patch by (Loïc Le Loarer < lll+ffmpeg m4x org)
      Fix compile with --disable-muxers, patch by Loïc Le Loarer, lll+ffmpeg m4x org.

Luca Abeni (227):
      fix rounding errors with NTSC patch by (Luca Abeni <lucabe72 at email dot it>)
      don't compile ffserver if network is not enabled patch by (Luca Abeni <lucabe72 at email dot it>)
      sample_rate and channels in aac audio patch by (Luca Abeni <lucabe72 at email dot it>)
      A/V Synch in ffmpeg patch by (Luca Abeni <lucabe72 at email dot it>)
      YUV420P to UYVY422 conversion patch by (Luca Abeni <lucabe72 >< email >< it>)
      let user set codec even if CODEC_ID_NONE is default patch by (Luca Abeni: lucabe72, email it)
      MPEG4 streaming over RTP patch by (Luca Abeni: lucabe72, email it)
      Parameter added for user to set the cutoff frequency when encoding with faac.
      ffmpeg forgets to set AVFormatParameters::pix_fmt when capturing from a
      video4linux2 input support
      Set interlaced_frame and top_field_first according to coded_frame.
      Correctly set the interlaced_frame and top_field_first fields.
      Add video4linux2 detection to configure.
      ff_asts_init needs to be inside #ifdef CONFIG_MUXERS, remove some misplaced
      Add support for drivers that can change the capture size on VIDIOC_S_FMT
      Rename the "-loop" option defined in ffmpeg.c to "-loop_input" ("-loop"
      Fix crop & pad + rescaling
      Tell the user why video capture is failing
      Do not be too verbose
      Change img_convert() from img_fmt to img_fmt with img_copy()
      move enum PixelFormat from libavcodec to libavutil
      Introduce swscale interface in libavcodec
      Baptiste COUDURIER's padding patch (reworked by me a little bit).
      Make cropping, padding, and rescaling independent (after this patch,
      Cosmetic patch: fix indentation (broken by previous cropping / padding
      Disable v4l2 is videodev2.h is not "sanitized"
      Simplify the code used for assigning video_resample, video_crop, & video_pad
      Fix cropping & padding when capturing from a v4l(2) card
      Include the correct video4linux2 header
      Fix typo
      Properly set the image depth
      Move postproc ---> libswscale
      Remove unneeded code
      Move REG_* macros from libavcodec/i386/mmx.h to libavutil/x86_cpu.h
      make libswscale build system similar to the libav* one
      Move av_log() & friends to libavutil
      Move av_malloc(), av_realloc(), and av_free() from libavcodec to libavutil
      Install log.h
      Use libavutil in libswscale, and allow it to be built out of the mplayer tree
      Fix a corner case in sws emulation
      Replace MIN() and MAX() with FFMIN() and FFMAX()
      Remove sws_global_init() from swscaler emulation
      Add support for ffmpeg's pixel format names in libswscale
      Do not set the codec's pixel format in the format's write_header()
      Move ffmpeg.c to the swscale interface
      Move output_example.c and ffplay.c to the swscale interface
      Fix "-aspect" when "-pad*" is used
      Avoid segfaulting if the swscale context cannot be allocated
      Another PIX_FMT_ ---> IMGFMT_ conversion
      Enable swscale usage in ffmpeg
      Fix compilation as a win32 shared library (libavutil must be explicitly linked)
      Allow swscale emulation header to be included by C++ programs
      Allow to compile swscale tests
      Add some new pixel formats to libavutil
      Minimal support for the new pixel formats in libavcodec
      Fix the PIX_FMT_* ---> IMGFMT_* conversion (PIX_FMT_RGB32 is IMGFMT_BGR32)
      Remove the dependency of libswscale on img_format.h
      Improve swscale-example to test conversions between all the possible
      Add support for YUVJ formats
      Implement sws_getCachedContext() in swscale emulation
      Make read_packet fail is the v4l2 driver returns an unexpected frame size
      Properly set EXTRALIBS
      Do not use a fake libavcodec/swscale.h, but always use the real one
      AVPicture has 4 planes, not 3
      Add support for conversions from the rgb565 and rgb555 formats
      Synch with ffmpeg's r6938 "allow spaces in source and build directory names..."
      Clarify that some of the non-SIMD code is now LGPLed.
      Allow to compile swscale's non-SIMD code under the LGPL license.
      Fix compilation of non-MMX code with gcc 2.95
      libswscale can now be compiled under LGPL too
      Remove useless include (not needed now that libswscale uses libavutil)
      libswscale is not LGPLed, sorry
      Opss... My last commit was wrong; sorry!
      Clarify that yuv2rgb_init.c cannot be used under LGPL
      Remove some void pointers arithmetic
      Add some explicit casts to avoid some warnings
      Add some other casts, to remove void pointers arithmetic
      Add some explicit casts to avoid "assignment from incompatible pointer type"
      Mark some variables as possibly unused to avoid warnings
      change all the occurrences of "FFMIN(FFMAX())" to clip_uint8() or clip()
      replace some av_malloc();memset(...,0,...) sequences with av_mallocz()
      __attribute__((unused)) works with gcc 2.95 too
      Add "support for both orderings of the slices (top->down / bottom->up)"
      deprecate sws_scale_ordered(), as it now is a duplicate of sws_scale()
      Increase the av_log() levels, and add some documentation for them
      Pass a context to av_log(), when possible
      Directly access av_log_level instead of calling av_log_{set, get}_level()
      Pass a proper context to av_log()
      Pass the v4l2 buffer into the AVPacket, instead of allocating a new buffer
      Do not jump to "fail" without having a proper sws context in img_resample_ctx
      Fix the generation of RTSP reply headers (the "DEF()" macro is not used
      Fix the poll() loop (ret == 0 means that a timeout expired, so it must be
      Remove unused callbacks from ffserver
      Fix an use-after-free bug
      Fix ffserver crash when closing the second RTP session
      Fix reading PTSs and DTSs in libavformat, by reverting r8465.
      Reindent code after last commit
      Do not try to set a video standard unless "-tvstd" is specified
      Remove unused file
      Add support for non-blocking input formats
      Support for the AVFMT_FLAG_NONBLOCK flag (non-blocking input) in v4l2.c
      Resurrect yuv2rgb_init.c, to add a proper license (and some comments about
      Remove unused file (this time for real)
      Implement av_strlcatf(): a strlcat which adds a printf style formatted string
      Introduce an SDP generator
      Fix linking when RTP is disabled and libraries are dynamic
      Change avf_sdp_create() to get a pre-allocated buffer as input, and to
      Use avf_sdp_create() in ffserver
      Call av_find_stream_info() immediately after opening an input file.
      Print an SDP when all outputs are RTP (the SDP can be saved in a text file
      Initialize the "stc" variable before using it in ff_find_start_code()
      Move the RTP packetization code for MPEG12 video in its own file (rtp_mpv.c)
      Remove RTSP special casing in ffplay startup
      Reindent the code after the last commit
      Remove copyright from a function prototype
      Use correct timescale in RTSP seeking
      Remove commented code (payload header extension)
      Correct packetization for MPEG video frames, and correct setting of the
      Correctly set the "P" field in the payload header
      Set the "TR" field in the payload header
      Set the "B" flag in the payload header
      Fill the buffer with 0 before writing an SDP in it
      AAC support in the SDP generator
      Properly set RTP and NTP timestamps in RTCP SR packets
      Fix timestamps in RTP packets (now, MPEG1 video with B frames works correctly)
      Test extradata_size instead of the CODEC_FLAG_GLOBAL_HEADER flag to check if
      Support for AAC streaming over RTP. Fragmentation is not implemented yet
      Support fragmentation for AAC frames
      Use AVFMT_FLAG_GENPTS for the input file (the RTP output format needs all the
      Fix an ffserver crash when a non-existing rtsp: url is requested
      Fix a warning by removing an useless assignment (buf_ptr should be only
      Add my name to the copyright header, since rtp_mpv now contains
      Add myself as a maintainer for the RTP muxer and for the SDP generator
      Allow to set the maximum number of frames per RTP packet (and add support for
      Remove prototype for non-existent function
      Explicitly include rtp.h (needed for rtp_get_payload_type())
      Remove the inclusion of rtsp.h and rtp.h from avformat.h, and
      Remove redundant prototype
      If local port n is not available, try n + 2 instead of continuing to bind
      Mark the source buffer as "const"
      Reindent after last commit
      Use a symbolic name for the payload size of an RTCP Sender Report packet
      Do not send too many RTCP packets (according to RFC 3550, the minimum
      Remove redundant "if(len)"
      Reindent the code after last commit
      Add support for some more audio formats
      Remove some ifdefs by using the "dest_addr_len" field in both the IPv4-only
      Avoid to duplicate the multicast code between the IPv4-only and
      Give better names to multicast functions (they are not IPv6-only)
      Remove a now unused field
      Use recv() instead of recvfrom() (removes some other differences between
      Simplify set_remote_url(), and remove some code duplication
      Remove some useless assignments (the UDPContext is allocated with av_mallocz())
      Use the same code to set local_port in the IPv4-only case and in the
      Remove the useless variable "my_addr1"
      Remove some more duplicated code
      Remove the "AVRtpPayloadTypes[i].pt == i" assumption from RTP and RTSP
      Add MPEG2 support to the RTP muxer
      Remove perror() usage from libavutil
      Do not use perror() in audio, video, and DV grabbers
      Remove perror() usage from udp.c
      Convert perror() in av_log() in bktr.c.
      Generate a linking error if perror() is used in libav*
      Include os_support.h only when needed
      Introduce libavdevice
      Fix the creation of alldevices.c (create it without any rename).
      Link libavdevice against libavutil, libavcodec, and libavformat
      Suppress the "redirector hack" from libavformat/utils.c:av_open_input_stream(),
      Fix the RTP muxer dependencies
      Rationalize some dependencies between rtsp and rtp
      Split the RTP demuxing functions out of rtp.c, to simplify the RTP muxer's dependencies
      Split the RTP muxer out of rtp.c, to simplify the RTSP demuxer's dependencies
      Update maintainership information after the rtp.c split
      Pass a proper context to av_log()
      Do not install rtp.h (which should have never been part of the public API)
      Make avc_find_startcode non-static, so that it can be used from other
      Add minimal support for H.264 video in the SDP generator
      Add support for H.264 video in the RTP muxer
      Support out-of-band parameter sets in SDP for H.264 video
      Include poll.h instead of sys/poll.h
      Fix
      Fix
      Fix computation of the "NTP time" field in RTCP SR packets, and do not
      Add some information about the stream bitrate, if available
      Support mp3 audio in the RTP muxer
      Remove useless entries from AVRtpPayloadTypes
      Use the correct size for the enc_name field (removing the arbitrary "50" size)
      Add a comment about missing entries
      Remove the "multicast=" tag from UDP and RTP URLs
      Fix receiving from SDP with unicast destinations
      Fix type of dest_addr_len to respect return value of udp_set_url.
      Fix neverending loop which might happen when sending high-bitrate MPEG2
      Do not set sample_rate = 90000 for mp2 and mp3 audio over RTP
      Remove improper usage of AV_TIME_BASE_Q
      Do not free the priv_data field of AVStream on close (it is already
      Fix A/V synch for RTP streams that do not contain MPEG1 or 2
      Reindent after last commit
      Remove outdated comment
      Fix MPEG video packetization for RTP
      Do not rescale AV_NOPTS_VALUE in output_example.c
      Reindent after last commit
      Do not set timestamp information for a non existing AVStream
      Fix udp.c compilation with recent glibc (>= 2.8)
      Bump Major version, this commit is almost just renaming bits_per_sample to
      Remove unused types
      Remove useless typedef
      Rename rtp_payload_data_t to avoid clashes with the POSIX namespace
      Remove the frame rate from struct video_data, and allow using a variable
      Do not assume that if pkt->destruct != av_destruct_packet then it is not
      Do not reallocate AVPacket's data when muxing a packet
      Reindent after last commit
      Merge rtp_internal.h in rtp.h, and remove rtp_internal.h
      If AVPacket->data == NULL, the packet does not contain any buffer to
      Introduce a new num_frames field in RTPDemuxContext so that rtp_aac.c
      Split rtp.h in rtp.h, rtpdec.h, and rtpenc.h
      Merge rtp_aac.h and rtp_mpv.h in rtpenc.h
      Do not use RTP_MAX_PACKET_LENGTH in the RTP muxer
      Move RTP_MAX_PACKET_LENGTH to rtpdec.h: it is not used in the muxer anymore
      Cast a __u64 variable to uint64_t so that it can be printed without warnings
      Rename rtp_get_payload_type() to ff_rtp_get_payload_type(), as it is not
      Document ff_rtp_get_payload_type()
      Move to rtp.h the prototypes of all the functions defined in rtp.c
      Remame rtp_get_codec_info() to ff_rtp_get_codec_info(), as it is not
      Document ff_rtp_get_codec_info()
      Document ff_rtp_enc_name()
      Fix typo in a comment
      Document ff_rtp_codec_id()

Luca Barbato (159):
      postproc/yuv2rgb_altivec.c compile fix
      "put_s should store signed values. Spotted on #mplayerdev by someone I
      add libdir support to configure patch by (Luca Barbato )lu_zero gentoo org)
      SONAME support for shared libs
      Fix AltiVec H.264 decoding on FSF gcc.
      altivec support for snow
      altivec support for snow
      Cast cruft in order to support Apple gcc
      Proper typecast for vector bool
      hadamard8_diff* enabled on linux/ppc
      typo in the gcc handmade regalloc trigger
      40% faster decode
      13% faster inner_add_yblock
      Macosx fix
      put_pixels16_l2_altivec and avg_pixels16_l2_altivec
      oprofile reference for Linux/PPC
      Cosmetics: should not hurt performance, scream if are
      Make gcc-3 happy again, thanks to Olivier Castan <castan.o at free.fr> for pointing the issue
      fix 24bit flac support, revised from Thibaut Mattern <thibaut.mattern at gmail.com>
      minor cleanup
      Make debug work early on ffplay
      cleanup patch from Diego Pettenò <flameeyes at gentoo.org>
      make distclean clean everything
      big endian fix
      fix endianess build in a better way
      Make ffmpeg better support pic - from Diego Pettenò <flameeyes at gentoo.org>
      Cosmetics
      avoid possible segfault situations
      Cleanup
      vorbis_inverse_coupling_altivec
      avoid alignment hacks, luckly gcc does the right thing on arches different from x86
      Minor fix
      Almost cosmetic changes in dsputil_init_ppc and vorbis_inverse_coupling_altivec:
      standalone snow dsputil init
      Fix a stupid typo and another error, thanks to Emanuele Giaquinta <exg at gentoo.org> for pointing out the issue and the patch
      Yet another typo
      Cosmetics: 2->4 spaces and some braces
      Fix x86 SIMD asm and pic, patch from Martin von Gagern <Martin.vGagern at gmx.net>
      keep in sync with dsputil, makes --disable-decoder=vorbis build
      Clean up:make dsputil subfile names consistent
      altivec float optimizations
      10l, thanks to Emanuele Giaquinta <exg at gentoo.org> for testing and finding the issue
      add --disable-ipv6 option to configure
      Fix float_to_int16, unaligned case, broken by the previous commit
      Revert previous commit
      Proper fix for the corner case that would have been corrected before, praise&blame to me and exg in equal shares
      Align the input buffer in ffplay, introduce a public macro for aligned declarations
      Cleanup, from Emanuele Giaquinta <exg at gentoo.org>
      New single instruction math operation header
      Use mathops
      move memory functions from avcodec to avutil
      make postproc use avutil
      Fix postproc build, thanks to Pavlov Konstantin <thresh at cpms.ru>
      fix blackfin support, thaks to vapier for pointing out
      fix a warning
      fix the dcbz check and move it in configure
      Non Altivec optimizations already present at the top
      add_bytes passes tests
      removing ALTIVEC_USE_REFERENCE_C_CODE, since has no use anymore
      ppc generic prefetch
      Make ffmpeg work with x264 r592
      use --as-needed just on linux, since *bsd way is broken... (patch from Diego Pettenò <flameeyes at gentoo.org>)
      Make sure --as-needed and the other flags got checked as the last thing, workarounds linking checks issues due broken system libraries
      Simplify
      typo
      Unused variable removed
      fix a gcc warning, avoid an unnecessary operation
      Cosmetics
      Avoid branches in the loop and solve a gcc warning
      Constantize AVOption, solve few warnings, patch from flameeyes at gentoo.org aka "the other Diego"
      10l, there is a corner case afterall...
      Fix warnings and simplify/reorder the code
      Simplify and avoid a warning (should be faster on Cell and certain G4 revisions)
      av_logify
      Make example and test build and link correctly
      stray fflush spotted by Aurel
      Missing prototype added
      Even powerpc/power needs PIC for shared objects today
      typo fixed
      get_packetheader() forgot to read the header_checksum in big packets
      Reverting stray commit part I
      Reverting stray commit part II, r8156 had the base64 export patch mixed with the nutdec patch
      Many steps to avutilized this mess: DECLARE_ALIGNED
      Typo
      expose av_base64_decode and av_base64_encode
      Typos in comments
      ssd_int8_vs_int16_altivec, not completely benchmarkedwith svq1
      Add support for Cell (requires gcc-4.2 or patches)
      h264_idct_add_altivec, based on an old patch from Mauricio Alvarez <alvarezATac.upc.edu>, polished by David Conrad <umovimusATgmail.com>
      bump micro version: new string functions
      wish priority description and ffmpeg-issues vs ffmpeg_issues
      Typos removed
      Introducing Topics
      rtsp multicast fix by Thijs Vermeir $1$2 at gmail.com
      adding myself as rtsp.c maintainer
      RTP clarifications
      Editorial cleanup, thanks Diego =)
      fix the packetization mode parsing
      PowerPC/CELL/Altivec pointers
      Stray commit reverted
      snow altivec is broken
      Backport fix for mpegvideo rtp, fixes Issue119, original fix from Dario Gallucci <dariodotgallucciatpolito.it> in feng
      Cosmetics
      Sanitize altivec code so it can be built with runtime check properly
      Make av_read_frame with rtsp client return EINTR on interrupt
      Make av_read_frame with rtsp client return EINTR on interrupt
      Provide sha1 to outside applications
      kill a warning, av_sha1_update accepts uint8_t * not uint64_t *
      Make strict altivec parsers happy (gcc-4.3 and others)
      Cosmetics
      Factorize common code (almost cosmetic)
      Reindent
      Avoid a vec_add, directly start with sum
      Partially address issue299, no performance change apparently
      Reindent
      Add C/B == 0 cases, 2% slower on CELL but should address Issue299 eventually
      10l do not load after the buffer...
      Minor fix
      Reindent
      Real RTSP support, from Ronald S. Bultje rsbultje gmail - part 1 Comment
      Real RTSP support, from Ronald S. Bultje rsbultje gmail - part 2 x-pn-tng support
      Real RTSP support, from Ronald S. Bultje rsbultje gmail - part 3 Reindent
      gcc2 doesn't have bogus mergel, patch from Sigbjørn Skjæret cisc...broadpark___no
      Reorder declarations in vertClassify_altivec
      Remove declarations after statements from vertClassify_altivec
      Remove declarations after statements from doVertLowPass_altivec
      Remove declarations after statements from doVertDefFilter_altivec
      Cosmetics: reindent and whitespaces
      use FF_NETERROR to make winsock happy, patch from prossATxvidDoTorg
      Error message for EPROTONOSUPPORT, patch from prossATxvidDoTorg
      Eliminate movdqu in vp3dsp_sse2, patch from Alexander Strange astrangeAtithinkswDoTcom
      Expose max_packet_size from the rtp protocol
      Cleanup comments and make them doxygen parsable.
      doxy header for udp protocol
      libdirac and libschro common code, patch from Anuradha Suraparaju anuradhaAtrdDoTbbcDoTcoDotuk
      Import Dirac demuxer/muxer from SoC branch; written by Marco Gerards,
      Build sparc arch specific code using the Makefile
      Cosmetics: reorder and align targets
      Split simpleCopy into packedCopy and planarCopy. SoC Patch from Keiji Costantini
      Reindent. SoC Patch from Keiji Costantini
      Remove ASSERT() macro. SoC Patch from Keiji Costantini
      Split AVOption/AVClass in a separate file. SoC Patch from Keiji Costantini
      100L, avoption splited code added, I should double check with svn status...
      Introduce float_to_int16_one_altivec
      Introduce float_to_int16_interleave_altivec, tested with vorbis
      Remove unused variables
      Missing static in float_to_int16_altivec declaration
      Make ToY and ToUV family of function consistent part I
      Factorize unit32_t* casts for palette pointer
      Split mono2Y in monowhite and monoblack
      Uniform *ToY and *ToUV function signatures
      gcc-apple specific fallback not necessary anymore (btw no apple hardware is less than a Intel core, thus it won't come there w/out disabling all the optimizations)
      Flush SDP description
      Rename type to be consistent
      Unbreak imgresample altivec
      Cleanup types
      Let imgresample altivec code build
      Cleanup _t types in libavcodec/ppc
      Remove stray code

Maarten Daniels (2):
      H.261 decoder improvements
      H261 fixing and cleaning:

Magnus Damm (2):
      Simplified and fixed altivec/altivec.h detection under Linux and Darwin (and others). Patch by Magnus Damm <damm at opensource.se>
      altivec swscale under Linux patch by (Magnus Damm <damm at opensource dot se>)

Maksym Veremeyenko (1):
      * clarifying the AVInputFormat::read_seek return value

Marc Hoffman (57):
      add bfin cpu specific build options, patch by Marc Hoffman % mmh P pleasantst P com %
      Fix compilation on Mac OS X, patch by Marc Hoffman, mmh pleasantst com.
      Handle proper compilation on Mac OS X.
      Replace emms() with emms_c(), patch by Marc Hoffman, mmh pleasantst com.
      Remove illegal call to exit().
      platform-independent restructuring and code simplification
      bfin dsputils, basic pixel operations sads, diffs, motion compensation
      Remove stray object from list of Blackfin assembler objects.
      Blackfin is a strictly aligned processor.
      Faster 32 bit byteswaping code for Blackfin.
      Remove large automatics from stack.
      Blackfin optimized YUV420 to RGB CSC Color Space Converters.
      Blackfin DSP utilities: add DEFUN_END
      Adding Myself to the Maintainers list for the Blackfin architecture
      Blackfin optimized uyvytoyv12 routine. the implementation includes chroma averaging.
      convert if then else video palette to pix_fmt construct to simpiler table based lookup.
      extending video4linux support to handle the additional formats UYVY, YUYV and RGB565
      correcting spelling error, yuv2rgb888_line was renamed to yuv2rgb24_line
      Blackfin IDCT has no special permutation it uses normal order
      Blackfin dct_quantize_bfin routine
      integrating MPV_common_init_bfin into the build system and mpegvideo.c
      Blackfin - read_time primitive
      removing underscores from block profiling renaming to xx_
      correct chroma skewing bug, caused by incorrect stride adjustments
      re pipeline loop, to eliminate extra chroma reads
      document pointer offset by -8
      adding Blackfin swscale_bfin infrastructure, and the first swscale routine uyvy to yv12 video resampler
      integrating ff_bfin_get_unscaled_swscale into the system
      renaming L1CODE to attribute_l1_text, which is defined in dsputil_bfin.h
      removing dependency on bfin_sram.h, which is not used right now
      cosmetic changing hardwired 20 to SUBBAND_SIZE
      moving automatic allocation of joint_decode/decode_buffer
      separating saturation codes so that we can support other data formats
      separate the actual math for recovering stereo from a signal channel, this is to allow fixpoint implementation
      fixpoint: separate windowing arithmetic imlt_window_float
      fixpoint: lowlevel functional abstraction for all buffer arithmetics
      fixpoint: move cplscales to context structure and provide hook for data format conversion
      float_t was a typo that compiled on macosx
      cosmetics: adding some white space to align the arguments of a couple of functions
      fixpoint: renaming all lowlevel arithmetic routines to xxx_float
      adding blackfin optimized yuyvtoyv12 converter
      removing reference to bfin_sram.h its not used anylonger
      integrating blackfin optimized yuyvtoyv12 converter
      adding myself to the CREDITS
      making reference models have the same API
      --enable-fixedpoint, for configuring fixedpoint optimizations, this is just the config aspect
      correct order of fixedpoint in CONFIG list
      special handling for vp3 idct, use the c reference
      make clean bfin
      Blackfin 32bit multiplication primitives MULH MUL16
      Hooking bfin/mathops.h
      renaming ARCH_BFIN to HAVE_BFIN
      renaming ARCH_BFIN to HAVE_BFIN
      using isnan instead of d==d, to signal correctly parsed option
      properly tell the compiler that A0 and A1 are clobbered
      blackfin vsad primitives
      blackfin optimized vp3 transform and infastructure for idct

Marc Mason (2):
      Check function prototype in check_asm function to avoid compiler warnings.
      Make unsigned int a variable that does not need to be explicitely 32 bits.

Marco Gerards (4):
      THP PCM decoder, used on the Nintendo GameCube.
      cosmetics: indentation fix
      Fix an underflow/overflow that was causing some crackles when playing
      Make the Golomb decoder work for Dirac

Marco Manfredini (2):
      Fix CHECK_BIDIR macro so it works with Intel's Compiler
      Fix building with --disable-opts but MMX enabled.

Marcus Engene (3):
      vhook watermark patch by (Marcus Engene: ffmpeg, engene se)
      FreeBSD support as well as some assorted fixes
      Watermarking with exact colors.

Mark Cox (5):
      simplify, patch by Mark Cox, melbournemark plus ffmpeg minus devel gmail com
      patch so that the deprecated items show up correctly
      Remove redundant assignment of CODEC_ID_BMP.
      Add description for feature_request/open/approved.
      Fix unused variable warnings.

Mark Hills (5):
      oggvorbis support patch by (Mark Hills <mark at pogo dot org dot uk>)
      * Ogg/Vorbis patch by  Mark Hills
      fixes crash patch by (Mark Hills <mark at pogo dot org dot uk>)
      ogg deallocate patch by (Mark Hills <mark at pogo dot org dot uk>)
      ffserver deallocate ctx->streams on closing patch by (Mark Hills <mark at pogo dot org dot uk>)

Mark Weaver (1):
      WMA MAX_CODED_SUPERFRAME_SIZE too small patch by (Mark Weaver: mark-clist, npsl co uk)

Marko Kreen (1):
      adding ff prefix to avoid global name conficts with xvid (patch by Marko Kreen <marko at l-t.ee>)

Martin (1):
      allow amr bitrate to be changed "on the fly" through avctx->bitrate.

Martin Boehme (2):
      Fix for overflow issue in mpegvideo.c patch by (Martin Boehme: boehme, inb uni-luebeck de)
      Since there have been quite a few questions recently on how to use the

Martin Drab (2):
      Exporting some more options patch by (Martin Drab <drab kepler.fjfi.cvut cz>)
      optimization and gcc 4.0 bug workaround patch by (Martin Drab >drab kepler.fjfi.cvut cz<)

Martin Lindhe (1):
      Move 2 defines from mpegaudiodec.c to mpegaudiodata.h to allow compilation

Martin Olschewski (1):
      sun solaris compilation bugfix, patch by (Martin Olschewski <olschewski at zpr dot uni-koeln dot de>)

Martin Storsjö (2):
      fix bug when __GNUC__ isn't defined and when compiling for a non-x86 architecture, patch by Martin Storsjö, martin at martin st
      Fix memory leak in libgsm wrapper.

Martin Zlomek (2):
      Fix deblocking for PAFF content
      Fix mix-up of values in r10690 (deblocking for PAFF).

Mathieu Malaterre (5):
      Fix a potential out of array writing.
      Wrong constant used in default threshold value calculation for
      Return an error when ff_jpegls_decode_picture fails.
      cosmetics: s/const static/static const/g
      Use default system include path for an installed OpenJPEG library.

Mathieu Velten (2):
      mlp: Fix compilation under CONFIG_AUDIO_NONSHORT.
      Use H264 MMX chroma functions to accelerate RV40 decoding.

Matthias Fritschi (1):
      writing corrupt files on MinGW patch by (Matthias Fritschi <choi at netlabs dot org>)

Matthias Hopf (1):
      fix some potential security issues, patch by Matthias Hopf, mat at mshopf dot de

Matthias Schneider (1):
      Remove cross_prefix from ln command.

Matthieu Castet (16):
      attribute_used patch by (matthieu castet <castet.matthieu at free dot fr>)
      detect sse on athlon-xp patch by (matthieu castet <castet >.< matthieu >at< free >.< fr>)
      10l patch by (matthieu castet <castet.matthieu free fr>)
      extract from the subid the options
      fix theora header parsing
      Make rle_code int everywhere instead of signed char.
      Correctly skip unknown matrices (patch by Matthieu Castet) and disable comment reading
      fix playback of theora.ogg on mphq patch by (matthieu castet %% castet dat matthieu ot free dat fr %%)
      pld instruction are used in libavcodec/armv4l/dsputil_arm_s.S which can be used
      check alac EOF marker
      When compiling with -fmudflap configure fill EXTERN_PREFIX with garbage
      simplify decoding of uncompressed samples. patch by matthieu castet <castet matthieu free fr>.
      cosmetics: indent after last commit. patch by matthieu castet <castet matthieu free fr>.
      Define mm_support() for ARM.
      tools/trasher: Don't mix srand/rand and srandom/random interface.
      alac : check readsamplesize to ensure get_bits() doesn't

Matthijs Douze (1):
      Print error message instead of crashing when combining -vcodec copy and -vhook.

Matti Hamalainen (1):
      Off-by-one bug in libavcodec/idcinvideo.c, caused

Max Krasnyansky (5):
      mpeg4 raw output support patch by ("Maksim (Max) Krasnyanskiy" <maxk at qualcomm dot com>)
      img_convert() (YUV to YUV) patch by (Max Krasnyansky <maxk at qualcomm dot com>)
      dv patch by Max Krasnyansky (maxk at qualcomm dot com)
      variable UDP packet size patch by Max Krasnyansky
      Frame rate emulation patch by (Max Krasnyansky <maxk at qualcomm dot com>)

Maxim Poliakovski (1):
      Fix decoding of 01-Untitled(1).oma, patch by Maxim Poliakovski

Mean (1):
      prevent going out of the buffer if the nal size does not fit in the buffer.

Michael Benjamin (3):
      Blackfin processor support
      Add support for 8-bit grayscale to V4L grabber
      ADI Blackfin optimizations

Michael Donaghy (1):
      fix 8- and 24-bit FLAC decoding

Michael Kostylev (8):
      djgpp port, first part
      Check for the presence of sys/select.h and conditionally #include it.
      Check for the presence of llrint(), lrint(), round() and roundf()
      Add system-specific network link flags to inet_aton test.  This should fix
      fix extern prefix detection when cross compiling.
      Don't prefix yasm with cross prefix.
      Use coff as objformat for dos.
      Add truncf() replacement function.

Michael McConnell (1):
      Fix to loss of sync in ffplay when paused patch by (Michael McConnell /// soruk put at here eridani and here a dot co.uk)

Michael Niedermayer (4982):
      bugfixes: last 3 lines not brightness/contrast corrected
      final changes to convert to C
      fixed a rounding bug thing in the X1 Filter
      minor QP scaling & auto-contrast bugfix
      fixed a sig4 bug an non mmx2 cpus (in case of more sig4 errors please send me a "disassemble $eip-16 $eip+16" from gdb)
      rewrote the horizontal lowpass filter to fix a bug which caused a blocky look
      minor cleanups
      quick fix for movies with a height which is not a multiple of 8
      MMX & MMX2 optimizations (MMX2 is buggy and commented out)
      fixed out of regs "bug"
      24bpp support (untested)
      fixed the height%8!=0 bug
      increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
      fixed a bug which caused horizontal lines
      vertical lines bugfix
      horizontal lines bugfix
      horizontal scaling bugs fixed, should be mostly bugfree now
      fixed a bug in the tmp buffer
      added compiletime option to turn width%8==0 on
      first line bug fixed (happend if the image was scaled down or 1:1)
      more logic behavior if the altenative deblock filters are used (turning a alt filter on without turning the deblock filter on uses the alt filter instead of using no filter now)
      fixed alignment (static variables where sometimes not 8-byte aligned)
      BGR24 bugfix
      auto brightness/ contrast bugfix
      negative black bugfix
      brightness / contrast fix/copy optimizations +2% speedup
      optimizations (+2% speedup)
      and another +2% speedup
      more speed
      more speed
      much better horizontal filters (transpose & use the vertical ones) :)
      fixed difference with -vo md5 between doVertDefFilter() C and MMX / MMX2 versions
      downscale
      dering in mmx2
      just a few "not implemented yet" removed
      3dnow dering
      dering in c
      faster dering
      yuv2rgb bugfix
      yv12 to yv12 scaler
      3dnow prefetch & sfence
      tell the c compiler that the memory changed
      patch from Martin Decky <deckm1am at ss1000.ms.mff.cuni.cz> applied and unnecassery "memory" removed
      cant compile on non x86 bugfix
      fixed a warning
      bugfixes
      interpolate chrominance for every second line in the 1:1 vertical scale function
      minor speedup
      c optimizations (array is faster than pointer) (16bpp variants tested and 2% faster)
      c optimizations
      c optimizations
      15/16bit in C speedup
      c speedup
      c speedup (x1, rk1 filters)
      deinterlace bugfix
      right green line bugfix for width not %8==0 (untested -vo vesa doesnt work)
      faster bgr15/16
      fixed warnings
      green line fix for dstw%8!=0
      bgr32to16, bgr32to15 (needed for palette stuff)
      palette to bgr24
      faster 15to16 bit rgb (the mmx routine is limited by memory speed so there is no difference ): but the c routine is faster
      yv12 <-> yuy2 in C
      yv12toyuy2 in MMX
      yuy2toyv12 in MMX
      fixed yv12toyuy2
      yuy2toyv12 fixed and speedup
      unsigned stuff
      only 6 registers used
      faster bgr24 output
      common code moved to a #define
      15 bpp support
      16bpp is 10% faster
      vars are aligned by 8 now (even faster, yes they where not aligned correctly)
      mmx2 bgr24 stuff from swscale (slightly faster)
      minor changes (fixed some warnings, added attribute aligned(8) stuff)
      better dithering
      15/16 bpp dithering
      gcc does optimize writes to non volatile variables away if it didnt know that they were read in between
      rgb32tobgr32 / bgr32torgb32
      bugfix
      .align 16
      .balign or we´ll align by 64kb on some architectures
      yuy2toyv12 bugfix
      QP_store==null bugfix and no opendivx bugfix
      uyvytoyv12 in MMX (untested)
      temporal noise reducer in C (-pp 0x100000)
      temp denoiser changes: (a-b)^2 instead of |a-b| and MMX2/3DNOW version
      temp denoiser:
      faster mmx2 / 3dnow deblocking filter
      cleanup
      oops, too much cleanup ;)
      1% speedup
      faster dering
      use fastmemcpy for chrominance if no chrominance filtering is done
      runtime cpu detection
      runtime cpu detection
      runtime cpu detection
      oops
      runtime cpu detection
      runtime cpu detect optional (compiles faster)
      runtime cpu detection optional (compiles faster)
      compiletime pp-mode support (luminance = chrominance filters though) 1-2% faster with -benchmark -vo null -nosound
      faster brightness correcture in MMX2
      swscaler cleanup
      sliced scaleing bugfix
      horizontal up/downscale linear & cubic
      typo
      vertical cubic/linear scaling
      dithering info wasnt displayed
      print more info if -v
      rgb24toyv12 bugfix
      minor bugfixes (noone noticed them)
      new -pp options stuff (-npp ...)
      101
      output shifted by a few pixels on extreem scalings bugfix
      customizeable deblocking (2 thresholds)
      minor brightness/contrast bugfix / moved some global vars into ppMode
      only print last line warning if verbose is on
      -npp fq:<quantizer>
      -npp help
      fixed -npp help ... i think i need more sleep ;)
      fix slices when code=0x18, patch by Michael Niedermayer <michael at mplayer.dev.hu>
      print cpu flags
      fix different UMV handling for mpeg4 vs. h263 - patch by Michael Niedermayer <michaelni at gmx.at>
      fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni at gmx.at>
      AC table reset (memset) optimization - patch by Michael Niedermayer <michaelni at gmx.at>
      another bitstream reader code (faster on intel cpus) - patch by Michael Niedermayer <michaelni at gmx.at>
      aligned bitstream support (optional) - patch by ichael Niedermayer <michaelni at gmx.at>
      s->c_dc_scale was 7 if s->qscale==2 but should be 8 (the bug is visible in deep red areas in high bitrate clips) - patch by Michael Niedermayer <michaelni at gmx.at>
      different fix, s->mbintra_table used only if h263_pred set. - patch by Michael Niedermayer <michaelni at gmx.at>
      (commited by michael / arpi was crazy enough to give me his password)
      (commit by michael)
      (commit by michael)
      dont print -npp parser debug info if only -v
      (commit by michael)
      (commit by michael)
      (commit by michael)
      (commit by michael)
      (commit by michael)
      (commit by michael)
      (commit by michael)
      removed global vars so that multiple swscalers can be used
      101
      general convolution filtering of the source picture
      a few filters (should be removed/merged when arpis videofilter stuff is finished)
      copyright(swscaler) = GPL
      15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
      minor bugfix in the -v output
      (commit by michael)
      (commit by michael)
      (commit by michael)
      (commit by michael)
      minor QP bugfix
      nearest neighbor / sdl emulation ;) scaling (-sws 4)
      area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
      top row bugfix
      YV12 like formats support (I420/IYUV)
      -sws 2 is default now
      YUY2, BGR24, BGR32 input support (no mmx yet)
      printing correct format info
      MMX(2) optimized YUY2 input
      filterPos bugfix (only nearest neighbor scaling was affected)
      printing error messages if something is wrong instead of just return NULL;
      use multiply instead of divides for DC prediction on X86
      swScale internally uses yuv2rgb now if possible
      101 (yuv2rgb with mpeg1,...) ;)
      rgb32 & rgb24 input support
      width or height %16 != 0 bugfix
      fixing green line at right side bug (reported by Nick Kurshev <nickols_k at mail.ru>)
      unscaled converters
      avoid copying input when encoding non intra stuff too
      bgr16 input support
      faster bgr16 input
      bgr15 input support
      unscaled BGR15->BGR16
      fastmemcpy
      workaround gcc 2.95.2 bug
      bgr24toY in MMX
      bgr24toUV in MMX
      rgb24toyv12 in MMX (untested)
      unscaled BGR24 -> YV12
      unscaled memcpy variant bugfix (negative strides)
      alternative bitstream writer (disabled by default, uncomment #define ALT_BISTREAM_WRITER in common.h if u want to try it)
      aligned bitstream writer (1% slower on p3 but perhaps its faster on p4?)
      optimize mpeg1_encode_block()
      optimized encode_dc() (+2% speed on P3 for mpeg1 intra only encodings)
      minor dststride bugfix for i420 output
      optimized the normal bitstream writer, its faster than the alternative one on p3 now ... lets hope its at least not slower on p4 & k7
      block_last_index was too large (in mpeg1 decoding)
      avoid overflow of picturenumber*fps*10000
      unscaled yuv2rgb fix for big-endian systems
      fixing bgr32 output on big-endian systems
      101++
      use assert.h
      fixing msmpeg4 decoding if fps < 16 (i thought it was a indicator for the ext header, its the fps indeed)
      fixing warning
      parsing more of the mpeg4 header & print some "not supported" stuff
      gmc bitstream decoding support (the real motion compensation isnt implemnted yet)
      divx5-gmc support
      q-pel mc fixed
      divx5 draw_edges bugfix
      mpeg4 b-frames :)
      generate even more complete mpeg4 headers
      support decoding (with mplayer) of 3 .mp4 files from mphq
      use ac prediction in mpeg4 encoding (5% smaller intra-blocks/keyframes)
      optimization
      hopefully better bitrate controll
      10l
      scene change detection
      fixing 4MV
      msmpeg4 bugfix (wrong frame displayed if some frames are skipped)
      use the width & height from the mpeg4 header ... in the case that its complete
      trying to fix qscale==0
      fixing bgr15&16 on big_endian
      new motion estimation (epzs) not complete yet but allready pretty good :)
      oops
      -me epzs
      mpeg1 & epzs fix
      fixing epzs & mpeg1 (hopefully now really ...)
      mpeg4 aspect_ratio_info in AVCodecContext (requested by alex)
      more aspect stuff
      fixing MV range in h263
      better inter/intra decission algo (same as xvid)
      dont return -1 if mpeg4 scalability ...
      statistics for 2-pass encoding
      fixing motion estimation with h263p hopefully
      10l (motion_val was uninitilized)
      seems the old intra/inter decission is slightly better with a threshold, than the new one
      resync marker support, needed for some mp4 files
      b-frame decoding bugfix
      fcode_tables where too small, found by Klaas-Pieter Vlieg <vlieg at eurescom.de>
      optimizing mpeg4_encode_dc
      byte interleaving for mga
      10l
      4MV motion estimation (not finished yet)
      mpeg4 4MV encoding
      sizeof(s->block) isnt 64*6*2 anymore bugfix
      force frame type for 2-pass encoding
      minor optimizations / simplifications
      msmpeg4v2 header parser & some dump bits code behind #if 0
      divx5-gmc bug workaround
      fixing 1/0
      overread in the mmx2 horizontal scaler fixed
      fixing MVs in hq mode
      emms was missing, found by juanjo but he didnt commit it?!
      dump_stream support (nicely formated 0s and 1s)
      msmpeg4v2 decoding (no encoding yet)
      b frames & slices bugfix
      fixing msmpeg4v2 bugs (is bugfree now afaik)
      msmpeg4v2 encoding
      wmv1 header parser (unfinished)
      mpeg4 mpeg quantizer support
      fixing black stripe at right if width%8!=0 && stride%8==0 reported by kabi
      fix typo found by Milan Cutka <cutka at szm.sk>
      rgb24->bgr24
      yuv422p -> yuy2 (untested)
      fixed mpeg4 time stuff on encoding
      mpeg2/mpeg4 dequantizer support (c & mmx)
      fixing hq mode with mpeg1 and 2-pass
      much better ME for b frames (a bit slow though)
      fixed ratecontrol & b-frames
      rounding bugfix
      fix b-frame MB skip
      divx 5.01 support
      x1 deblocking filter bugfix
      b-frame mpeg4 fix when seeking
      export delay info
      segfault fix
      avcodec_flush_buffers()
      fixed msmpeg4 infinite loop if buggy stream
      export mbskip_table for direct rendering
      detecting xvid/divx4/opendivx and set low_delay flag
      fixing msmpeg4v3 dc-scale for quantizers 24-31
      set C4 to 16383
      C4=16383 for the c version too and even for some outcommented code
      freeze fix
      correct quantizer bias
      move put_string() to common.{c,h}
      move ABS() to common.h
      shorter user data with formating by juanjo
      MV visualisation support
      3dnow temporal denoiser bugfix by Rémi Guyomarch <rguyom at pobox.com>
      segfault fix for b-frame encoding with height%16!=0
      there are divx5? encoded files without a userdata section but with b-frames :(
      enable simple-idct by default (needed cuz otherwise files encoded with ffmpeg will show stripes with the reference IDCT during decoding)
      fixing end overwrite bugs (some at least)
      print frame decoding time support (x86 only)
      new hopefully faster MC
      hopefully faster mmx2&3dnow MC
      shared lib support (req by kabi) ...
      qmin==1 bugfix
      fixing alignment for alignment picky cpus
      workaround dc_scale bug in old ffmpeg msmpeg4v3 encoder (set workaround_bugs=1 for this)
      using 1 operand less and slightly faster put_pixels_y2
      fixing stackoverflow
      bugfix
      sqrt
      fixing get_bits_count at the end
      error concealment / error resilience
      error concealment / error resilience
      dont double check vectors
      single coefficient elimination
      msmpeg4v1 decoding
      data partitioning
      b_quant_offset (unfinished, should be 0 for now)
      single coefficient elimination
      divx5 bug workaround
      typo (found by Bohdan Horst <nexus at hoth.amu.edu.pl>)
      100l divx501+ seeking bugfix
      run1 can be <0 bugfix (found by fabrice)
      optimizing MPV_decode_mb
      grayscale only decoding
      passing avctx->flags so that grayscale only decoding works with mpeg1/2 too
      grayscale only encoding
      8bpp support (no dithering yet, use -vop noise for now)
      ordered dither for 8bpp
      8bpp bugfixes (brightness, contrast ...)
      1bpp support
      10l
      dc scale simplification/optimization
      wmv1 support
      wmv1 support
      wmv1
      fixing wmv1 bugs, it was foolish to belive that m$ would use the same dc prediction as mpeg4 just cuz they changed it a bit from msmpeg4v3
      cleanup & grayscale support
      more mpeg4 fourcc's
      selftest support
      cleanup
      yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
      100l (non mmx versions didnt compile)
      10l
      4mv bugfix
      killing globals
      init dc_scale tables to defaults (fixes mjpeg sig11)
      fixing scaling/contrast for 1/4/8 bpp
      use unified yuv2rgb init
      -sws 6 (luma bicubic & chroma bilinear)
      yvu9 -> yv12 unscaled converter with linear chroma scaling
      colorspace converter tests (memory corruption tests at different width/src/dst)
      fixing memory overwrite bugs again ...
      merging changes from mplayerxp (rgb2rgb*.{c,h} only)
      test new converter stuff too
      fixing memory overwrite bugs in the new converters
      fixing rgb24tobgr32
      last line messed up in Y800 bugfix
      using the new rgb2rgb converters (the LQ ones are only used if -sws 0 or -sws 4)
      more scaling algos
      wmv1 slice_height != mb_height support
      new bitstream reader API (old get_bits() based one is emulated and will still be supported in the future cuz its simpler)
      fixing sinc filter (seems the problem was caused by rounding in pmulhw -> solution use shorter filter, its long and slow enough anyway)
      gcc3 bug workaround
      c std doesnt like negative shifts -> use asm
      return the number of consumed bytes instead of 0 or buf_size
      -O3
      get_vlc() optimization
      get_vlc() optimizations
      returning the number of consumed bytes (2nd try)
      dont trash table in GET_VLC
      rl vlc decoding optimizations
      return buf_size again, i hope kabi is finally happy now ;)
      optimization (get_vlc -> get_vlc2)
      optimization (get_vlc() -> get_vlc2())
      get_vlc -> get_vlc2
      rl decoding optimization
      faster get_bit_cache
      rl vlc table optimization (not yet used)
      direct rendering method 1 support
      4mv & qpel edge emu
      uvlinesize
      direct rendering for SVQ1
      dr1
      dr1
      grayscale only decoding
      hurry up support
      exporting mbskip_table after it has been allocated
      fixes fire.avi & subtitles.avi
      low_delay / has_b_frames fix
      interlaced mpeg2 fix ... replacing linesize>>1 by uvlinesize brainlessly wasnt a good idea
      fixing gmc with emu edges
      larger edge emu buffer (might have been too small)
      hmm ABS(dc-diff) > 512 fix
      fixing NULL reference frame with dr1
      use edge emu for encoding of width or height %16!=0 files
      BeOS patches by "shatty" <shatty at myrealbox.com>
      pix_abs mmx2 isnt bit-exact ...
      minor optimizations
      fixing bitrate vs. kbitrate in header
      -bug comand line option to workaround encoder bugs which cannot be detected automatically
      different edge positions fixed with edge emu / dr1
      dump bits per frame / qp / frame complexity support
      better big/little endian test
      b frame direct mode bugfix (bug found by CM <chenm001 at 163 dot com>)
      max_b_frame=4 bugfix
      planar yuv 444 422 411 support
      10l found by Rémi Guyomarch <rguyom at pobox dot com>
      fixing normal mmx idct
      fixing mpeg4 mpeg quantizers
      last_picture should be never == NULL (it was with dr1) this might fix a segfault with error concealment
      making some error checks optional
      supporting rare overflow mess even with error_resilience>=0
      fixing slices which start at mb_x>0
      hopefully fixing unreprodceable segfault when overreading the end if error_resilience==-1 (it didnt overread here though)
      fixing level overflow check for qp=1
      4mv + bframe decoding bugfix
      mpeg4 mpeg quantizer encoding
      1000l
      faster vlc table selection
      skip blocks with small chroma dc too (if the user wants it) needed to avoid IDCT(input coeffs !=0) == 0 problems which cause catastrophic error accumulation at qp=1
      same dc skip behavior for chroma & luma elimination, its confusing otherwise imho
      dont trash mm0, its still used in the next iteration of the loop (fixes strange horizontal lines on some blocks in b frames on mmx only cpus)
      regression test checksum update (seems the files where even in different directories ...)
      darwin patches by Heliodoro Tammaro <helio at interactives dot org>
      better scene change detection
      faster ff_sqrt()
      error concealment needs the mbintra_table so it should allways be allocated
      new ratecontrol code
      new ratecontrol code
      fixing qmin==qmax && qsquish==1
      forgot the const bits in 2pass curve matching (patch (with rounding removed) by Rémi Guyomarch <rguyom at pobox dot com>)
      fixing some 64bit bugs
      checksum update due to LIBAVCODEC_BUILD 4619->4620
      mpeg4 interlaced decoding support (not completly implemented/tested due to lack of samples)
      mmx & non-mmx regerssion tests
      fixes rv10 decoding crash
      working around undefined behavior of get-vlc if the vlc s invalid
      use extradata for h263_long_vector mode detection
      base64 encoded patches posted to the devel list are prefered
      slow but accurate integer dct from IJG (should be ok with the LGPL as the old DCT is the fast integer DCT from IJG)
      checksum update cuz of build change 4620->4621
      undefined M_PI / M_E fix
      using only the fast int dct for regression tests
      not used anymore
      bitexact hack for the simple mmx idct
      checksums for the bitexact simple idct
      croping patch by (talus25 at speakeasy dot net) with fixes from atmos & me
      dct-test update
      dct cleanup
      higher accuracy
      ar -s -> ranlib
      fixing encoding at high QPs (was broken after the dct accuracy increase due to too small inverse table)
      10l (compilation if defined ARCH_ALPHA)
      fixing segfault
      do regression tests for rv10 too
      fixing tests
      apiexample needs EXTRALIBS
      fixing rv10 encoding (ffmpeg can at least decode its own rv10 files now)
      support decoding of the last mpeg "packet" even if no startcode is immedeatly afterwards (fixes bugs with mpeg in avi)
      regression tests for msmpeg4v2 & wmv1
      optimizing mpeg4_encode_block(), generates allso slightly shorter bitstream as some codes can be represented as esc1 and esc2 and esc2 is shorter for a few of them
      10l (forgot checksums after mpeg4_encode_block() optimizations which generates slightly shorter bitstream)
      fixing y422p output
      reducing sizeof MpegEncContext to avoid stack overflow on crap M$ windo$
      put/avg_pixels16
      fixing raw yuv input
      set pict_type & key_frame on decoding too
      export pts from stream if available
      m4v input support
      statistics for forw & back p-MBs instead of just one counter for both
      mpegvideo_probe cant distinguish mpeg4 and mpeg1 raws :( so m4v needs "-f m4v" now
      uninitialized var ...
      regression checksum change cuz of LIBAVCODEC_BUILD change
      avoid nonsense frame-skip messages
      vbv parameter parsing (more skiping actually ;) )
      parsing resync header extension support
      error resilience command line option
      fixing interlaced mpeg4 decoding
      fixing 1/0 if the headers are incomplete (like what M$ encoder generates)
      interlaced mpeg4 b frame decoding
      interlaced b frames cleanup
      10l (field select readed at the wrong place ...)
      interlaced qpel mpeg4 fixed
      fixing qpel 4mv
      interlaced b frames fix
      alignment info
      dont store version for bit-exact tests
      fixing memalign
      fixing rc of non-intra-only streams which only contain i frames
      fixing
      better min/max rate handling
      10l (forgot to commit)
      avcodec_alloc_context()
      fixing *pixels_tab stuff (hopefully, as its untested ...)
      documenting most stuff in AVCodecContext
      fixing apiexample
      fixing i_quant_factor, this should finally fix the bitrate bug with ffserver hopefully
      pix_sum16_mmx()
      regression test for ratecontrol & adv mpeg4 stuff
      adaptive quantization (lumi/temporal & spatial complexity masking)
      fixing variance scaling for b frames (messed adaptive quants up)
      fixing ac prediction encoding with adaptive quantization
      darkness masking (lumi masking does only bright stuff now)
      adaptive quantization for mpeg1
      adaptive quantization for h263
      mpeg4 interlaced dct encoding
      h263_dc_scale() cleanup
      automatically detect broken opendivx codec and workaround, so bug=1 shouldnt be needed anymore for mpeg4 files, if there still are files which need bug=1 then tell me ...
      autodetect UMP4 (by adding a fourcc field to AVCodecContext)
      ff_get_fourcc() & XVIX support
      minor optimization of h263_encode_motion()
      idct permutation cleanup, idct can be selected per context now
      renaming variable
      indenting preprocessor stuff, as its completly unreadable otherwise
      convert only needed matrixes in convert_matrix() (mjpeg calls it for every frame)
      forgot zigzag_direct[] behind #ifdef DEBUG (found by Klaas-Pieter Vlieg <vlieg at eurescom dot de>)
      optimizing mpeg1_decode_block()
      fixing rv10, this isnt the cleanest solution (parsing the packet header in the codec & creating it in the muxer) but it was that way before things broke, and its the simplest solution
      rv10 cleanup (de)muxer still needs to be cleaned up (still searching volunteer for that ;) )
      optimizing mpeg2 decode block stuff
      move scantable init from block-decode to header parser
      sliced mode for mpeg1/2
      optimization
      fixing custom quantizer matrix decoding
      trying to fix the non-x86 IDCTs (untested)
      set low_delay on encoding if no b frames
      mpeg1 2pass encoding
      yuy2 output support
      better/cleaner error resilience (done in a 2nd pass after decoding)
      mmx yuy2 output
      fixing yuy2 upscaling (use -sws 0 for the faster but uglier variant)
      fixing wmv1 slice decoding
      mpeg4 header parser clenup (needed for parsing of VOL header in avctx->extradata)
      better padding bug detection
      fixing unused variables & adding missing options
      warning fixes
      fixing alt_scan for the first frame (variable was reset)
      complete mpeg4 GMC decoding support
      init flags before using it (found by arpi)
      avoid multiple init
      fixing interlaced dct + hq mode
      fixing h263 decoding (did segfault)
      fixing interlaced dct with 4mv
      fixing h263 slice decoding (again)
      fixing msmpeg4v3 encoding at bitrates <128k
      new video grab device for ATI All-In-Wonder cards based on the GATOS km driver patch by (Fred <foohoo at shaw dot ca>)
      xvid qpel bug workaround
      xvid qpel bug autodetect
      more defaults
      fixing interlaced MC & edge-emu
      optimize block_permute()
      clip MVs for direct mode
      slice encoding cleanup
      regression test for H263+
      idct_permutation_type variable, so the permutation type can quickly be identified
      fixing decoding of streams with no keyframe at the start
      handle direct rendering buffer allocation failure
      75% smaller rl_length table
      mergeing inter & intra rl_length tables (50% smaller & no meassureable difference in filesize/quality)
      get_vlc2()
      postprocessing cleanup:
      using fewer registers (fixes compilation bug hopefully)
      -fomit-frame-pointer is needed for the PIC code
      better deblocking filter
      cosmetics ;)
      cleanup
      10l
      cleanly passing the cpuCaps
      using same numbers for cpuCaps as libmpeg2 & liba52
      exporting the internal qscale table, this allso fixes the ordering
      change qscale type to int8 and fix qscale ordering
      per context cpuCaps (idea by kabi)
      removing "-npp" (found by arpi)
      fixing q>0.0 assert failure caused by overflow of variance for b frames
      fixing another assert q>0.0 issue caused by variance < 0, this fix allso changes the inter/intra decission very slightly -> all regression checksums need to be updated
      using fewer registers ... to workaround something
      disabling compilation of .so if its not needed, gcc 3.1 seems to have problems with it
      cleanup
      dont call emms_c() for each MB
      put a few large tables under #ifdef CONFIG_ENCODERS or dynamically allocate them
      cosmetics requested by kabi
      hopefully fixing sprite_warping_points==0
      fixing msmpeg4v3 with mb_height % slice_height != 0
      put MIN/MAX under ifndef MAX/MIN
      better padding bug detection
      cleanup
      Altivec Patch (Mark III) by (Dieter Shirley <dieters at schemasoft dot com>)
      cleanup
      cleanup
      support forcing low_delay during decoding
      use cp ; strip instead of strip -o
      -p
      pp docs
      fixing dependancies
      10l
      vc++ sucks, found by ("csharprules" <csharprules at hotmail dot com>)
      dont use fastmemcpy for PIC
      lrintf detection (based upon a patch by François Revol <revol at free dot fr>)
      cleanup
      cleanup
      cleanup
      cleanup (breaks compatibility, requested by fabrice)
      cleanup
      fixing -sameq with mpeg2
      use FF_INPUT_BUFFER_PADDING_SIZE
      hurry_up>=5 -> skip everything except header & set pict_type&key_frame
      avoid duplicate ";" warnings (based upon a patch by mru at users dot sourceforge dot net (Måns Rullgård))
      reversing header game
      100l for kabi (fixing qpel)
      1l found by kabi
      huffyuv
      YV12 support (warning this is experimental & wont work with offical huffyuv but there is a approx. 20% speed & compression gain)
      0.1l
      dr1
      slices & encoding optim
      lower case div3
      hurry_up=5 support
      fixing compilation of cs_test
      10l
      aspect ratio cleanup
      moving init of some variables (hurry_up,...) to MPV_frame_start()
      fixing aspect
      10l
      fixing h263+ slices if the padding at the end of the frame is >7 && <16 && there are no zero bytes afterwards
      apiexample doesnt send complete frames to the codec
      100l (mixing chroma & luma linesize)
      100l (sad8x8_x2 had params in wrong order)
      cleanup
      use spatial error concealment if we dont have a previous frame
      fixing illegal 3. esc bug (the mpeg4 std only requires encoders to use unescaped symbols but not esc1 or esc2 if they are shorter than esc3, andjust beause its logical to use the shortest possible vlc doesnt mean encoders do that)
      specific debug output support
      workaround old xvid bug
      print a warning if something allocates 0 bytes
      new PSNR code (now works with chroma, b frames, ...)
      fixing files where the first frame isn a keyframe
      infinity fix by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      drop b frames if there are no reference frames
      fixing hq + adaptive quant
      better? scene change detection
      missing "\" (found by nick)
      fixing 2pass assert failure
      fixing aspect (hopefully, i couldnt reproduce the bug)
      mpeg1&2 aspect decoding
      aspect ratio encoding for mpeg1
      fixing api-example
      ump4 decoding fixed
      remove fake wmv2 codec
      correct mpeg4 vo type
      cleanup / messup?
      AVVideoFrame -> AVFrame
      100l
      fixing playback of DaveMatthews_Crash_PocketPC.avi
      more debug output
      support dumping the qscale stuff to the screen
      mb qp limits
      10l
      my TODO ...
      fix segfault with -vcodec copy
      huffyuv 4cc & -strict
      qpel encoding
      update
      motion estimation changes
      buffer alloc cleanup / 10l
      xvid build 3 still has the padding wrong in 1/8 of the cases :(((((
      better set pix_fmt explicitly
      removing PIX_FMT_ANY
      ooops forgot this (ffserver regression checksums update after motion estimation changes)
      cleaning corners of green dirt ;)
      correcter mpeg4 headers
      10l (int i)
      trellis quantization
      more optimized trellis quantizer
      trellis quantization for mpeg1
      mpeg1 rounding fix
      shape adaptive diamonds for EPZS
      regression checksums change cuz of motion estimation changes 10min ago
      bugfix
      motion estimation pre pass
      pre motion estimation cleanup/bugfix
      mpeg4 header encoding bugfix
      qpel in mmx2/3dnow
      negative linesize support (so mplayer -flip works)
      YUV 411/422/444 support for pp
      slowdown / gcc 2.95.* bug workaround (this should be reversed as soon as gcc 2.95.* support is droped)
      qpel in b frames bugfixes
      b frame segfault fix
      1000l (push & esp) using mangle now ...
      1000l, but for whom? (zero ME sucks and caused the many I frames in the regression tests)
      gcc 3.2.2 -O3 bug workaround (older gcc are very likely affected too but didnt check)
      more gcc bug workarounds
      PIC / ebx fix
      workaround dropable p frame after first frame bug
      fixed wmv2 slices
      qpel fix
      typo (noticed by kabi)
      optimization tips
      more links
      gmc1
      get_format()
      shorter test streams / faster regression tests
      average motion vector rounding like the reference says and not what the standard says
      /* align 16 */
      dct_unquantize_h263_altivec by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
      gmc1 segfault fix (found by kabi)
      better padding/stuffing bug autodetection
      padding bug autodetection improvement 2nd try
      10l
      more compare functions (rd & bit)
      fixing DCTELEM != short
      update
      AltiVec perf (take 2), plus a couple AltiVec functions by (Romain Dolbeau <dolbeau at irisa dot fr>)
      bugs
      finetuneing thresholds/factors
      oops forgot that ...
      add/put_pixels_clamped update
      align
      fixing yuv422 -> yuv420p (i need that for the vceq videos ...)
      trying to fix mb skip bug in mpeg1/2 if slices are not used
      GetBitContext.size is allways multiplied by 8 -> use size_in_bits to avoid useless *8 in a few inner loops
      cleanup
      reversing my own stupidity ... (raw packed yuv422 files dont use YUY2 but UYVY)
      UYVY input
      feeding doxygen
      drawing cute arrows instead of lame lines for motion vector visualization
      optimizing av_log2
      10l / 4mv segfault fix
      --disable-risky support
      bugfix
      100000l
      fixing rgb4 & bgr4 (2 pixels per byte)
      fix for ppc compile broken by (Romain Dolbeau <dolbeau at irisa dot fr> and Colin Leroy <colin at colino dot net>)
      10l found by (Dominik Mierzejewski <dominik at rangers dot eu dot org>)
      divx503 decoding fix
      yuv9 -> yv12 bugfixes
      cliping MVs, i dunno if its correct but it looks better then without it
      b frame segfault fix (+-1 bug)
      checksum changed cuz of b frame segfault fix?
      mpeg1 bframe encoding patch by (Raphaël LEGRAND) with some modifications by me
      changes to f_code calculation for b frames -> changed checksum
      update
      mpeg1 b frame regression test
      direct blocksize in bframes fix (might fix qpel+bframe bug)
      Creative YUV (CYUV) decoder by (Mike Melanson <melanson at pcisys dot net>)
      check for qscale==0 (fixes 1/0 on one corrupted stream)
      cleanup (unsigned stride -> int stride)
      yv12 -> yuy2 converter in alpha asm (from mplayerxp)
      -fPIC compileable
      #ifdef c++
      avoid #include "../
      cleanup
      cleanup
      brightness / saturation / contrast / different yuv colorspace support for some yuv2rgb converters (many converters still ignore it)
      cleanup
      improving parsing of incomplete headers
      moving postprocess to ffmpeg/libavcodec
      simpler & faster
      bigendian bug (fixes?)
      1000l
      fixing 410 -> 420
      optimize
      cleanup
      reverse incorrect stride < (width+7)&~7 fix
      yuv2rgb brightness/contrast/saturation/different colorspaces support finished
      100l (gcc 3.2 .....)
      32  vs. 64bit bugfix by (Dave Huang (dahanc))
      sws_ prefix, more seperation between internal & external swscaler API
      round readed bits up to next 32bits, as orginal huffyuv cant handle less then 32bit blocks
      cleanup
      get_vlc2() "docs"
      huffyuv v1 tables, as they are essential and the only possible way for decding of v1 files they very likely cant be copyrighted ...
      move repeat_pict field from AVCodecContext -> AVFrame (closes bug #683536)
      user specified motion estimation range limit
      h263 me range increased -> checksum for h263, rv10 updated
      fixing padding bug autodetection for some rare files, closes bug #647941
      h263(+) clenaup & bugfixes
      10l
      fixing inaccurate frame_rate
      MpegEncContext.(i)dct_* -> DspContext.(i)dct_*
      typo (fixes wmv1 regression test in c)
      extended option for h263+ patch by (fixounet at free dot fr) with some minor modifications
      disabling aic for h263p regression test
      field picture decoding support (16x16 MC blocks only as i dont have any samples which use other modes ...)
      10l
      mpeg2 field pictures + sliced mode (doesnt work with mplayer though, dunno why)
      cleanup
      ff_block_permute() dox
      more idiot proof dox
      more dox
      doxygen config file, so ppl just have to call doxygen in the libavcodec dir and a beatifull doxy is generated ;)
      cosmetics
      doxy
      per file doxy
      altivec compilation fixes by (Magnus Damm <damm at opensource dot se>)
      typos & cosmetics
      doxy / cosmetics
      doxy / cosmetics
      typo
      more cosmetics so that doxygen output is readable ...
      reversing not yet reversed changes from r1.7 -> r1.8 except the static/const stuff
      merging fourcc with codec_tag
      read BITMAPINFOHEADER extra stuff (huffyuv decoding fixed)
      huffyuv encoding fixed
      update
      avi fps fixed
      wmv2 & huffyuv regression test
      forgot ....
      fixing frame_rate accuracy
      10l
      looks better IMHO
      bug workaround
      more fixes
      more fixes
      even more fixes
      fixing parsing of opendml avis
      per context frame_rate_base, this should finally fix frame_rate related av sync issues
      100l
      sorry forgot
      rvlc decoding support
      oops
      halfpel MC doxy
      cleaner & more flexible edge bug workaround
      field pic decoding cleanup
      10l
      qmin/qmax cleanup
      merging a small amount of the changes from BroadQ, the rest is either not clean / doesnt apply / or is PS2 specific (someone with a PS2 should merge/send a patch for the later)
      error resilience cleanup (its faster too...)
      db video chunk in avi files support (idea by Dmitry Poplavsky <dima at thekompany dot com>)
      better non conformant divx packed bitstream detection, so unpacked (no b frames) divx MPEG4-ES streams can be read
      10l
      bitstream tracing support
      10l
      binary compatibility fix
      user setable quantizer bias
      allways test ffserver too by default
      10l
      faster hadamard transform
      test ffserver even if ffmpeg regresion test fails
      10000l (YUV vs. YVU swscale fix/cleanup)
      10l
      fixing cliping of c deinterlacers
      faster C linear blend & interpolate deinterlacers
      disabling ac3 decoding regression test (its different between different compilers/platforms due to float use)
      simplify slice encoding (variance of packet size is practically identical)
      mpeg1 slice encoding support
      some static -> dynamic alloc & 16->8 bit
      fixing msmpeg4v2 encoding
      msmpeg4 2pass support & some related cleanup
      10l
      optionally optimize for size (libavcodec.a is >30% smaller), no i didnt test the speed ...
      fixing typo in some outcommentet code
      fixing mem corruption
      support for DV aspect ratio and erroneous audio patch by (Dan Dennedy (dan at dennedy dot org) and Roman Shaposhnick <rvs at sun dot com>)
      H264 decoder & demuxer
      h264 update
      multiple reference frames support
      #ifdef TRACE printf() -> tprintf()
      pict_type & quality fix
      some checks to avoid segfault
      fixing slice decoding, dunno why the regression tests didnt catch that ...
      b frames + slices bugfix
      simplified adressing of most mb based arrays (mb_x + mb_y*s->mb_stride) now instead of mb_x + mb_y*mb_width and 1+mb_x + (1+mb_y)*(mb_width+2) and ... mixture
      10l
      smaller slices for the regression tests, i hope that way we'll catch the next 10l slice bug earlier ...
      check quality too, not only md5 (but dont take the scores too serious, the different codecs use different quality/bitrates and its just based on a artificially generated testvideo)
      10l (adaptve quant fix)
      alternative test video generator by (Sebastien Bechet <s dot bechet at av7 dot net>)
      finished integrating rotozoom test in ffmpeg
      100l (mpeg2 decoding fixed)
      some checks to avoid segfaults if the decoder is feeded with junk
      use bilinear interpolation for the rotozoom
      checksum update for the avi zeros ...
      fail if a test fails instead of contuing as if everything was ok
      10l (use before malloc)
      faster 8x8 & 16x16 plane prediction by skal (massimin at planet-d dot net)
      10l
      10l (hadamard fix)
      minor cleanup / some warning fixes
      check validity of some flags
      10l
      meanigfull dr1 comment
      mplayer idependant (not really yet) swscale example
      segfault fix
      we really shouldnt use M$* as default codec -> use MPEG4 as default
      mpeg2 qscale flag
      better? default thresholds, if this is worse for any files, then tell us ASAP
      memleaks
      cleanup mpeg1/2 bitstream parser
      removing sws - global verbose var dependancy
      doc fixes (forgot to commit)
      fixing mixed dr1 + internal buffers
      moving getFilter stuff back (vf_scale.c -> swscale.c)
      typo in a comment ...
      remove remaining cpudetect dependancy
      bigendian fix by (Samuel Kleiner <kleiner at cd dot chalmers dot se>)
      2pass stats fix (by ffdshow cvslog)
      10l
      10l (returning negative number of consumed bytes if the first startcode of a frame was split between 2 buffers)
      tired
      10l
      doxy
      VOB stream patch ba (Brian Foley <bfoley at compsoc dot nuigalway dot ie>)
      Tiny patch for OpenBSD by (Björn Sandell <biorn at dce dot chalmers dot se>)
      release buffer cleanup
      more #ifdef CONFIG_ENCODERS by (Wolfgang Hesseler <qv at multimediaware dot com>)
      svq3 decoder by anonymous
      print obmc bug info only if avctx->debug&1
      more idct_permute docs
      100l (document buffer padding requirements)
      optimizations
      optimizations
      some libmpeg2 style bitstream reader fixes (no dv doesnt yet work with it)
      libmpeg2 style bitstream reader fixes
      make sh4 IDCT optional, so the user can choose
      libmpeg2 style bitstream reader 17 vs 16 bit bugfix
      cleanup
      optimize
      moving the svq3 motion compensation stuff to dsputil (this also means that existing optimized halfpel code is used now ...)
      CODEC_CAP_DRAW_HORIZ_BAND
      bigendian fix
      seems i guessed correctly (last 411 chroma block isnt scaled but cut and reordered)
      improving 3IV1 support (still far from perfect) note, disabled as it requires some checks in the inner loops which would slow things down a tiny bit
      ASV1 codec
      cleanup
      altivec gcc-3 fixes by (Magnus Damm <damm at opensource dot se>)
      truncated h263 decoding support / H263-ES "demuxer"
      cleanup
      optimize
      some benchmarking code
      c frame size debug stuff
      10l
      10000l (__CPU__ wasnt set)
      motion vector vissualization for mpeg1/2
      asv1 regression test
      CABAC
      mpeg1 decoding fix
      fix mpeg1/2 decoding if there are no 0 bytes after the bitstream
      (truncated) unary binerization
      rgb15 & rgb16 fix
      4xm codec
      10l (array[-1] ...)
      fix 4xm yuv->rgb565 transform
      vf_scale.h & related cleanup & some small warning fix by dominik
      4xm adpcm
      cleanup
      more comments
      FFV1 codec (our very simple lossless intra only codec, compresses much better then huffyuv)
      10l
      reducing memory consumption (reducing number of contexts and other stuff, no noticeable loss in compression rate)
      golomb rice codes
      lossless jpeg decoding support
      cleanup
      lossless mjpeg encoding (planar yuv & RGB) and somerelated bugfixes
      update
      make ff_emulated_edge_mc() independant of MpegEncContext
      fixing "internal error in orderYUV" bug
      fixing concealment MVs prediction
      the return of the idct with 16bit output by ("Ivan Kalvachev" <ivan at cacad dot com>)
      CONFIG_ENCODERS cleanup
      blinking blocks around thin vertical lines and dots bugfix
      2x100l
      gcc 2.95 workaround
      fix CHECKED_ALLOCZ(0)
      try to filter _all_ PATs if no SDT could be found patch by (<gbazin at altern dot org>)
      PPC970 patch + cpu-specific tuning support by (Romain Dolbeau <dolbeau at irisa dot fr>)
      20% less memory needed for 8bit AC state (worst compression rate loss was 0.2%, all other files had <=0.1% loss)
      cleanup / cosmetics
      segfault fix by (Giancarlo Formicuccia <ilsensine at inwind dot it>)
      fix last_block_index<=0 bug
      Missing ffserver.c format parameter fix by (Giancarlo Formicuccia <ilsensine at inwind dot it>)
      undefined local_port fix by (Giancarlo Formicuccia <ilsensine at inwind dot it>)
      mpeg1 bad frame_rate_base fix by (Arthur van Hoff (javanator))
      av_free_packet  SEGV fix by (Arthur van Hoff (javanator))
      memleak fix by (Jake Page <jake at CS dot Stanford dot EDU>)
      merge U and V statistics, 33% reduction in memory requirement, compression rate better for some files worse for others, worst compression rate loss 0.05%
      CODEC_FLAG_LOW_DELAY fix
      mpeg 2 seeking fix
      the reference sw doesnt like some legal headers ...
      few more error checks
      FFmpeg & evaluating performance on the PowerPC Architecture HOWTO by (Romain Dolbeau <dolbeau at irisa dot fr>)
      flv regression test
      flv cleanup / simplify
      fix? flv escape codes
      1000l (forgot to commit)
      postprocessing support
      Building ffmpeg with gcc-2.95.3 encountered a problem due to C99 initialization fix by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      lossless jpeg regression test
      10l
      cleanup
      simpler
      golomb rice code cleanup / simplify (~0.5% compression gain and slightly faster)
      10l
      10l
      flv1 fix
      fix qmin==qmax==1 && msmpeg4 && intra bug
      fix edge repeating bug for %16!=0 files, this fixes Quicktime mpeg4 (they arent buggy)
      edge fix
      AVCodec.flush()
      svq3 b frame slices fix
      default for slices is display order again
      bitstream changes to match JM7.2
      last frame decoding fix
      last frames & b frames encoding fix
      ATI VCR1 decoder
      fix
      ATI VCR2 decoding support
      ATI VCR2
      vcr2 cleanup
      ati vcr2 uv swap
      debug stuff
      rate distortion mb decision support
      patch for flv deblocking by (Garrick Meeker <gmeeker at theoryllc dot com>)
      namespace fix (triggered by gcc-alivec-bugfix + gentoo or so) by (Magnus Damm <damm at opensource dot se>)
      make the h263 decoder more error tolerant
      FF_INPUT_BUFFER_PADDING_SIZE
      clear bitstream buffers in AVCodec.flush()
      cleanup / fixes
      check input pix_fmt
      skip ext header missing message for msmpeg4v2
      fixing thread safety, this may cause alignment problms (if it does than the stuff should be moved to MpegEncContext)
      use codec_tag for encoding too
      fix AVFrame.reference (the frames are used for decoding future frames so it should be !=0)
      MC rounding fixed?
      fixing chroma MC
      AC3 encoding patch ba (Ross Martin <ffmpeg at ross dot interwrx dot com>)
      PSX MDEC decoder, based upon some code from Sebastian Jedruszkiewicz <elf at frogger dot rules dot pl>
      custom quant matrix encoding support
      xvid edge bug (if i understand the mpeg4 spec correctly) autodetection
      mpeg2 encoding
      oooooops, having 2 Eterms open and doing cvs diff in one and cvs commit in the other sucks, especially if they are in different directories (MPlayer/main/libavcodec vs. ffmpeg/libavcodec)
      mpeg2 regression test
      fixing CODEC_FLAG_GLOBAL_HEADER
      mov, 3gp, mp4 muxer by (Thomas Raivio <tjraivio at cc dot hut dot fi>)
      100l (forget to reset me.skip for b frames)
      M_PI diff for output_example (handles undefined M_PI) by (Bill Eldridge <bill at rfa dot org>)
      ffplay diff commenting out rtsp for no network by (Bill Eldridge <bill at rfa dot org>)
      10l (closed gop bit)
      bit_rate in mpeg1/2 should be 0x3FFFF for vbr or at least >= max_bitrate
      Small Patch for "ffplay -" instead of "ffplay pipe:" by (Bill Eldridge <bill at rfa dot org>)
      libavformat/avio.h compilation problem in VisualC++ by (lethean at realtime dot ssu dot ac dot kr)
      10l (big vs. little endian)
      fixing unused f_code values
      ASV2 decoder
      asv2 encoding
      asv2 regression test
      some fixes (still buggy though)
      more fixes
      100l
      fix MPEG-TS missdetected as MPEG-PS
      CODEC_ID_MPEGVIDEO
      10l
      MingW patches by (Glenn Maynard <g_sf at zewt dot org>)
      case insensitive codec_get_id()
      Compile fix when no CONFIG_ENCODERS by (Glenn Maynard <g_sf at zewt dot org>)
      fixing copyright header (the code is from strptime.c so the copyright header should be from there too instead of strptime.h)
      rewrite
      disable a52 & pp by default
      big/little endian fix
      padding fix
      memory leak fix by (Tom Dexter <devel at www dot digitalaudiorock dot com>)
      kernel header bug workaround by (Ronald Bultje <rbultje at ronald dot bitfreak dot net>)
      memleak fix by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      --disable-risky fix by (Luca Abeni <lucabe72 at email dot it>)
      mov/mp4 muxer cleanup (mostly cosmetics/simplifications & global header fix)
      detect avcodec_open() on an already opened AVCodecContext
      theres no SVQ1 encoder -> it cant be default
      detect old xvid with fourcc=DIVX
      get_bi(), be/le fix
      AVI stream header FCC / more reliable detection of old xvid files
      mpeg sequence end code regression checksum update
      another non const static, maybe its thread save now
      scenechange_threshold
      constraint_parameter_flag fix
      slight quality improvement ( < +0.04 PSNR) for high bitrate videos with trellis quantization
      fixes levis.avi
      h263 stuffing decode fix
      srcSliceY % 4 != 0 fix
      100l
      hurry_up fix?
      simpler
      AMR-WB support by (Johannes Carlsson <joca at rixmail dot se>)
      init AVFrame with defaults
      10l
      Patch for attribute(printf) by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      rate distortion optimal cbp support (h263/mpeg4 non intra only)
      ac prediction cleanup
      nut regression test
      optionally try to encode each MB with MV=<0,0> and choose the one with better RD
      rd cbp + single coeff elimination fix
      Patch for AAC encoding with libfaac by (Gildas Bazin <gbazin at altern dot org>)
      10l ?
      mpeg1 + trellis quant segfault fix
      use lagrange multipler instead of qp for ratecontrol, this may break some things, tell me ASAP if u notice anything broken
      10l
      1E10l
      yv12->uyvy MMX
      recommit of
      recommit of
      recommit of
      recommit of
      forgot to commit, i hope its correct
      intra dc treshold != 0 decoding fixed
      h263 GOB decoding fix
      fix decoding if the aspect ratio changes
      dont check if the ac esc 3 could have been stored as vlc as this detects only very few errors allthough it detects several buggy encoders
      10l
      mmx2 optimization of huffyuv median encoding
      alignment 10l
      fix the case where all vlc codes are 8 bits long (repeat=256)
      recommit of
      recommit of
      rounding fixes
      dither scaling coefficients so they add up to 1.0
      fix "first frame is no keyframe" bug
      align
      default_get_buffer() fixes
      picture buffer cleanup, this fixes a assertion failure (if assertion checking was manualy enabled for utils.c) and a memleak at least
      emu_edge checks simplification and avoid redundant checks for mpeg1/2 if emu_edge is set
      OS/X 10.2.x gcc-3.1 bug workaround by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      export mpeg2 active display area / pan scan
      bottom_field_first -> top_field_first
      AVRational
      10l
      use continued fractions to approximate a fraction if its numerator or denominator is too large
      mpeg2 aspect ratio encoding fixed
      BSD doesnt have log2
      start time hack by (devik <devik at cdi dot cz>)
      forgot to apply this chunk appearently, no, i dont really remember where it came from
      dynamic alloc of picture structs instead of putting them in MpegEncContext
      minimum slice size fix (fixes 422P 1 line per slice, used by huffyuv)
      move q_*_matrix out of MpegEncContext (40k ->23k) dct_quantize() is even slightly faster now, dont ask my why ...
      move mmx quantizer matrixes out of MpegEncContext (23k -> 7k) (no meassureable slowdown)
      move blocks away MpegEncContext (7.4 -> 5.8k)
      112 bytes smaller MpegEncContext
      floating point AAN DCT
      oops forgot
      10l
      optionally merge postscale into quantization table for the float aan dct
      4.9-RC FreeBSD doesnt like lrint() so change to lrintf(), not that bsd supports that but we emulate it ...
      inline -> always_inline (842 -> 690 cpu cycles for dct_quantize() difference for the dct itself should be even bigger)
      minor mmx2 optimization if the dct
      optimizing
      diff -up looks better then diff -u
      segfault fix
      10l
      reorder table instead of wasting instructions to reorder the input to match the table
      fixing the remaning block-align checks
      make convert_matrix() independant of MpegEncContext
      linear blend 1 line shift fix
      1/0 fix by (Tim Allen <tim at proximity dot com dot au>)
      dumping funny numbers from first zygo i frame header
      pts fix and related fixes
      100l (SAR is written as height/width instead of width/height in the MPEG1 standard)
      initalize various uninitalized variables and avoid coded_picture_number as its not always correct (later should be reversed after fixing the picture_number mess)
      move identical code into its own function
      int_fast??_t emulation for broken systems which dont conform to ISO C
      ;
      ffv1 rgb support
      10l ...
      noise reduction of dct coefficients
      10l (forgot that the dstW offset is hardcoded in the asm)
      10l (fixes dct coeff noise reduction)
      only add prefix after \n
      flush fix (untested)
      av_log patch(2 of ?) by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      2nd attempt at OS/X shared/mdynamicnopic fix by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      ffv1 regression test
      removing borderline spyware
      use .mp4 in one of the mpeg4 regression tests
      -sws 0 && -ssf chs fixed
      fix context=1
      1000l
      last frame
      rate distored optimal lambda->qp support
      ac prediction with rate distorted qp fix
      small mjpeg fix by (James Stembridge <jstembridge at users dot sourceforge dot net>)
      better fix
      qprd needs mbd=2
      qprd with b frames fix
      av_d2q() documentation
      Patch for PPM probing by (Rob Joyce <rjoyce at twcny dot rr dot com>)
      10l
      16x8 MV vissualization by (Wolfgang Hesseler <qv at multimediaware dot com>)
      dont read from b frames if draw_horiz_band==NULL
      h263 OBMC & 4MV support
      cleanup
      10000l
      intra16x16 fix
      h263 alternative inter vlc support
      forgot
      rv20 decoder
      rv10 fix
      cleanup
      rv20 sub_id == 0x20200002 support
      4th frame type ...
      h263 loop filter
      10l
      h263 MV prediction doesnt match mpeg4, for some slices configurations (fixes RV20 MVs)
      h263_v_loop_filter_mmx
      h263_h_loop_filter_mmx
      split ff_h263_decode_mb() into h263 and mpeg4 versions
      h263 modified quantization fix
      cleanup
      altivec.h bug workaround (untested)
      rv20 (h263) b frame decoding support
      cleanup
      set has_b_frames correctly, fixes rv20 dr1 with mplayer
      export top_field_first & interlaced_frame
      AVFrame.interlaced_frame fixes
      removing unused var & converting 64->32bit
      h263 slice structured mode
      slice flush / position fix
      alternative inter vlc encoding fix
      AIC/MQ encoding fix
      Advanced intra coding, modified quantization and alternative inter vlc regression tests
      better get_packet_size and probe
      motion_subsample_log2
      slice structured fixes
      100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l100l
      MB_TYPE_INTERLACED info fix
      10l (uninitialized var ...)
      interlaced dct encoding fix
      set AVCodecContext.width/height before get_buffer()
      obmc encoding flag (forgot to commit...)
      fixing h263p encoding without UMV
      CBR improvements
      stuffing to stay above min_bitrate
      mb type & qp vissualization
      cleanup
      error concealment regression test
      split intra / inter dequantization
      10l (printing qscale before parsing it)
      mv vissualization & field picture fix
      100l (forgot alpha & ppc)
      motion estimation & mb compare functions command line options
      vbv_delay
      svq3 fix
      fix crash
      rv20 / h263 b frame fix
      10l (forgot to copy top_field_first flag)
      its nice to include 'what it does & why' in submitted patches
      10l (segfault fix)
      picture number fixes
      nicer looking PSNR stats (YUV seperate and print the whole file PSNR at the end instead of the last frame's)
      100000l
      interlaced motion estimation
      info about h and tpel
      fixpoint log() for tiny_psnr so it can output actual PSNR
      print vbv buffer size & bitrate when decoding with -debug 1
      minrate=0 for SVCD & DVD, this matches mpeg2enc and the mpeg1/2 vissual standard
      rate limit tips and checks
      test interlaced motion estimation too
      dont show uninitalized motion vectors
      put_pixels width comment fix
      trellis quantization regression test
      optimizing (trellis) quantizer
      cleanup
      more trellis quant optimizations
      noise reduction regression test
      denoise_dct_mmx()
      forgot ...
      SVCD scan offset stuff
      warn the user if we had to clip some dct coefficient due to a crappy format which doenst support the whole needed range (msmpeg4/wmv mostly but mpeg1 too to a lesser extend)
      non YV12 vissualization fix by (Wolfgang Hesseler <qv at multimediaware dot com>)
      mpeg2 QP clamping fix
      frame rate emulation "fix" by (Mean <fixounet at free dot fr>)
      10000l
      10l
      gop timestamp fix
      simplify
      interlaced dct decision cleanup
      10l
      LDFLAGS fix by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      fix assertion failure when writing mov files
      mpeg1 overflowing dc coeff fix
      patch for configuring amr_wb by ("Murphy Chen" <murphychen at mail2000 dot com dot tw>)
      memleak fix
      uninitalized variables fix by (Gildas Bazin <gbazin at altern dot org>)
      optimize compute_antialias() and add a floating point based alternative (2x faster)
      something unimportant ...
      alpha amr tip
      cleanup
      more sane inttypes emulation behavior if libavcodec is used outside ffmpeg
      copyright year update of the files i touched and remembered, things look annoyingly unmaintained otherwise
      grayscale mjpeg decoding support based upon a patch by (Leon Bottou (leonb))
      fixing some seek failure issues
      audio pts fix
      fix b frame after seeking blockmess
      oops
      ignore damaged video frames
      type fixes, found by Wolfgang Hesseler
      caching of timestamps for mpeg-ps so seeking is faster
      fixing twice added offset bug, was triggered by 4mv + sub_cmp != mb_cmp
      dont calculate stuff if its not used ...
      memleak fix
      seeking fixes
      unused
      use packet number instead of byte number internally for seeking
      dont let the demuxer thread die before the video has actually been played as seeking would become impossible shortly before EOF
      replic_size==0 fix, needed so that G.726 asf file, so ffmpeg can at least decode the video without segfaulting
      header parser fix
      xvid "packed" stuff support
      store searched distance in index, so we dont waste time searching for keyframes where we already searched
      seeking stuff
      CONFIG_ENCODERS fix by (Ronald Bultje <rbultje at ronald dot bitfreak dot net>)
      rv20 mv prediction in b frames fix
      removing broken framerate conversation hack in mpeg1/2
      segfault fix
      avcodec_flush_buffers() fix
      more tolerance for dc=0
      closed gop support & flags2 as all bits in flags are used
      segfault fix
      optimization
      optimization & bugfix extracted from the 4k line diff between ffmpeg 0.4.7 and http://www.alicestreet.com/ffh263.html
      mpeg4 pts fix
      int64max fix by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      10l
      first coded vs. first displayed pts (segfault & pts fix if b frames are used)
      4mv + b frame encoding segfault fix
      SSE2 fdct by (Balatoni Denes <pnis at coder dot hu>)
      b frames + cbr fix
      segfault fix
      Patch for MPEG-2 VOB headers by (Jimmy Blair <blueskyjb at verizon dot net>)
      quantizer noise shaping
      simplify
      fixes by Gildas Bazin <gbazin at altern dot org>
      cleanup ugly code (also appearently works around a bug gcc 3.4 which fails compiling it) by (mru at kth dot se (Måns Rullgård))
      remove incorect motion estimation table
      quantizer noise shaping optimization
      analyze the gradient of the error surface to avoid a few checks in the quantizing noise shaper
      segfault fix
      h263 4mv encoding fix
      no limits!, warning is enough
      100l
      g726 regression test
      use clip() from common.h
      pass strict_std_compliance to audio context too
      segfault fix
      M$ ADPCM encoding & some simplifications
      ima & ms adpcm regression test
      2 byte shorter userdata for mpeg4
      rv10/obmc fixes
      ppc chroma mess workaround (real bug is that the motion compensation code assumes that 2*uvlinesize == linesize and fixing this would mean a slowdown)
      multithreaded/SMP motion estimation
      simplify MV availability check / dont use below last row MV which is always zero
      replace a few pbBufPtr() by put_bits_count(), one of them was actually wrong
      activate the memcpy special case for the byte aligned case in ff_copy_bits(), wasnt working before because of a incorrectly used pbBufPtr()
      threadless threads warning
      ff_copy_bits() optimization
      multithreaded spatial complexity calculation for i frames
      mpeg4 b frame + adaptive quantization assertion failure fix
      normalize adaptive quantizatiuon fix (based upon a patch by (Jindrich Makovicka <makovick at kmlinux dot fjfi dot cvut dot cz>))
      msadpcm encoding
      fix image resize
      more error checks for the sequence header
      noise reduction with threads bugfix
      flac decoder by (Alex Beregszaszi <alex at fsn dot hu>)
      flac fixes:
      remove residual array, it isnt really needed
      crc8 checking, based upon a patch by (Miroslav Lichvar <lichvarm at phoenix dot inf dot upol dot cz>)
      simplify
      flac_flush()
      check if threads_count is too large
      update libavformat regression checksums (it should be checked why exactly they changed ...), and always test libavformat in make test so this doesnt happen again
      auxdatae, crcrs fix by (Jean-Francois Panisset <panisset at comcast dot net>)
      simplify
      w32threads by (Gildas Bazin <gbazin at altern dot org>)
      (f)printf() is disallowed in libavcodec, compilation will fail now if its used, except that codecs which where added after the printf->av_log change which did ignore av_log() and used prinf are now silent and wont print anything, they should be changed to use av_log, i could do that, but its better if the orginal developer decides which AV_LOG level each message should get
      c99
      av_log()
      multithreaded mpeg2 decoding
      set AVFrame.pts based upon framerate if its 0
      use AVFrame.pts=AV_NOPTS_VALUE instead of AVFrame.pts=0
      1l
      Wolfgang Hesseler (mv/dct coeff export & mv vissualiztion)
      100l (field picture decoding fix)
      10l
      INT64_MIN fix
      AVCodec.supported_framerates
      10l (vdpart fix)
      Metrowerks CodeWarrior patches by (John Dalgliesh <johnd at defyne dot org>)
      mpeg2 field pic decoding fix
      interlaced mpeg4 + edge emu fix (fixes decoding of mermaid.avi)
      simplify MPV_motion()
      skip b frames before first i frame
      optimizing/cleanup of mpeg/qpel_motion()
      100l (divx5 b frame + seeking + flush fix)
      10l (segfault fix)
      macrofree av_freep()
      av_log() cleanup
      count > thread_count for execute()
      10l
      fix motion compensation with (x+1/2,y+1/2) MVs
      write trailer before closing codecs as otherwise accesses to extradata in write_trailer() will cause a segfault
      h263 framerate & aspect ratio fixes
      channel count & samplerate fix
      regression test fix (forcing specific audio codecs as the default often depends upon HAVE_MP3LAME or similar)
      find_start_code() optimization (about 2x faster now) this may improve decoding speed with multiple threads
      flv fixes
      10l (1 vs. descrLength(1)) found by (Jean Le Feuvre - jeanlf)
      disable lames leading zero "feature", fixes a few more flv encoding issues
      guess framerate
      h263+ custom pcf support
      mb stuffing in partitioned frames fix
      nicer output for 'ffmpeg -formats'
      10l
      matroska demuxer by ("Ronald S. Bultje" <R dot S dot Bultje at students dot uu dot nl>)
      big endian compile fix by ("Ronald S. Bultje" <R dot S dot Bultje at students dot uu dot nl>)
      (f)pritf() vs. av_log() and 'make test' update
      removing broken mpeg_decode_gop()
      no read loop tcp/http and http CRLF fix by (Leon van Stuivenberg <l dot vanstuivenberg at chello dot nl>)
      width,height%16!=0 fix
      alignment fix
      init cleanup
      1000000l (using uninitalized variable for initalizing bits per MV table)
      moving f/b_code init to where it should be
      adjust_write_index() fix by ("Curi Fabio Eduardo (SFL)" <curif at TELEFONICA dot COM dot AR>)
      wrong help text
      h.263 ssm + aic fix by (Maarten Daniels <maarten dot daniels at student dot luc dot ac dot be>)
      slice structured encoding support & interlaced spelling
      enable .ffm regression test as it works again
      h264 loop filter for progressive I&P frames by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      h264 loop filter optimizations
      do loop filter immediatly after each macroblock is decoded instead of after a frame is decoded
      static allocation rewrite (old code was plain a broken mess)
      1000l in av_mallocz_static()
      powerpc PMC doc fix by (Romain Dolbeau <dolbeau at irisa dot fr>)
      Compile fix by (Chip <szarlada at freemail dot hu>)
      1000l for iive
      test more (i)dcts
      add size of encoded stream to the regression tests
      untested win32 binary pipe fix
      forgot (bitstream size info)
      cleanup frame_rate code, this may also fix some large file bugs
      always parse APP* as it sometimes contains startcodes which confuse the decoder otherwise
      huge vbv buffer size + CBR fix
      h263 loop filter option
      fix decoding of qp<6
      ensure that the bitstream buffers for encoding partitioned frames are aligned
      diff -w
      hmm av_log(..., get_bits());
      fixing alignment, 2nd try
      split stream into valid mp3 frames, at least flv & nut absolutely need this, but probably most other formats too
      pass frame_size for stream copy
      gmc segfault fix
      skiped mbs in interlaced b frames fix
      interlaced chroma MC divx-bug workaround
      rv20 (0x20201002) support
      10l (vismv fix)
      nicer looking arrows for the motion vector vissualization
      nut (de)muxer update
      remove spyware from ogg*.c
      kill obnoxious ogg_packet passing from demuxer to decoder
      time & (s)rand is forbidden
      fix obnoxious ogg_packet passing from encoder to muxer
      remove function call from muxer->encoder and cleanly pass global headers
      pts hack (correct solution would be to pass the pts from the encoder to the muxer)
      fix global header passing from demuxer to decoder
      move zero size hack from ogg.c to utils.c
      large extradata fix
      some random id for vorbis so we can do some experiments with vorbis in various containers, anyone knows if vorbis in WAV/AVI has a official id?
      store extradata, vorbis in avi works now, dont expect av sync though
      10l
      make vorbis default again
      info packet read/write & simplify
      10000l
      10l (Titel->Title)
      antitime
      checksuming for nut & nice checksum API for libavformat
      search for undamaged headers
      10l (bytes_left() -> url_feof())
      dont reduce the min_distance of AVIndexEntry
      fourcc & language code v->vb type
      regression checksum update for alexs .mov time change
      require the user to explicitly specifiy --enable-gpl before enabling of gpl parts becomes possible
      fixing alignment problems -> SSE2 support enabled again in libavcodec (from ffdshow / milan_cutka)
      removing keyframe prediction and checksum threshold
      checksum threshold removial
      prefix global
      set codec_tag
      avoid seeking to the end, as it confuses some crappy code
      some tag for mpeg2, so looking MPEG2VIDEO up works
      4xm codec tag
      codec tag for FLV
      moving nearly identical binary search code from nut/mpeg/asf to utils.c
      10l
      distclean by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      bit-exact
      store correct pts
      assert(time_incr >= 0) so its clear that this MUST be the case
      10l
      print video/audio/global_header size and muxing overhead
      10l (stats didnt include av_write_trailer())
      unify table indexing (motion_val,dc_val,ac_val,coded_block changed)
      segfault fix
      +-1 bugfix regression test checksum update
      t/dprintf printf -> av_log
      10l by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      seems wmv8 doesnt use intra-inter prediction
      stream copy pts fix
      cleanup
      flv dropable p frame support
      passing AVFrame instead of AVPicture around in ffmpeg, that way stuff like motion vectors can be passed from the decoder to the encoder
      10l (segfault fix)
      pts fix / b frame segfault fix
      various fixes
      fix exported motion vectors
      fix y scale of mv vissualization of interlaced MVs
      10l
      exporting field_select/ref_index values
      hadamard8_diff8x8 in AltiVec, the 16bits edition by (Romain Dolbeau <dolbeau at irisa dot fr>)
      motion estimation cleanup
      10l (segfault fix)
      corrupted field pictures segfault fix
      RSHIFT(x,0) fix
      reuse motion vectors/mb types/field select values of the source video, if the SSE for a macroblock which is predicted with these values is below me_threshold
      h264 nal decoding fix by (Laurent Aimar <fenrir at via dot ecp dot fr>)
      sanity check
      VIS optimized motion compensation code. by (David S. Miller <davem at redhat dot com>)
      minor motion estimation cleanup
      sanity checks
      -qpel option
      qpel & SSE/SATD regression test
      support reusing mb types and field select values of the source file, but use motion vectors just as additional predictors
      ring buffer fix by (Luca Abeni <lucabe72 at email dot it>)
      rv10 decoding fix
      fix decoding with quant matrixes which contain elements <16
      lmin/lmax
      H.263 MCBPC fix
      NULL pointers ...
      export pts if available
      simplify
      >2 channels decoding fix
      remove duplicated find_frame_end() code
      10l
      10l
      ffmpeg -version
      OPT_INT
      intra_dc_precission>0 encoding support
      10000l
      cleanup & memleak fix
      dont missuse assert()
      simplify
      -bug
      AMV bug workaround
      mmx dct optimization
      dont use always_inine with -Os
      match latest spec
      bigendian fix by (Romain Dolbeau <dolbeau at irisa dot fr>)
      cleanup
      arbitrary precision integer support
      use AVInteger in av_rescale() so it can finally do 64*64/64 instead of just 64*32/32
      mpeg4 dc prediction simplification
      support decoding mpeg4 with buggy dc clipping
      coder, context, predictor command line options
      next_pts fix by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      vbr audio fix
      10l
      rate distorted intra only encoding
      color and 10l
      imlib2 needs -lm
      width height %64 != 0 fix
      using pointers so the code could eventually be used for P frames
      p frame encoding, only with 0,0 motion vectors yet though
      dummy avcodec_thread_init() to avoid linking issues
      reduce dependancy between motion estimation and MpegEncContext
      moving motion estimation specific variables from MpegEncContext -> MotionEstContext
      UINT64_MAX fix
      motion estimation for SVQ1
      support skiped blocks in SVQ1
      segfault fix
      svq1 regression test
      automatic pixel format selection
      automatic framerate selection
      removing unused encoding code
      segfault fix
      mention avcodec_align_dimensions()
      some of the warning fixes by (Michael Roitzsch <mroi at users dot sourceforge dot net>)
      bytestream tracing support
      muxing overhead statistic fix for stream copying
      forgotten prototypes
      oops, forgot to commit that change from the warning fixes by (Michael Roitzsch <mroi at users dot sourceforge dot net>)
      data_size = 0 cleanup
      more data_size=0 cleanup
      move time_base (pts_num/pts_den) from AVFormatContext -> AVStream
      ooops
      mpeg audio timestamp fix
      100l
      store index for seeking in the native timebase of each stream
      cur_dts in native timebase
      100l (AVStream.duration)
      100l (avi seeking)
      warning fixes by (Michael Roitzsch <mroi at users dot sourceforge dot net>)
      avcodec_find is in no header file, and appearently not used or very usefull so lets make it static ...
      10l (interlaced b frame decoding fix)
      10l fixes by ("Debabrata Banerjee" <davatar at comcast dot net>)
      altivec optimizations and horizontal filter fix by (Romain Dolbeau <dolbeau at irisa dot fr>)
      accurate/slow (per line instead of per block) deblock filter spport which is identical to what is recommanded in the mpeg4 spec
      faster c lowpass filter
      mmx optimized version of the per line/accurate deblock filter
      per line lowpass filter in mmx
      pass AVPacket into av_write_frame()
      exporting keyframe flags, fixes keyframe stuff with streamcopy
      export low_delay for encoding too
      correctly interleave packets during encoding
      much better AV sync, for some files
      skip subtitle streams instead of perishing
      10l
      h261 decoder by (Maarten Daniels <maarten.daniels at student dot luc dot ac dot be>)
      rewrite h261 loop filter
      simplify
      removing h263 specific code
      h261 dequant fix
      dv regression test
      fix playback of http://www.zonque.org/test-long.ps
      10l fix by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      remove stupid YV12-huffyuv warning
      10l (-top 0)
      -qprd
      infinite loop fix
      change qscale -> lambda for the motion estimation
      small linux/altivec compile fix in postproc/ by (Romain Dolbeau <dolbeau at irisa dot fr>)
      check for sdl-config
      update todo list
      qscale + qprd fix
      -cbp
      hadamard/AltiVec: fix to compiler fix by (Romain Dolbeau <dolbeau at irisa dot fr>)
      avg_pixels8_xy2_altivec in AltiVec, enabling avg_pixels8_altivec, hadamard fix by (Romain Dolbeau <dolbeau at irisa dot fr>)
      mini-fix for compatibility with other compilers by (Hauke Duden <H.NS.Duden at gmx dot net>)
      optional and disabled by default memalign hack for SSE/SSE2 on that alternative OS
      Building with faad2-cvs fix by ("Steven M. Schultz" <sms at 2bsd dot com>)
      cubic filter instead of truncated sinc
      sanity checks
      fix normalization
      noise preserving sum of squares comparission function
      nsse weight
      sse8 and nsse in mmx
      hadamard/AltiVec: fix to compiler fix, again by (Romain Dolbeau <dolbeau at irisa dot fr>)
      gcc bug workaround by (Jonathan Gray <khalek at linuxgamers dot net>)
      ac3 bugfix, bug found by "Ben Cooley" <bcooley at dstreamtech dot com>
      more fourccs
      clip input motion vectors
      chroma me fix
      timestamp discontinuity correction try #1
      fewer overflows
      file id string
      coefficient saturation fix
      pass timestamps correctly for -sync 0
      10l fix by (Bohdan Horst <nexus at irc dot pl>)
      polyphase kaiser windowed sinc and blackman nuttall windowed sinc audio resample filters
      avoid useless normalization and 10l fix
      10l
      regression checksum changes due to new resampling code
      100l
      segfault fix
      1000l to the ffsvq3 author, our default bitstream reader is only guranteed to be able to read 25bit at a time
      1000l to myself for FFMIN(...,0) instead of ...,6)
      1000l
      increase size of crop table
      flush audio encoder buffers at the end
      try to allocate the buffer before usig it :)
      Documentation for -itsoffset switch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      ogg with ffplay fix
      transmeta crusoe mmx detection by "Nobody/Anonymous - nobody" / sf feature request tracker
      optimization
      overflow fix
      use libmpeg2 idct to decode xvid videos unless the user forced some other idct
      simplify
      support skiping of mb rows during decoding
      altivec yuv->rgb converter
      which codecs does windows support
      -t + capture bugfix
      10l
      timestamp duplication bugfix
      forgotten
      various resampling fixes
      audio timestamp drift compensation
      10l
      fix first timestamp when decoding mpeg-ps
      better audio drift compensation
      1000l
      10l (fixes delayed frames during encoding)
      10l (fixes delayed frames during encoding)
      10l
      100l(audio sync) fix by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      bttv patches
      assert(0) -> retunrn -1
      100l fix by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      ugly missing lrintf workaround by ("Steven M. Schultz" <sms at 2bsd dot com>)
      dc_precission=11 fix
      h264 in mpeg-ts fix by (Måns Rullgård <mru at kth dot se>)
      overread fix
      spaces in the patch fix by (Nicolas Boos <nicolas.boos at wanadoo dot fr>)
      update version for next release
      update
      FAQ: '-f jpeg'
      field pic timestamp fix
      1000000l
      warn user if ms style codec tag is used
      play more broken files
      comment is no longer true
      libdts support by (Benjamin Zores <ben at geexbox dot org>)
      in/outfile options
      more info about why init failed
      all non trivial functions, structurs and their member variables should be documented
      simplify
      Doxygen example
      more comments
      moving Doxyfile from ffmpeg/libavcodec to ffmpeg
      image2 / image API cleanup phase-1
      support single images, remove copy/paste junk
      debugging symbols ...
      segfault fix
      10l and better MMX/SSE detection for VIA1000
      disable mov_read_elst
      mov_read_elst fix, based upon a patch by (Sebastien Bechet <s.bechet at av7 dot net>)
      raw rgb support
      make comments doxygen compatible
      seems noone knows what to do with bugreports against old ffmpeg versions
      multi frame duplication fix
      ffserver fixes by (Koos Vriezen <koos.vriezen at xs4all dot nl>)
      my experimental wavelet codec
      typo
      TIMER dummies for non-x86
      typo fix by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      libdts support improvements by (Aurelien Jacobs)
      argh
      better context model (0.2-1% lower bitrate)
      log2() & bsd
      realloc(NULL) fix
      use parent sample to predict significance & magnitude
      cleanup
      optimize quantizaton (about 3x faster)
      dont segfault without -strict -1
      1/0 fix
      use non zero qscale for svq1 test
      svq1 mv array size fix
      min rate should be 0 by default
      raise build number for the snow
      /usr/include -> /usr/local/include
      various subband encoders (all either worse or complicated so they are commented out)
      cleanup
      dont be so picky
      fix one file, will probably break some others but i cant find any samples which need this ...
      improved magnitude coding, 0.2% lower bitrate (foreman at 352x288 qscale 1 and 8)
      cleanup
      lossless support
      10l (segfault)
      segfault fix
      cleanup
      make dep fixes by (Måns Rullgård <mru at kth dot se>)
      10l (dont store 2 different things at the same place ...)
      give the user a chance to override codec_tag
      spelling fixes, closes bug #1005426
      cross compile fix by (allanc at chickenandporn dot com)
      user overrideable level & profile
      minor optimization
      Minor Patch for shared libs on Mac OSX by (Bill May <wmay at cisco dot com>)
      stack overflow
      10l (lrintf(a)+b vs. lrintf(a+b))
      get_bits(32) fix
      cleanup
      memleak found by ("wxWindows_study" <quicktime2008 at hotmail dot com>)
      big endian fix (untested)
      more untested bigendian fixes
      nonsense removial
      copy & paste bug
      4x4 SSE compare function
      lrintf emulation improvments
      edge bugfix
      typo fix by (Marcin 'Morgoth' Kurek <morgoth6 at box43 dot pl>)
      quad tree based motion compensation (currently only 16x16 & 8x8 OBMC blocks, but can be extended to other block sizes easily)
      yuv median encoding 10l fix
      50% faster decode_subband()
      46% faster decode_subband()
      5% faster decode_subband()
      11% faster decode_subband()
      merge decode_subband() and dequantize()
      10l
      cleanup
      10l (keyframes and context resets)
      forgotten parameter
      optional non spec compliant optimizations
      optional non spec compliant optimizations for mpeg1
      typo
      20% faster predict_plane()
      10000l fix and use more mmx2/3dnow code for mpeg4 qpel which has been written and commited long time ago but appearently never used, qpel motion compensation is 5% faster
      useless
      h264 luma motion compensation in mmx2/3dnow
      optimization
      optimization
      optimization
      10l (wrong coded_picture)
      note about our speed with replying to patches
      h264_qpel4_hv_lowpass_mmx2/3dnow
      h264_qpel8_hv_lowpass_mmx2/3dnow
      mess cleanup
      fail if SSE && !memalign && !memalign_hack
      get/set_sr_golomb() cleanup
      use pointer arithmetic in mc_block(), 25% faster
      fixed framerates
      10l (gcc 2.95 compilation)
      use h264 MC code if possible
      100l (forgot to set the 8x8 mc functions, so ME did use the mpeg ones ...)
      pre_cmp fix
      qprd + mpeg2 segfault fix
      comment fix
      changing size segfault fix
      various fixes for the main chunk demux loop
      per picture psnr
      10l
      make distclean
      warning the user about possible incompatibilities
      huffyuv writes to AVCodecContext.stats_out only once every 32 frames,
      passing an array or double precission parameters for the scaling function, instead of missusing a few bits of the flags
      av_rescale with user specified rounding
      check if the user specified timestamps are strictly monotone
      1/0 fix for files with size=0
      initial timestamp fix
      gcc 2.95 compile fix by (Jan Knutar <jknutar at nic dot fi>)
      8 -> FRAC_BITS
      last timestamp and -vsync 0 fix
      removing useless code
      write shorter headers if possible
      10l
      jpeg regression test
      segfault fix by (Kostya <cannonball at bw-team dot com>)
      imlib2 vhook rgba fix by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      merge predict_plane() with DWTELEM->8bit conversation (21% faster)
      100l
      unused
      trellis quantization for h263(p)
      H.264 max reference pictures fix by (Loren Merritt <lorenm at u dot washington dot edu>)
      assertion when playing AVC/H.264 streams fix by (Loren Merritt <lorenm at u dot washington dot edu>)
      some credits for all these patches
      1/2 resolution decoding
      10l (wrong context)
      1/4 resolution decoding
      low resolution decoding fix
      1/8 resolution decoding
      low resolution interlaced support
      lowres slice fix
      optimizing 4x4 idct
      porting optimizations from 4x4 dct to 8x8
      lowres slice fix
      lowres chroma fix
      set AVCodecContext.width/height to the picture width/height instead of the one stored in the bitstream (that only matters if lowres!=0)
      10l
      10l (segfault)
      lowres fixes for the parser
      lowres width/height cleanup 3rd try
      dont print "vop not coded" by default
      move h264 idct to its own file and call via function pointer in DspContext
      10l forgot to commit
      100l (signed vs. unsigend)
      pads stream
      fix assert
      less printf spam
      uninitalized variables
      dont just resample half of the data sometimes
      avoid using local arrays outside the block where they live ...
      10l (wrong variable)
      broken mess removial
      move packet interleaving function into AVOutputFormat, so it can be overriden easily instead of doing reordering twice if the muxer needs some other interleaving then dts based
      do not call (av_)abort()
      do not call (av_)abort()
      64bit and reading over the end of the array fixes
      move free() of AVStream priv data to av_write_trailer()
      simplify put_padding_packet()
      SCR timestamp fix try #1
      cleanup
      cleanup video buffer size
      pts/dts 100l fix
      remove wrong 33bit truncation of internal timestamps
      initial dts fix
      fix initial vbv buffer underflow
      user selectable packet_size and mux_rate
      10l pts vs. dts
      do not randomize unknown timestamps
      prevent multiple av_register_all()
      write error handling
      nonsense removial
      h.261 loop filter fix
      lowres needs EMU_EDGE
      lowres 4mv
      use native timebase for seeking
      AMR links
      100l (forgoten seeking functions)
      flags, rounding and cliping fix by (Nathan Kurz <nate at verse dot com>)
      generic cur_dts update
      fix assertion failure in case of timestamp discontinuities
      fix forcing of individual frame types
      support strictly enforcing gop size
      xvid fake divx version workaround
      user setable fourcc
      OPT_FLOAT
      user setable preload and max_mux_delay
      1000l (copy & paste)
      infinte recursion fix
      set correct timebase
      fix constrained parameters flag
      fixing demuxing of vc2_intro.avi
      add the long_name to the -formats output patch by (john Galloway <jrg at monsterbot dot com>) and (Måns Rullgård <mru at mru dot ath dot cx>)
      10l
      make filter size, int32/int16 and a few other things selectable at compiletime
      optimization
      make most resample filter parameters selectable at runtime
      100l (dont reset the sample position at the block end)
      user selectble cutoff frequency
      set keyframe flag at a more central place instead of in every demuxer for containers which only store intra only streams
      decode intra cabac mb_type cleanup
      cleanup of decode_cabac_residual()
      cbp_table cleanup
      fix CABAC+LOOP missmatch
      optimization
      optimzation
      dvr-ms video decoding support
      stdin detection fix
      mjpeg parser
      image2pipe fix
      type fix
      type fix
      optimization
      stride align cleanup
      YUV420P to YUV422 conversion routine by (Danny Laarmans <dlaarmans at androme dot com>)
      H.261 encoder by (Maarten Daniels <maarten dot daniels at luc dot ac dot be>)
      h.261 regression test
      suppress uninerresting output
      cleanup
      fix and cleanup h261 mb_type selection
      simplify
      simplify ff_h261_reorder_mb_index
      cleanup
      range coder
      CABAC -> range coder
      segfault fix
      remove debug stuff
      switch ffv1 from CABAC to the range coder
      switch ffv1 to the put/get_symbol() from snow
      remove weird offsets
      cbr audio muxing fix
      fix image stream copy
      doxy
      porting png support from -f image to -f image2
      prefer image2 over image
      10l (CONFIG_ZLIB)
      move p*m from image/lavf -> image2/lavc
      av_resample_compensate() doxy
      last packet size fix
      try to guess the codec for imag2pipe too
      pnm parser
      switch pnm from imagepipe to image2pipe
      merge pam_decode_frame() into pnm_decode_frame()
      automatic workaround for another padding bug
      allocate a few bytes more for extradata so the bitstream reader if its used by the codec for parsing extardata, doesnt read over the end
      raw h261 demuxing fix
      plane predicted non-interlacd rgb decodeing fix
      attribute_used
      disable redundant image formats, otherwise we wont get any bugreports about -f image2
      extradata and FF_INPUT_BUFFER_PADDING_SIZE doc
      overflow fix
      mv/mb reuse regression test
      loop hack
      fix parser
      reset ref_index
      rv20 demuxing fix
      rv20103001 support
      skip unknown streams instead of failing
      fixing rv20201002
      workaround apps which dont provided correctly padded streams
      parts of the dvd patch from ("Chris" <chris <at< garveycocker >dot< com> and Paul Curtis <pfc >at> terrapin <dot< com>)
      cvs policy
      fake regression errors caused by randomly choosen dct algo
      int / int64 fix by (Wolfram Gloger <wmglo @@@ dent:med:uni-muenchen:de>)
      cygwin fixes by (Sascha Sommer <saschasommer O freenet o de>)
      check num_entries too to avoid reading over the array end and dont check i>0 it must be >0 there
      fix for build on IRIX by (Michel Bardiaux {mbardiaux peaktime be})
      credits for Michel Bardiaux
      10l (encode bit buffer too small for 1920x1080 raw)
      10l (dont use codec_id for error simulation)
      fix some codec_id values so that new ones can be added at the end of the subsets without breaking binary compatibility
      max frames command line options
      fixing playback of ftp://ftp.mplayerhq.hu/MPlayer/incoming/blender_raw.avi
      raw rgb flip fix
      kill a annoying debug printf
      fifo_realloc()
      next try at fixing rv20 decoding
      fixing pipe seek bug
      use pts based duration/bitrate guessing code for mpeg-ts
      timestamp fix
      rv10 thread safety fix / rv10 ffplay fix
      various demuxer fixes
      try to recover from errors instead of failing fataly
      rm encoding fix
      seeking in rm
      seeking in rm 2nd try
      rv20 encoding
      rv20 regression test
      suppress 100 pages of missing enum value in switch spam
      frame skip support
      more flexible frame skip decission
      out of buffer access fix
      uppercase codec_tag=AVC1 support
      warn the user if the quantization code could overflow with the selected custom matrix and qp limits
      flush remaining data from parser at EOF
      another try at fixing rv20 ...
      x86 cpu capabilities detection rewrite / cleanup
      fix timestamp prediction for low_delay mpeg streams
      some sanity checks on what is muxed, invalid timestamps in mpeg are very common and lead to strange errors in the mpeg muxer otherwise
      try to set next_pts to a reasonable value in case of streamcopy
      quantizer overflow warning + AAN dct fix
      workaround some broken mpeg-ps timestamps
      try to guess the fps if they are variable
      broken avi / div by zero fix
      better motion_subsample_log2 docs
      10l (double free)
      stream copy block_align
      fix regression test failure
      guess fps for the variable fps case
      prefer integer fps if possible when guessing
      1/0l fix
      preroll is in ms
      fopen "r" -> "rb"
      r/w -> rb/wb
      motion_val doxy
      common.c -> bitstream.c (and the single non bitstream func -> utils.c)
      common.h -> common.h/bitstream.h
      ffm_init() for win32, so the regression tests dont fail there
      special case for filter_length==1
      10l
      faster and slightly less accurate nearest neighbor resampler
      WV1F support
      spaces in path fixes
      skip INDX chunks
      10l
      .{Y,U,V} image2 support
      disable yuv_image_format
      workaround missuse of url_feof()
      Patch for creating m2v files by (Fabrizio Gennari <fabrizio.ge tiscali it)
      fix decoding of http://mplayerhq.hu/~diego/problem.mov
      fixing decoding of http://mplayerhq.hu/~atmos4/ffrv20-crashing-atmos.rm
      RV30/RV40 demuxing (untested)
      print "D" instead of "P" for dropable frames in flv with -debug 1
      integer overflows, heap corruption
      10l (ret vs. ret>0)
      picture heap corruption fix
      fixing selftest
      various security fixes and precautionary checks
      dissallow sprintf
      sprintf->snprintf
      cvs policy entry about security
      disallow strcat
      10l (gcc 2.95 fix)
      never freed buffer fix
      wav timestamp truncation fix by (Wolfram Gloger <wmglo dent.med.uni-muenchen de>)
      timestamp truncation fix?
      10l (dont modify the input picture pts)
      assertion about bits statistic and minor bits stat fix
      simplify, null pointer, selftest
      simplify
      trying to fix the -ss bugs
      simplify
      x86_64 pointer typecast fix by (Martin Drab <drab kepler.fjfi.cvut cz>)
      svq3_get_se_golomb() fix
      div by zero aspect fix
      jpeg style yuv fixes
      1/0 fix
      count errors instead of printing 2^31
      support discarding of uninterresting packets
      fixing demuxing for short files where the framerate detection failed
      support discarding uninterresting packets
      simplify
      use clip_uint8()
      simplify deblock
      framerate check
      10l
      more detailed error messages
      avoid buf_size == 0 checks in every decoder
      simplify
      wmv3
      mention table duplications, note, please avoid duplicating tables!
      1000000000l typos
      dont use several 100 mb memory for a tiny 120 element table
      fixing many bugs in bitplane_decoding()
      norm-6 / diff-6 support (untested, parts of it just guessed as the spec is unclear)
      more 10l typos
      check norm6 vlc validity as there are some bit sequences which dont corespond to any codeword, the other vlc tables all seem to be huffman tables though
      dont duplicate decode012()
      10l (negative strides)
      make decode012() static inline
      dont predict missing timestamps if we lack the required information to do so
      require a few valid and equal mp3 headers for resync
      fixing missaligned memory accesses in fill_rectangle()
      quick and dirty PSNR check for audio
      dithering for the mpeg audio decoder
      10l (didnt test code after cvs up ...)
      100000l (typos in the #define for selecting the NORM6 method -> wasnt compiled, after fixing typos new method doesnt even compile -> swicth back to old until its fixed)
      10l (missing overflow check)
      more accurate dequant of near zero coefficients
      support 32bit output for the mpeg audio decoder
      10000l vbr mp3 fix
      signed 16bit support
      10l (array gets padded with 0 which is CODEC_ID_NONE -> parsers claim to support CODEC_ID_NONE)
      make keyframe default for AVFrames
      support comparing files with some posiion offset relative to each other
      compare audio files 16bit wise instead of 8bit
      optimizing imdct36()
      faster, simpler and more accurate l3_unscale()
      optimize antialias
      mb_lmin/max to limit the per mb quality for the ratecontrol independant from the frame limits
      optimizing imdct12
      merge a few shifts into the dequantization, faster and more accurate, lets hope it doesnt overflow
      overflows
      10l (use of deallocated memory)
      fixing an integer overflow, which could lead to overwriting the end of a malloced buffer by 8 bytes
      seperate bitstream encoding skip from PASS1 flag
      support skiping some bitstream encoding
      emulate (idiotic) MS behavior for inconsistant dimensions
      reversing the 1.52->1.53 change (handle fixed sample_size patch)
      cleanup
      set timestamps correctly
      regression test checksum change due to the mov header field and compressor name patch (movenc.c 1.24->1.25)
      max/avg bitrate change (movenc.c 1.25->1.26)
      max/avg bitrate change (movenc.c 1.25->1.26)
      dont kill the buffers while the decoder is using them
      checksum change due to (don't crash on null codec patch which changed the stored length of the compressor name, movenc.c 1.28->1.29)
      .wma
      audio duration fix?
      buffer overflow
      nonsense comment
      memleak fix
      buffer overflows
      switch MPEG4 fourcc to FMP4
      -strict -1 comment fix
      set supported pix_fmts
      watermark filter by (Marcus Engene <ffmpeg engene se>)
      shorten decoder by (Jeff Muizelaar <jrmuizel gmail com>)
      10l (shorten.o missing)
      ra288 demuxing support (doesnt really work, might be demuxer or decoder bug)
      set block align
      cast to uint32 as the table is uint32_t, this also worksaround a gcc or libc bug related to frexp()
      04-vp3 fix(by matthieu castet).patch
      mp4a fix based on some code by (Song.Itany inventec-inc com)
      mpeg-es bitrate parsing
      typo
      spam
      indeo3 for bigendian patch  by (demo-n <demo-n users.sourceforge net)
      group audio codec ids
      AVOption removial patch from (James A. Morrison >ja2morri csclub.uwaterloo ca>)
      fixing playback of xvid pretending to be divx (sample_vop_nc.avi)
      (10l) supported pix_fmts forgotten
      H.264 encoding with x264 by (Måns Rullgård <mru inprovide com>)
      optimize & simplify
      fix lib order and remove nonsense avcodec double linking
      keyframe & non keyframe index fixes
      use libavformats index system instead of the half duplicated mess in avidec.c
      set global_quality
      non interleaved avi support
      dont link with libavcodec twice on win32
      cleanup strh parser
      set global_quality
      sample_size fix
      reverse zero packet dissapearence "feature"
      zlib detection fix by (Gildas Bazin <gbazin altern org<)
      buffer overflow in sdp_probe() fix by (Gildas Bazin )gbazin altern org)
      more fine grained discarding of packets
      10l typo
      initalize has_b_frames
      alignment fix
      init has_b_frames for stream copy
      trying to fix a few minor values in the mp4/mov muxer
      change audio volume range so that the very slow division by 100 can be avoided
      fix time stored in mov
      support omiting various headers in mpeg4 as WMP seems to have difficulty with them
      create temporary file in current directory if it failed in /tmp
      copy extradata for streamcopy
      discard dummy packets before doing inapropriate checks on them and failing as a result
      segfault fix
      bypass internal buffer for "large" packets
      more checks, fixes assertion failure
      check mb/me_threshold range, fixes assertion failure
      make reverse[] non static
      replace custom bitstream reader with get_bits/vlc()
      fix psp muxing (probably this fix is wrong but its better then nothing)
      add Loïc Le Loarer
      motion estimation bitrate penalty compensation
      reverting 1.66 -> 1.67 change
      fix vorbis in avi
      snow regression test
      merge vertical lifting steps and a little cleanup
      fixing lossless snow
      10% faster unpack_coeffs
      fix QROOT != 8
      increasing precission of the quantization parameter
      testing cvslog
      testing cvslog
      rescale coefficients during IDWT, that way the lifting steps are much simpler and faster
      non 4:2:0 foobar.{Y,U,V} support
      optimizing unpack_coeffs()
      minor optimization
      replace complicated pointer dereference + index stuff by pointers in unpack_coeffs()
      store the number of runs to avoid storing the last run value
      read 32bit instead of 64bit to avoid overreading and missalignments
      various fixes related to the non alt_bitstream_reader
      avoid UPDATE_CACHE() in GET_RL_VLC() if not needed
      fixing demuxing of DV5700.asf
      optimizing non_intra ac coeff decode
      Indeo 2 decoder by (Kostya <> kostya.shishkov gmail com)
      buffer overflows
      remove duplicate clip to 8bit function
      simplify
      mpeg2_fast_decode_block_intra()
      avoid clearing block[] if its not used at all
      Video-Galileo_small.rm fix by (matthieu castet # castet.matthieu free fr)
      typo fix by ("Todd.Kirby" doubleshot pacbell net)
      fixing broken_rv20_1mb_.rm
      avoid unneeded clear_blocks()
      os2.diff by (Paul Smedley // paul smedley info)
      PIX_FMT_NONE and related fixes
      print pix_fmt if its known instead of if the raw codec is used
      fix nonsens timestamp calculation
      fix nonsense timestamp mess
      fix useless framerate messup
      switch to native time bases
      --enable/disable-codec / --disable-encoders (configure part only, no actual #ifdefs)
      typo
      #ifdefs around most register_avcodec
      also put the CONFIG_FOOBAR_EN/DECODER stuff in config.mak
      put most codecs under ifdefs
      replace local sensitive awk uppercaser with tr
      GCC4 fix by (Keenan Pepper (keenanpepper  gmail com)
      fix segfault with 'ffmpeg -i 1.avi -cmp 257 -4mv 2.avi'
      check for error
      always honor the user specified frame rate if set
      set time_base for variable fps too
      fix decoding of conformace streams AUD_MW_E.264 and BA1_FT_C.264 again (there where last correctly decoded with h264.c rev 1.96)
      prefer container time_base for frame duration guess
      division by zero fix
      default to YUV420P if none specified for rawvideo input
      better r_frame_rate guessing code
      reverse 1 hunk from 1.96->1.97, fixes decoding of (MR1_BT_A.h264 and MR1_MW_A.264)
      minor optimization of the h264 loop filter
      complain about mpeg4 limits only if codec_id == MPEG4
      overflow fix
      fix block corruption caused by clear_blocks() optimization
      fix PIC
      #defines for strict_std_compliance and split between inofficial extensions and non standarized things
      cast NULL to correct type
      change macros to inline functions
      optimization
      typo
      LE bitstream reader based upon a patch by (Balatoni Denes <dbalatoni programozo hu)
      support the LE reader, so it can be tested a little
      fixing decoding of AlanKay-245.asf
      b_frame_strategy sanity check
      some asserts()
      sanity check
      10l (overflows)
      10l (rescaling AV_NOPTS_VALUE)
      arm alignment fix
      typo found by Chengji Zhao
      10l (mb_type uninitalized and then changed and overwritten)
      alignment fix by (Gábor Kovács >picard demoscene hu)
      fix mimetype
      10l (%f vs. %Ld)
      fill missing reference pictures with something to avoid assertion failure
      preserve parse context
      round duration up in mov
      ffm.c rev 1.39+
      another try at decode_ref_pic_list_reordering()
      fix dequant matrix
      various 10l fixes for the loop filter
      some benchmarking code
      clear blocks after each idct instead of per picture
      fix last coeff
      disabling vp3 mmx&mmx2 idcts, they must be ported over to the lavc idct API, ill port the vp3 c idct
      remove permutation done to quant tables and then undone during idct
      porting vp3 idct over to lavc idct api
      fix idct permutation
      remove duplicate zigzag tables
      porting the mmx&sse2 (sse2 untested) vp3 idcts to the lavc idct API
      Vorbis decoder by (Balatoni Denes | dbalatoni programozo hu)
      SATURATE_U8 -> clip_uint8 (10% faster loop filter)
      some optimizations gcc should have done (10% faster loop filter)
      disabling buggy optimizations
      optimizing rle decoding loops a little
      reduce size of Vp3Fragment from 32byte to 16byte
      fix mpeg2 chroma_420_type
      use O(number of non zero coeffs) instead of O(number of coeffs) storage for the coefficient colleting/reordering
      10l (overwriting w/h with 0)
      disable broken loop filter optimization
      remove unused vp3 related function pointers
      support non interleaved mov files
      segfault fix
      update current_sample number for non split mode
      START/STOP_TIMER for ppc32 by (Luca Barbato: lu_zero, gentoo org)
      b frame strategy
      fix max b frames with b frame strategy 1
      av_find_stream_info() end check fixes
      libgsm warper by (Alban Bedel: albeu, free fr)
      AV_LOG_DEBUG is not a good default ...
      very tiny script to remove tabs and trailing whitespace from patches
      cleanup
      tr -d '\015'
      AVPacket.pos
      changing a few AV_LOG_DEBUG to AV_LOG_INFO
      Vorbis sse fix by (Balatoni Denes: dbalatoni, programozo hu)
      ilog() -> av_log2()
      faster float->short conversation
      optimizations
      trivial pre-C99 fix for rtp.c by (Wolfram Gloger: wmglo, dent med uni-muenchen de)
      maxdepth is supposed to be a constant, its faster if its too large then if its non constant
      more trivial optimizations
      decrease V_NB_BITS if possible
      add rounding bias before the horizontal idct (765->730 dezicyles on duron)
      avoid one transpose (730->680 dezicycles on duron)
      get rid of 2 movq (680 -> 658 dezicycles on duron)
      merge a few asm blocks so gcc cant unoptimize it (658->631 dezicycles on duron)
      100l for myself (breaking amd64)
      adding a few const
      one bugfix and a few gcc4 bug workaorunds by (Gianluigi Tiesi: mplayer, netfarm it)
      10l (segfault fix)
      a step toward CODEC_TYPE_DATA support
      fix stream class
      fix stream class
      remove some purely cosmetical hunks
      mpegts regression test
      typo fix by (lists: kenneth aafloy net)
      10l (dont read variables before they have been set)
      uint_fast64_t emulation by (Wolfram Gloger: wmglo, dent med uni-muenchen de)
      apiexample fix by (Chip: szarlada, freemail hu)
      set log level to AV_LOG_DEBUG if -debug ... is used
      make decoder a little bit more tolerant to missing NAL units
      typo
      AVStream.nb_frames and docs for r_frame_rate
      another chapter in the neverending mov demuxer cleanup
      *BSD bktr grab support by (Jacob Meuser: jakemsr, jakemsr com)
      tiny docs improvement which i apparebtly forgot to commit
      support changing in bitstream global headers into extradata style and back
      nut prefers extradata style global headers
      x86_64 + PIC fix by (Jacob Meuser: jakemsr, jakemsr com)
      better dont read the width/height at all, its random anyway
      width % 16 != 0 workaround by (Nicolas Plourde: nicolas plourde, gmail com>)
      keep decoding nal units after an error
      dont open and close codec at every call of try_decode_frame() as this is not only slow but also fails if the previous frame is needed for setting some parameters correctly
      check mb_y
      fixing first_mb_in_slice if frame_mbs_only_flag==0 && mb_aff==0
      support extracting sps/pps from bitstream and putting it in extradata
      RTCP_* conflict with live.com and they seem not to be used anywhere so commenting them out
      off by 1 error bugfix
      fix assertion failure
      .264
      av_seek_frame() docs timestamp clarification
      Here is the patch suggested by: unkaggregate, users sf net
      patch from http://www.freebsd.org/cgi/cvsweb.cgi/ports/multimedia/ffmpeg-devel/files/patch-libavcodec::x264.c
      remove -f singlejpeg as its identical to -f mjpeg
      segfault fix
      .m1v and .m2a (feature req #1178960)
      fix decoding of (broken) files with f_code=0
      error_resilience = FF_ER_CAREFULL
      fix infinite loop (suggested change by rjayne at convera dot com)
      fix segfault (bug #1165640)
      check len (should fix #1165694)
      ffmpeg version should be "CVS"
      check for CODEC_CAP_DELAY in audio decoders too
      skip_idct
      more fine grained skip_frame
      patches should not contain several unrelated changes
      yamaha adpcm nibbles in the wrong order fix by (Vidar Madsen: vidarino, gmail com)
      split string to avoid buffer overflow in native english speaking persons (fix suggested by The Wanderer)
      DVB subtitle decoder by (Ian Caulfield: imc25, cam ac uk)
      10l (forgot to commit this yesterday)
      make error concealment related options accessable from ffplays command line
      verify len field validity in mjpeg_decode_com()
      dont pre allocate uselessly large buffer and dont ignore FF_INPUT_BUFFER_PADDING_SIZE
      changing AVCodecContext codec -> *codec in AVStream so additions to AVCodecContext dont randomize AVStream and break binary compatibility
      ffserver segfault & other fatal failure fixes
      enabling ffserver regression tests
      portable IEEE float/double read/write functions
      portable IEEE float/double read/write functions
      extract duration if available
      kill duplicated get/put_be24()
      remove non portable get/put_be64_double()
      more non portable float parsing code ...
      minor user data reading bugfix
      10l (!= vs. >)
      mpeg1video m bit fix by (Luca Abeni: lucabe72, email it)
      fix user data parsing code so it suppors pre1 and cvs
      thread mess check for avcodec_open/close()
      yamaha adpcm regression test
      fixing colocated mv if colocated block is L1 predicted
      fixing colocated mv if colocated block is L1 predicted for the temporal direct case
      round pointer up to next packet_size multiple
      -ldl fix
      PIC for ia64
      motion vector limit for mpeg2
      check time_base.den for mpeg4
      try to check for nonsense time_base instead of setting fake one
      A.B.C versions
      workaround 'colocated mv if colocated block is L1 predicted' bug in x264
      fix assertion failure for mpeg2 encoding
      fix ffmpeg under mingw
      segfault fix
      mkstemp on win32 workaround by (Ivan Wong: email, ivanwong info)
      silly script to fix line wraped patches
      set pix_fmt
      set timebase correctly? this thing is always 15fps?
      a few checks, seems i forgot to commit this
      support fixing missing pts by parsing future frames
      parse pict_type for streams in avi
      automatically remove mess
      segfault fix
      reducing minimum supported picture size
      time_base description by Rich Felker | dalias, aerifal cx
      prevent assert failures and infinite loops with broken streams
      Compilation fixes part 1 patch by (Arvind R. and Burkhard Plaum, plaum, ipf uni-stuttgart de)
      regression tests for dv
      optimizations
      error concealment fix
      forgot to commit this after dv changes
      lowres support
      setting timebase in the encoder doesnt seem like a good idea
      add consts (this was in my local tree, dunno where it came from, probably forgoten from some const patch)
      simplify (d&a) and (d&~a) calculation, hint by skal
      print fourcc in a readable form if possible
      the user sets codec->time_base for video encding so she should also set it for audio encoding
      pts must be in time_base units
      fixme note
      sample_rate value is not always correct (is there anything in quicktime which is?) so try to guess it from time_scale
      AV_LOG_DEBUG vs. AV_LOG_ERROR
      set a few variables correctly
      AVOption first try
      a little script to turn the inside of a struct into a AVOption array
      a few more raw yuv fourccs
      AVOption enumeration support and some flags to classify AVOptions
      set time_base as soon as time_scale is known instead of in a unrelated chunck which might be missing
      dont be so picky with .mp4 ... fixes Mr&MrsSmith.mp4
      xvids mmx&mmx2 idcts
      select xvid idct for xvid files if possible
      remove unused table entries
      add xvid idcts to dct-test
      do not rescale AV_NOPTS_VALUE
      forward_ptr doesnt contain its own length anymore
      checksum doesnt cover the forward_ptr anymore
      AVOption API improvments
      flags and named constants with type checking of course for AVOption
      missing include noticed by g0th
      common questions people ask me all the time, lets hope this reduces the stuff in my inbox, otherwise ill train bogofilter on them ...
      checksum doesnt cover the forward_ptr anymore (seems i forgot to commit this ...)
      avoid stdio.h
      convert some options in ffmpeg.c to AVOptions
      converting a few more options to AVOptions
      clarify GPL patch acceptance policy
      dont include opt.h in avcodec.h
      the people want lower case ...
      print more time_base fps stuff if av_log level is at debug or above
      set pts for h.263 decoding
      dont print NULL
      doxygenize a few comments
      precautionary checks
      changing a few options to AVOptions
      replace a few mov + psrlq with pshufw, there are more cases which could benefit from this but they would require us to duplicate some functions ...
      extradata endianness comment
      allow NULL write_header() and write_trailer()
      per frame crc support
      frame skip fixes
      check MPV_frame_start() return value
      default_get_buffer() cleanup
      make selftest code compile again
      check block_max_depth (fixes crash)
      psnr calculation with threads>1 fix
      nan fps fix
      total == 0 seems like it should be valid (based upon patch by Steve Lhomme | slhomme divxcorp com)
      wrong sizes, part of [PATCH] from DivX, Part 9: bitstream crashes from (Steve Lhomme | slhomme divxcorp com)
      bitstream related fixes from [PATCH] from DivX, Part 9: bitstream crashes by (Steve Lhomme | slhomme divxcorp com)
      fix custom matrix permutation if xvid idct is autoselected and you have bad luck
      typo
      unfinished jpeg-ls codec (feel free to finish it ...)
      fix q=0 bug
      support printing compact qp histogram
      forgot
      use correct qp & lambda for ME on the second pass
      10l
      also autodetect xvid idct if bitexact is used
      improved lambda/qp guessing for motion estimation RD in first pass
      better f_code guessing in 2nd pass
      minor typo
      bidir refine support
      10l (segfault fix)
      slightly better mv_penalty table
      hmm did i mess the scaling of these up at some point?
      dont discard least significant bits
      b frame strategy 2
      10l (didnt remove option from old ones after converting to AVOption)
      support downscaling frames for dynamic b frame decission
      iterative overlapped block based motion estimation for snow
      10l
      4mv + iter ME 10l fix (still not bugfree but better)
      second try of improved lambda/qp guessing for motion estimation RD in first pass
      mem corruption fix
      av_freep() recommandition
      8x8 integer dct from x264 as cmp function (under CONFIG_GPL)
      cleanup
      10l (assertion failure)
      regression test for iterative ME & wavelet cmp functions
      use h264 MC functions for rectangular blocks too
      use h264 MC functions for 2xX Xx2 blocks in snow too
      fixing second last time François Revol applies a patch without running the regression tests, next time his account is disabled
      bugs found by valgrind
      split snow53 test out
      pts fix by (Bryan Mayland / bmayland O leoninedev o com)
      audio in dvr-ms demuxing support by (John Donaghy <johnfdonaghy gmail com<)
      >4gb fixes by (Dirk Musfeldt | d.musfeldt meilenstein de)
      scale video for snow iterative me down so the egression test is faster
      xvid ratecontrol support
      10l forgot #ifdef CONFIG_XVID
      try avi video name -> codec_id mapping if the mov one has no match
      avoid using non constant fields of AVCodecContext in avi muxer
      fixing bframe strategy 2
      JPEG-LS codec by (Kostya | kostya.forjunk gmail com)
      list of supported pix_fmts
      dynamically increase probe buffer until format is detected
      os2 thread, network and freetype2 support by (Paul Smedley | paulat a t smedleydot d o t info)
      improve mpeg1/2-es detection
      AIFF format support by (Patrick Guimond <patg a.t patg d.o.t homeunix d.o.t org)
      jpeg-ls regression test
      output last ac3 frame and simplify
      sanity checks for seeking
      improve fps guessing code
      simplify timebase if possible
      change sws sharpen filter a little
      fix regression tests
      3rd try for mpeg-es probe ...
      faster find_startcode()
      simplify
      1 goto 1 fix
      fix memleak, based upon a patch by Wolfram Gloger
      replace probe() by one similar to MPEG-ES
      fix segfault on broken streams
      mpeg-ps probe 2nd try
      Fix for url_fclose() being called on an already closed file based on a patch by (<Colin Ward> hitman codehq org)
      stsd fix by (Baptiste COUDURIER )baptiste.coudurier smartjog com)
      10l for myself
      fix demuxing of Video006.3gp
      useless code removal by (Baptiste COUDURIER <baptiste.coudurier smartjog com)
      mov fourccs for handling FCP files by (Baptiste COUDURIER )(baptiste.coudurier smartjog  com)
      RMP4 ("xvid fork") decoding support
      print a big warning if we mess up and run out of space ...
      fixing dv_guess_qnos()
      dont lowwer qnos of more blocks then needed
      10l (forgot EOBs)
      replace for loop with 2 asserts()
      remove wrong assert()
      v4l2 maintainer
      replace if by assert as it should never by true
      ODML index parsing support
      add size to AVIndex
      add movi_list offset into index
      fix demuxing of XviD_with_3_AAC-HE_audio_streams.avi
      10l forgot to set packet_size
      support seeking in RenderAvi.avi (audio stream == single huge chunk)
      fixing av sync in videotest.avi (index doesnt match chunks, header doesnt indicate that)
      fix decoding of mpeg2-trunc.vob
      cleanup IMA-ADPCM WAV decoder
      fixing demuxing of minorityreport.AVI
      dwstart support
      timebase fix
      indention
      fixing total_frames
      fix infinite loop
      fix segfault with BLUR8.AVI
      PAL8 support (fixed BLUR8.AVI)
      generic crc calculation code
      switch ac3enc to av_crc
      switch flac to av_crc
      unused variable
      very minor optimization
      fixing uninitalized dc_pred_dir
      add assert to clarify that we know what we are doing
      fixing out of array access (only cosmetic, this should never have had a end user vissible effect)
      cleanup picture / debug output code
      out of array read fix
      out of array read fix
      asserts to check if assumed conditions really are true
      fix %8 != 0 w/h
      av_free -> av_freep
      simplify AVFormatParameters NULL checks
      some mov files have invalid pts so we need to consider these pts too in calculating the timabase
      reset interlaced_dct flag
      fix coverity warning CID: 255 (uninitalized variable used to build tables which arent used, well there is a slight change of a FPU exception maybe ...)
      fixing coverity warnig CID: 253 (reading over the end of the aspect_ratio array for illegal aspect ratios)
      dont pass intra flag if its always 0
      extract duration
      segfault fix
      fixing bquant_offset
      fix frametypes with xvid_rc
      various fixes for xvid_rc
      fixing interlaced direct mode field select values
      provide more info on which streams lack codec parameters
      fixing current_sample vs. next_sample mess
      fix aspect ratio
      cleanup codec_id mapping (untested)
      fixing timebase
      dont set random timestamps if they are not known
      optimizations by arpi
      minor optimization of decode_block()
      avoid i++ in decode_block()
      avoid if(i>=63) check at the end of decode_block()
      fix gcc 2.95 compilation
      dont forget table_size in the decode_frame return value
      reverting 1.31->1.33
      avoid hard to predict branch (idea by arpi)
      + (sign&1) -> -sign (1 instruction less)
      do UPDATE_CACHE only when needed (which should be VERY rare)
      simplify loop a little
      remove useless code
      get_xbits() optimization
      remove mpegvideo.c img resample dependancy
      the PSP rejects video with too high bitrates so lets claim they arent too high ...
      little h.264 related hunk from diff between cvs and mobile hackerz
      ffv1 and ffvhuff havnt changed since a long time and noone proposed any
      fix "concealing 0 DC, 0 AC, 0 MV errors" bug
      avoid AVInteger usage in av_rescale
      missmatch control for mpeg2 intra dequantization if bitexact=1
      use previous qscale for intra_dc_threshold check
      <= vs. >= 10l bug
      extend range used for testing (larger range should work too but testing code overflows)
      d10 decoding fix
      removing esp usage
      make zero motion vector threshold user setable
      add outcommented chunk of code to handle stuffing MBs at the end of slices (IMHO the standard doesnt allow this and there are no real world files which need it)
      allow insane gop sizes if the user doesnt care about playable streams (doesnt want compliance to any standard)
      decoders should (IMHO) not duplicate frames, the application can do this if it wants
      cosmetics (initalize variables where they are used)
      sanity checks, some might have been exploitable ...
      sanity checks some might have been exploitable
      simplify
      URLProtocol faq
      simplify meta tag writing code
      dont write LAVF_ID if bitexact and any meta tag is set
      dont store mov style meta tags in mp4 (fixes ffmpeg -title + gtkpod)
      fixing title on PSP (0x60 != 'a') someone needs a 10x10km ascii table ...
      ipod & psp encoding guide
      working H.264 + AAC in mp4 encoding example for the psp
      suggest solution for 2 fixmies
      cbp must be >0 here or we are in troubble
      typo me thinks ...
      fabs is better then abs for floats ...
      improve selftest
      drop stuff which seems to have been finished
      random ideas
      support doing motion estimation and compensation without any residual transform or coding
      fix probe of high bitrate mpeg2video
      dont load the index if we cant seek back
      10l to whoever wrote this
      vfr mov/mp4 todo item
      simplify
      first rudimentary version of (Justin Ruggles  jruggle earthlink net) flac encoder
      flac regression test
      simplify & optimize things a little
      10l
      stereo decorrelation support by (Justin Ruggles  jruggle earthlink net>)
      dont read over the end of a data chunk and at the end search for the next
      md5 support
      new files for the CAVS decoder by (Stefan Gehrer <stefan gehrer gmx de)
      simplify
      simplify
      dont use memcpy for copying structs
      simplify compute_autocorr
      simplify
      3 lines while -> 1 line for loop
      CAVS decoder by (Stefan Gehrer  stefan.gehrer gmx.de)
      more maintainers, if someone doesnt want to maintain something just say so and ill remove you again
      simplify
      i wont split h263.c, too much work, and not important at all ...
      minor simplificationn
      check cbp for validity, avoids possible out of array reads / segfaults
      simplify intra prediction mode decoding
      return -1 on bitstream errors instead of continuing, as the following stuff almost always will be useless until the next startcode
      add rounding bias earlier in the idct, avoids many adds
      vbr requires version=1 and samplesize=0 per QT spec
      check for unparsed AMR input
      100l to whoever write this
      AVBitStreamFilter (some thingy which can modify the bitstream like add or remove global headers or change the headers or ...)
      noise bitstream filter
      increasing version for AVBitStreamFilters
      unresponsive maintainer policy (its pretty much what has always been done in practice anyway)
      table contains vlc with 19 bits and 8*2 < 19
      seems i forgot to commit this simplification ...
      moving utf8 reading function to libavutil
      put the code which is specific for the large crc table under #ifndef CONFIG_SMALL
      set CONFIG_SMALL if user wants small ffmpeg
      remove STATS code (probably hasnt been used for years ..., and its not completely clear what it was good for anyway)
      cosmetic
      simplify
      avoid redundant local variables
      simplify
      dither lpc cpeffs
      linear least squares solver using cholesky factorization
      adding a \ so we can add more lines without having to do a "cosmetic" change in the previous line
      unneeded #include
      optionally (use_lpc=2) support Cholesky factorization for finding the lpc coeficients
      cleanup
      calculate all coefficients for several orders during cholesky factorization, the resulting coefficients are not strictly optimal though as there is a small difference in the autocorrelation matrixes which is ignored for the smaller orders
      log prediction order search method
      cleanup
      check how many bits are left after decoding extradata
      some fixmes if someone ever wants to implement theora support correctly ...
      useless memcpy
      use RSHIFT
      replace *_dequant (2 intra, 1 inter) by qmat (3 intra, 3 inter) (needed for theora according to spec)
      attempt to implement xiphs useless and stupid quantization matrix mess
      100l for me
      parse all QIS values (we still ignore them though ...)
      add header info to flv format based on a patch by (Philipp Klaus >ffmpegdevel ad pylonsoft mot ch )
      simplify b64_encode()
      simplify
      disable the vp3 mmx and sse2 idcts, their output doesnt match the c idct (tested with -f crc) and the theora spec does not allow different idcts not to mention the difference is quite vissible ...
      fix loop filter
      simplify
      iabs -> ABS
      simplification
      more simplifications
      simplify
      remove a few useless casts and avoid the useless t1/t2 variables
      remove 166 useless underscores, and make this file valid c code as a result
      avoid 4 +
      simpler, smaller and faster
      for floats fabs is better & smaller then ABS
      simplify
      fix regression failure after adler simplification
      removing redundant mess next time we break compatiility
      update
      disable libtheora, this change leaves all the libtheora related code in svn for now but just disables it in configure
      vertical scaler with accurate rounding, some people on doom9 can see +-1 errors
      a tiny bit of cleanup
      video filter todo
      set supported pixel formats correctly
      things to take care of for the next soc
      floating point "emulation" code
      change vorbis_inverse_coupling_sse2() so it works on sse1 cpus
      both timestamps are dts, (checked trailer.gxf, spec is unclear)
      timestamp debugging av_logs()
      dont be too picky about timestampsbeing wrong if the destination container is without timestamps and raw of the raw video / raw audio sort
      convert  vector_fmul_reverse_sse2 and vector_fmul_add_add_sse2 to sse
      mailinglist
      align av_realloc()
      trying to fix av_realloc()
      fix h261 parsing bug
      fix codec timebase and timestamps
      insufficient alignment
      cleanup
      revert aligned realloc() changesm this should be identical to r5784
      av_free* cleanup
      simplify
      fix probing of 02-Penguin.flac
      audio format conversion
      SWAP
      cosmetics
      attribute_deprecated
      cleanup PixelFormat a little
      compatibility warning
      reorder butterflies to reduce memory load/store
      replace a few MULL by MULH
      replace MULL by MULH + shift in dct32() (code is very slightly faster, and the compiler should be able to optimize the shift away on 64bit archs)
      make high precission synth filter 3 times faster on x86
      indent preprocessor directives
      replacing MULH by asm for x86
      replace MULL with asm too, no significnat speedup but its probably better to not take any chances, some versions of gcc will almost certainly mess it up too if they can
      MUL* for ARM code based on a patch by Siarhei Siamashka
      total_size should be 64bit (1 hunk of the asf seeking patch from DrDivx / Steve Lhomme)
      read and use index (based on a patch by John Donaghy on the 23-03-2006 '[Ffmpeg-devel] dvr-ms seek help request')
      fix fallback if theres no index
      simplify
      use memset() instead of slow c loops
      optimizing high freq decode
      faster handling of zero coeffs in the low freq decode
      dont copy the whole GetBitContext around
      simplify
      optimize reorder_block() though this function seems to be executed too rarely for this to make much difference
      fix non ALT bitstream reader
      avoid a few if(x==0) checks in the inner loops
      10l typo
      maybe fixing the segfault on ARM
      replace l3_unscale(1, exponents[pos]) by LUT
      try to avoid l3_unscale() by using a table
      reduce size of vlc table, thats slightly faster here
      avoid reading exponent twice
      move the +400 out of the innermost loop
      move variable declarations so that their scope is smaller
      avoid branch
      support some totally broken ODML variant
      prevent bitstream reader to be overriden
      move align_get_bits() to .h to avoid conflicts between different bitstream readers in different codecs
      2nd try for a skip_bits_long() for the A32 reader
      3rd try :)
      make A32 reader align its ptr during init no matter what missaligned mess is given to it
      2nd try of skip_bits_long() for the ALT reader
      remove duplicated parser, people who pass random gibblets of the bitstream into the decoder will have to pass it through a AVParser like for all other codecs too
      10l found by valgrind
      another silly typo which surprisingly neither valgrind nor cmp of the file output cought
      make mpa_decode_header() 10 times faster
      dont copy frame if the whole mp1/2/3 frame is available in one piece in the input
      unused var
      check the validity of the amount of the remaining bytes in the bitsteam before memcpy
      fix low bitrate mp3
      overread AV_LOG_ERROR -> AV_LOG_DEBUG
      fix buffer handling
      make the C code of the swscaler which i wrote LGPL
      clarify
      fix b pyramid in mp4 muxing if no dts are provided to the muxer
      print warnings if
      seems the output of that code is never used
      ignore index parameter to ignore the ODML index in avi
      fix sound of SampleVideo.mov (mp2 in mov)
      print more correct error messges
      AVPROBE_SCORE_MAX doc
      10l typo
      support SI and some non SI prefixes
      K prefix
      use correct rounding functions
      set supported pixel formats for a few encoders (fixes crashes with grayscale)
      reduce scores if the mp3 frames dont start from the begin of the file (fixes flv deteted as mp3 issues)
      1000l copy and paste typo
      simplify url_fseek()
      segfault fix
      cosmetic
      mpeg1 b_strategy=2 segfault fix
      simplify seek() failure check
      support forward seeking in non seekable streams
      dont ignore MAX_READ_SIZE
      reindent
      drop broken telecine detection code
      fix r_frame_rate init code for groundhog.vob
      fix decoding of bastard1.rm (rv ver=0x10003001 ver0=0x8000000)
      fix decoding of bastard3.rm (ver=0x10003001 ver0=0x9000000)
      prevent infinite loop with VORC012.WMA
      fix rv20 b frames (broke since r5264 Mon Apr 3 07:52:24 2006 UTC 2x faster ff_mpeg4_set_direct_mv)
      simplify
      support resolution switching, fixes rv20errors.rm
      big endian fix
      ignore sub_id
      dont set sub_id as its completly redundant and silly
      seems only 3 bits are used for the number of resolutions, some rv30 files have this set to 10 even though there are just 2 resolutions in the extradata
      move things from my todo to the unassigned one so my lazyness doesnt prevent others from working on them
      marking AVPaletteControl as deprecated due to thread saftey amongth other issues (it doesnt work if theres any delay between demuxer and decoder)
      palette (if we memcpy it into AVFrame) must be uint32_t
      slightly saner indention
      1000000l for lu_zero (commiting untested? code)
      fix r_frame_rate detection in mpeg4 streams with non integer fps
      remove confusing divx note
      fix buffer underflows by reencoding the current frame with a higher QP
      fix decoding of ED-small-sample-file.ogg
      prevent >16bit tags to be truncated and used in creating .wav files
      flac fourcc (fixes mplayer -demuxer 35 with flac)
      several x86 renorm_cabac_decoder_once optimizations
      () 10l
      refill cabac variables in 16bit steps, 3% faster get_cabac()
      renorm_cabac_decoder_once START/STOP_TIMER scores for athlon
      disable benchmarking code
      5% faster get_cabac()
      branchless renormalization (1% faster get_cabac) old branchless renormalization wasnt faster because gcc was scared of the shift variable (missusing bit variable now)
      cleanup
      decrease ff_h264_norm_shift[] size
      remove bytestream_end checks, seems to work fine without them and the bitstream reader doesnt check for the end either
      always add -fomit-frame-pointer if optimizations are enabled
      first try of a handwritten get_cabac() for x86, this is 10-20% faster on P3 depening on if you try to subtract the START/STOP_TIMER overhead
      10l bugfix for some disabled code
      drop failed attempt to optimize *state= c->mps_state[s];
      move outcommented START/STOP_TIMER to a hopefully better place for benchmarking ...
      optimize branchless C CABAC decoder
      x86 branchless cabac decoder
      reverse remainder of the failed attempt to optimize *state=c->mps_state[s]
      10l
      on the P4 inc needs twice as much time a add
      reading 8bit mem into a 8bit register needs 2 uops on P4, 8bit->32bit with zero extension needs just 1
      replace a few and/sub/... by cmov
      move the &1 out of the asm so gcc can optimize it away in inlined cases (yes this is slightly faster)
      moving another bit&1 out, this is as fast as with it in there, but it makes more sense with it outside of the loop
      return cleanup
      factorize (untested)
      set duration and timebase (based on a patch by Simon Morlat simon.morlat linphone org)
      factorze another 2 lines
      more factoriztaion and a FIXME note
      remove dummy amr_read_close()
      unused #define
      doxy
      enable CMOV_IS_FAST as its faster or equal speed on every cpu (duron, athlon, PM, P3) from which ive seen benchmarks, it might be slower on P4 but noone has posted benchmarks ...
      make lps_range a global table its constant anyway (saves 1 addition for accessing it)
      10l
      make state transition tables global as they are constant and the code is slightly faster that way
      use ecx instead of cl (no speed change on P3 but might avoid partial register stalls on some cpus)
      write cabac low and range variables as early as possible to prevent stalls from reading them before they where written, the P4 is said to disslike that alot, on P3 its 2% faster (START/STOP_TIMER over decode_residual)
      moving lps state transition code a little up in the branched asm code (1% faster on P3)
      slightly faster on P3 slightly slower on athlon and probably faster on P4
      int->uint8_t for 2 tables
      prevent "mb level" get_cabac() calls from being inlined (3% faster decode_mb_cabac() on P3)
      10l int vs uint8_t
      -1 is a valid return value in ratecontrol_1pass() -> 100l for takis
      add "memory" to the clobber list we change memory so we need it, this also fixes some problems with gcc svn
      dehack *ps_state indexing in the branchless decoder
      simplify escape decoding
      10l CODEC vs. FORMAT
      shift CABACContext.range right, this reduces the number of shifts needed in get_cabac() and is slightly faster on P3 (and should be much faster on P4 as the P4 except the more recent variants lacks an integer shifter and so  shifts have ~10 times longer latency then simple operations like adds)
      this code will not work with PIC as it needs 7 registers and gcc doesnt support that in PIC
      10l
      fix !CMOV_IS_FAST case (iam not really happy with the fix but i didnt come up with a better one quickly)
      mod()
      asin() acos() atan()
      fix -a^b which was interpreted as (-a)^b
      cosmetic (%%eax->%0)
      x86 asm version of the decode significance loop (not 8x8) of decode_residual() 5% faster decode_residual() on P3
      copy cabac context onto the stack for the c code in decode_cabac_residual() (15% faster on P3 but still slower then the current asm)
      optimize sign decoding code in decode_residual()
      dont mess with ByteIOContext internal variables (fixes showdown2.mov)
      1 variable less in decode_cabac_residual() no speed change (but less vars means less things gcc can load and store redundantly)
      slightly faster decode_cabac_mb_mvd()
      unused variable
      factorize get_cabac asm (0.5% slower but its much cleaner)
      reordering instructions a little in decode_significance_x86() -> 2 instructions less / 1% faster decode_residual on P3
      1 instruction less
      move memory reading ByteIOContext init from mov.c to avobuf.c
      another instruction less in decode_significance_x86() -> 1% faster ion P3
      3 more TODO items for volunteers
      decode_significance_8x8_x86()
      prevent infinite loop
      set v_edge_pos correctly for field picture mpeg2
      merge 8 and 16bit functions, they are identical except the data type
      common subexpression elimination
      FFMAX
      factorize update state out
      remove map variable
      reverse my last change, it was wrong and broken
      simplify
      factorize AVEvalExpr alloc and init
      minor simplification
      support seperating expressons by ;
      add support for while() loops again ugly syntax while(condition, statements) but very simple implementation
      use lambda/lmax instead of qscale/qmax as end check for vbv retry
      prevent overflow if lambda2 is "too" large
      lambda* should be unsigned
      CANDIDATE_MB_TYPE_INTER heuristic doesnt work at really low quality where the distortion becomes less relevant then the overhead of intra blocks
      update copyright year, it looks odd otherwise :)
      avoid random noise on damaged frames
      dont set the sampling rate just because 1 mp3 packet header says so (fixes playback speed on some old mencoder generated avis which where then dumped to mp3)
      better timebase selection heuristic for video stream copy
      fix 1/0 b frames + CBR bug
      merge clip() into the alpha/beta/tc0 tables (10% faster filter_mb_fast() on P3)
      10l (forgot ALT_SCAN)
      new I/P frame selection heuristic
      optimize H264_DEBLOCK_P0_Q0
      1 instruction less
      preempt possible overflow
      2 instructions less in h264_loop_filter_luma_mmx2()
      move luma tc0 related init into asm
      comment about failed optimization
      2 instructions less (same speed)
      merging mov & and (no speedchange)
      ensure alignment (no speed change)
      doxy
      2nd try for -vcodec copy timebase selection last one choose 1fps due to broken opendivx stream in avi (avi had the correct 15fps)
      utf8 title support for psp (untested)
      100l for me, iam being to stupid to read the mpeg4 spec / eat stuffing mbs a the end of the picture
      fix missdetection of mpeg-ps (ps2_interdite.mpg)
      10000l to myself, fixing regression test failure
      16bit in avi is RGB555
      mp3 header (de)compression bitstream filter
      also set strict_std_compliance for stream copy, its needed for the bitstream filters
      store a identifer and the first header in extradata
      zero private bits
      libnut uses "MP3 " for mp3
      updating nut demuxer to latest spec
      disable nut regression test until we have a working muxer
      nut simple seeking first try
      dont do startcode search from last syncpoint if there are no errors
      simplify
      fix timestamp rounding in syncpoints
      fix checksum check (was my bug not libnut)
      more verbose r_frame_rate definition
      AVL tree
      improve enumerate so arbitrary ranges can be enumerated quickly
      split av_seek_frame_binary() so the code becomes idependant of AVInputFormat and AVIndex
      0.5l
      index less seeking in O(log n) time
      seeking with index
      indention
      call AVCodec.flush() for audio too
      more checks and error messages
      reset internal buffer during seeking
      making decoding more error robust
      remove skip spam
      resync from last syncpoint
      check for frames with 2*size > max_dist and no crc
      detect a few more errors (fixes libmp3-bug.avi again)
      removing unused code
      remove unused variables
      skip non keyframes after seeking between syncpoint and the first keyframe
      check max_pts_distance
      remove broken decoder locking code
      10l forgot (skip non keyframes after seeking between syncpoint and the first keyframe)
      fixing the lamest bug in swscale, all the rgb/bgr->* code did 2x2 downsampling for chroma, it should just be 2x1 (the rest of the code also belived its 2x1 ...)
      forgotten 2 converters (yeah svn up, svn di svn ci isnt enough i should actually look at the code after svn up not just the diff ...)
      fix support for avis with sample_size > packet size
      testing all dquant for qp_rd fixes the qp runaway to 31 bug
      100l qprd fix 2nd try
      infiinit loop fix for CBR + adaptive quant
      100l infinite loop fix
      fix b frames + -mbd rd without -psnr (yes that was seriously buggy and probably hasnt been noticed as everyone who tested stuff used -psnr)
      fix b frames + -mbd rd without -psnr (yes that was seriously buggy and probably hasnt been noticed as everyone who tested stuff used -psnr)
      PIC fix
      return optimal fraction
      fix overflow and remove wrong comment
      doxy
      1 step diamond search around the median MV predictor
      cosmetic (prevent name clashes of variables in a macro with surrounding code)
      UMH support (dia_size=-1 activates it)
      typo in umh
      remove pretty much useless skiping of some predictors
      hexagon search
      dia_size=-1 -> funny_diamond_search()
      fix indention
      make hex and large 2 small diamond search half their size instead or -1 if the initial size was a power of 2
      10l umh search needs hex size 2 now after the last change
      fix compilation
      ff_check_alignment to warn the user about a missaligned stack
      better warning
      set flags correctly in write header instead of seeking back and fixing it in write_trailer()
      get rid of AVFMTCTX_NOHEADER, create streams in read_header()
      amf_get_string() by Allan Hsu  allan aat counterpop doot net
      cosmetics
      use amf_get_string()
      move duration finding code into read_packet() so it can be skiped if duration has already been set
      fix playback of mp3_bits_left.avi
      rewrite r_frame_rate selectioon code again
      gcc 2.95 fix
      mv overflow in map fix (untested except regression tests)
      the next r_frame_rate try ...
      indent
      limit x/ymin/max to me_range
      warn user if the selected diamond size is larger then the motion estimation score cache size
      10l
      minor doc fix
      dont favor the zero MV if mv0 is used (psnr per bitrate gains ranging from 0 to 0.14, most are <=0.04 though)
      fix special casing of the 0,0 MV for b frames (slight PSNR/bitrate gain)
      try direct mode MB after QPRD, slight PSNR/bitrate gain if b frames + QPRD
      QPRD & adaptive quantization regression tests
      dont randomly dissallow direct MBs
      dont randomly disallow intr4v in adaptive quant
      skip motion estimation and encoding of non direct-0,0 MBs if the next MB is skiped (mpeg4 doesnt allow such MBs and in the past we did ME and encoding until at the end we droped them, so this should be faster though i didnt benchmark it, benchmark welcome)
      i think this if() is useless lets abuse our users to test this ;)
      simplify
      use qscale_table in adaptive quant
      simplify
      FIXME has already been fixed ages ago
      10l
      test DIRECT{0,0} too if flag mv0 is used
      skiprd
      skiprd doxygen comment from avcodec.h -> AVOption comment
      do not base frame rate guess on just one duration
      ignore duration of the first 2 frames in the max-time break check
      1st attempt to fix width/height
      remove code duplication
      10l (fix segfault with audio only files)
      cosmetic (move video_open() up to avoid prototype)
      check x/y validity
      dont use SDL_WM_ToggleFullScreen() as that doesnt seem to work reliably if a SDL_SetVideoMode() is called quickly afterwards
      remember the size after switching back from fullscreen
      revert regression test breaking change
      seekless filesize retrieving support in 7 lines of code, also doesnt break compatibility
      fix indention of previous commit
      AVSEEK_SIZE doxy
      some quick encoding tips
      store a few values in the header as difference to the last
      nth try of getting the frame rate detecton working
      indention
      missdetected a broken mp3
      dead code removial
      snow specification
      maybe fixing a segfault
      fix av_reduce() with things like 1/0 and 0/0
      fix 16 CABAC_BIT * 10l bug introduced in r6578
      replace a few hardcoded numbers with their correct named ones
      some interlaced direct sanity checks
      aes support (unfinished)
      simplify sbox init
      memleak
      cosmetic
      stealing reimars XOR_BLOCK
      stealing multbl trick from reimar
      copyright fix
      seems the larger rcon values are never accessed
      memcpy
      cosmetics
      simplify round_key generation by writing over the end but ensuring that theres some irrelevant stuff afterwards
      cosmetic
      ascii
      init multbl
      benchmark
      simplify special round
      fix CONFIG_SMALL case
      1 test vector
      merge sbox with mix on the encryption side
      merge inv_sbox into inv_mix
      unused stuff removial
      remove no longer corret comment
      dont use encode tables for decoding
      make decode look more like encode
      remove duplicate round_key
      indent
      merge encrypt and decrypt so the source is simpler and the compiler can choose with inlining if it wants speed or small size
      unused  think
      unneeded
      seems i didnt copy anything from aes128.c so this should be more correct ?
      fix 2 warnings
      simplify
      revert simplification (broke 192 bit keys)
      simplify selftest
      simplify multiply table init
      simplify
      er and strict also apply to audio theoretically
      avcodec_decode_audio2()
      the missing header ...
      av_aes_init()
      cosmetic
      replace / by >> this makes the .o ~150 bytes smaller
      check the last? entry written for detecting already initalized tables
      remove dependancy on *malloc()
      move #include log.h into the selftesting code where it belongs
      make aes_en/decrypt() static until we decided on the public API
      merge shift into mix
      fixing segfault with http://sam.zoy.org/zzuf/lol-ffplay.avi
      fix segfault with http://sam.zoy.org/zzuf/lol-ffplay.mpg and http://sam.zoy.org/zzuf/lol-ffplay.m2v
      fix segfault with http://sam.zoy.org/zzuf/lol-ffplay.wmv
      unused variable
      remove useless parameter
      use 2 state arrays so that fewer temporary variables are needed
      change subshift a little, this reduces the object size a litlle
      give crypt a src and dst
      avoid memcpy()
      cbc support
      public API
      10l typo
      30byte smaller object file
      fix CONFIG_SMALL again
      set data_size to 0 so that in case we return without setting it nothing funny can happen
      10l (aes_crypt -> av_aes_crypt)
      seperate redo (we dont handle this possibly valid packet) from
      reduce len checks, simplifies code
      cosmetic
      simplify pts/dts reading
      simplify
      simplify
      simplify
      remove near duplicate function
      simplify av_fifo_realloc()
      better to set things to NULL instead of random in case of out of mem
      avoid code duplication
      doxygenize
      simpler branch structure in init (16 bytes smaller object file)
      change while loops to do-while as the condition is true the first time and the check just wastes cpu cycles
      uint16->uint8
      checking bitstream values and other related changes
      fix indention (dunno why it was messed up ...)
      add codec_id <-> codec_tag tables to AVIn/OutputFormat
      export a few more codec_tag-codec_id tables
      get rid of the [4] limitation of codec tag lists
      this is wrong but it was that way before the AVCodecTag change, only reason why it didnt broke regressions was that the table wasnt used
      set codec_tag from codec_id if its not set and can be set based on the AVCodecTag tables
      simplify
      remove now useless codec_tag setting code in muxers
      the context is supposed to be freed in the generic code in utils.c
      simplify
      fix msmpeg4v3 tag in asf
      simplify GUIDs
      remove sensless casts
      useless &0x80
      get rid of a few idiotic asserts()
      these variables seem to be unused on the demuxer side so theres no sense in setting them
      cosmetic
      adding a few checks to the audio packet descrambling, this should hopefully catch all related out of array accesses
      remove a few unused and duplicated variables
      simplify
      get rid of some sensless malloc/free
      cosmetics (aligne stuff nicely)
      doxygenize comments
      10l
      more correct packet alignment check and correction
      move frame header reading into its own function
      check fragment offset and size
      replace nonsense code by something which makes at least some sense
      inden(ta)tion
      check to detect nonsense fragment size
      replace wrong return 0 by -1
      more correct error message
      better padsize check
      consistant error return for asf_get_packet()
      1/0 fix
      adding assert to protect delayed_pic
      check num_reorder_frames for validity
      flame
      command line option to select the audio stream
      seek based on currently active stream fixes seeking in ffwma2_broken.wma which had some "empty" streams
      improve probe and give it the max score
      flv follows in movs footsteps and has random trash in the width/height fields
      make av_find_stream_info() duration user selectable
      remove padding to 16bit code, i dont think this change can break anything and it makes the broken MPEG2-pub.avi from http://www.albx79.it/stuff/ playable other players arent that picky either ...
      fix segfault with darkkben.free.fr/corrupted_h264.mp4
      make the asf demuxer resistant against insertion/deletion errors
      maybe fix segfault with missing extradata (unchecked)
      enum
      explicit values for all enums
      crash fix by Carl Eugen Hoyos cehoyos ag.or at
      set last_picture for h.264 too, this significantly improves error concealment quality, i hope it doesnt break anything as last_picture_ptr was always NULL for h.264
      cosmetics / remove #if 0 code
      fix inifnite loop with http://sam.zoy.org/zzuf/lol-mplayer.mpg
      stealing a few field names from mplayers asf demuxer
      fix seeking in wma
      trigger for the audio scope vissualization
      replace if(x>>b) by if(x>C) as shifts are slow on some cpus and i have my doubts that gcc can replace the shifts as x is signed, it could in theory but well its gcc ...
      simpify state and make code 2% faster
      another >> vs >
      10l to whoever wrote this (wrong fps, fixes nirvana.nuv)
      extract aspect ratio
      use avcodec_decode_audio2()
      some (probably exploitable) out of array write fixes
      add FIXME
      packets must be correctly interleaved
      fix decoding of takethat.mp3
      factorize switch_buffer()
      useless as this file wont be compiled if this is not true
      seek regression tests
      10l uninitalized cola
      better generic index building and seeking code
      dynamic index building so forward and backward seeking in avi without an index is possible
      dont output information about what is demuxed after a seek failure
      indent
      indent
      a few words about libavutil
      simplify and remove useless index in AV_W*
      Revised patch for HD DVD .EVO demuxing by (Ian Caulfield <lowercase name seperated by . place at here gmail place a dot here com>)
      wma encoder
      kill debuging code
      10l use of uninitalized var
      workaround sign bug
      wma regression test
      doxygenize
      vorbis suffers from the same sign flip issue as wma (so this lame workaround, though it arguably doesnt matter much where the sign is fliped back)
      vorbis regression test
      supporting files with A,V,I,0x19  header
      muxing packets with unknown timestamps is not allowed
      iam pretty sure send time is dts and not pts ...
      remove senseless and wrong direct access to ByteIOContext internals
      av_random() by Ryan Martell rdm4 name-server-seperator martellventures d0t com
      fix segfault with lol-ffplay2.mpg (dunno if this is exploitable, probably not easily)
      sanity check
      merge asm fragments in H264_CHROMA_MC2_TMPL()
      factor out common subexprssion (gcc of course is too stupid to do this ...)
      fix rawvideo
      packets with half pcm samples are invalid
      fix broken memset(padding, 0) code
      use dvr-ms timestamps
      maybe fix x86_64 (untested)
      slightly improve *_TIMER
      dont set the keyframe flag for audio as microshitty doesnt do it and some things have problems with our asfs, maybe that has a positive effect?
      disabling vorbis regression test for now, both encoder and decoder behave strange and have to be fixed
      remove md5 check for wma as there are rounding differences between cpus
      increaseing preroll time to the minimum of the asf files i have
      oops forgot this one (increase preroll time to min of my asfs)
      always use audio spread (instead of "video no spread" for audio ...)
      support yet another broken avi (filedoesitbetter.avi) which has a wrong stream count
      doxy for avcodec_encode_video()
      pal8 input
      av_noinline
      prevent h.264 MC related functions from being inlined (yes this is much faster the code just doesnt fit in the code cache otherwise)
      add a few asserts to ensure alignment
      simplify
      factorize mv_cache and mvd_cache init
      fix ref_count check
      fix 2nd ref_count check
      replace list<2 checks by list<list_count
      factorize fill_rectangle()
      not checking error return, resulting double free
      BGR/RGB4 byte formats as input
      remove imgresample and convert with the next version bump
      negative stride and h_size-=8 fix
      rewrite H264_CHROMA_MC4_TMPL (20% faster)
      factorize fill_rectangle() calls
      factorize more calls to fill_rectangle()
      try to fill in missing bits_per_sample
      warn if bps from user and what will be stored missmatches
      output delayed frames
      gcc 2.95 fix
      dont check buffer size if the decode function wont be called at all
      fix chroma mc2 bug, this is based on a patch by (Oleg Metelitsa oleg hitron co kr)
      remove now wrong comment
      shortening dependancy chain in chroma mc2
      reorder paddws to reduce dependancy chain
      reorder a few more paddws to reduce dependancy chains
      be less noisy
      rewrite find_frame_end()
      decoding rtp
      copy src to prevent the user supplied argumet from being changed
      a small step toward 4 plane support
      support feeding individual NAL units to the decoder instead of just complete frames
      functional part of the commit below
      cosmetic part of the commit below
      recommit of the change below after reverting earlier cosmetic-functional mix
      try to move fixed mov.c over the one with cosmetic-functional mix
      check for no frame and prevent segfault
      off by one error
      all the time related fields seem to be 64bit so i guess preroll is too (this is just a cosmetic anyway ...)
      all asf files i have have this field at 2 so lets try to change it to that ...
      all asf files i have have the 2 redundant duration fields differ by the preroll time so lets set them up like that too
      create codec_comment_header which looks more like what M$ creates, sane or not ...
      update forgotten regression test chcksums for
      prevent audio packets from being fragmented, wmp mac doesnt seem to like that and microshitty also doesnt seem to ever fragment them
      remove usless if() before av_free()
      indention
      free opt_names at the end of main() based on a patch by Limin Wang
      make av_find_opt() available to the public and add a mask+flags parameter to search for specific AVOptions
      add av_opt_set_defaults2() which sets just defaults from AVOptions whos flags match a user specified & mask = flags
      only store command line parameters in contexts with matching codec_type
      add avcodec_get_context_defaults2() / avcodec_alloc_context2() which take CodecType as an additional parameter
      set defaults based on CodecType by using avcodec_get_context_defaults2()
      add a -ab with default of 64k
      bitrate sanity check (fixes assertion failure)
      jpeg fourcc
      @section My bugreport/mail to ffmpeg-dev/user has not received any replies
      fix demuxing of broken MVI_3011.avi
      some url_is_streamed()==1 related fixes
      fix gsize=0 / -1 case (if(<24) goto fail) before it is not a good idea
      simple SHA-1 implementation
      explain how to test it
      its faster to copy the data to the stack it seems ...
      avoid silly ring buffer logic (faster with -O2, -O3 is always slower then -O2)
      dont recommand testing with -O3
      remove bigger and slower code
      make count count bytes not bits (this is simpler and leads to a very slightly smaller object file)
      very slightly smaller object file
      10% smaller object file, 2% slower
      cosmetic cleanup
      cosmetic cleanup
      2 other variants of how to implement the core part
      zero byte idea by rich
      revert 2% speed loss change (r8360)
      factorize VARIANT2 (smaller and slower)
      double ;;
      10l
      remove middle variant (keep fastest and smallest)
      simplify
      dont memcpy() simpler and same speed
      cosmetic
      put state[5] last so no padding is needed on arch where uint64_t needs 8byte alignment
      use CONFIG_SMALL
      smaller av_sha1_update()
      get rid of b_used variable (same size with CONFIG_SMALL)
      borrow finalization algo from sha1 (100byte smaller)
      #undef printf under #ifdef TEST (needed for testing)
      s/context/ctx/
      cosmetic (order context variables like in sha1)
      deprecate av_mallocz_static()
      dont malloc() static mv_penalty arrays
      remove redundant memset()
      remove silly static int done protecting the initalization of 32 elements to 1
      cosmetic
      remove useless static cm variable
      10^10l to whoever wrote this
      kill a av_mallocz_static()
      kill another 2 av_mallocz_static()
      kill 3 more av_mallocz_static()
      kill av_mallocz_static()
      move RLTable stuff to its own header
      kill av_mallocz_static() calls in init_rl()
      10l to whoever wrote this
      deprecate ff_realloc_static
      explain where the T table comes from
      160 bytes smaller object file, ask gcc devels why
      10l for myself, fixing --disable-encoders
      Patch review process
      add a entry to the policy about fixing warnings (i think this is common sense but if anydeveloper disagrees, just say so
      patches should be split into incremental changes if possible
      add rule to keep disscussion on the public mailinglist (stolen from mplayers policy)
      gpg and passwords rule
      remove secure password delivery requirement after complaint by the SVN admin
      support reordering pts in the decoder (this might be equivalent to mplayers -correct-pts mode)
      fix indention
      remove special case for h.264 whatever this was supposed to do its not correct
      add a delay variable to hold the timestamp buffer size
      remove silly video check, theres nothing video specific in there
      simplify
      simplify
      remove useless if()
      reorder pts into dts support
      fix indention
      remove incorrect special case for audio
      fix full range (jpeg yuv) chroma
      i think this is more correct
      revert r8428, the check was not that useless
      libav* headers for pascal
      link to ffv1, msmpeg4, asv1, 4xm docs
      update links to old probably obsolete bttv patches
      remove dithering of filter coefficients, improves precision by 1-2 bits and
      code for testing the accuracy of the resampler
      increase bits for non default highprecision mode
      make high precision mode accessible at compile time
      make kaiser windows with other beta than 16 available
      select more sensible default windows (= attenuation beyond the dynamic range of your input is silly if it negatively affects other parameters)
      optimize bessel function instead of trusting gcc to do trivial optimizations (as gcc doesnt ...)
      AUDIOPHILE_KIDDY_MODE
      revert  r8469 (increase bits for non default highprecision mode)
      optimize linear filter coeff interpolation code, this also makes the code less prone to overflows
      change high precision mode to 30 bits again after making the code less overflow sensitive
      replace long double in the audiophile kiddy mode with doubles, its faster (and more portable) and the audiophile placebo 120db stopband attenuation isnt affected
      clarify sub_id
      clarify codec_tag
      update resgression tests after the resampler changes
      set more reasonable cutoff for audio rate conversion
      1000000l for me ive misstakely reverted the regression checksum change from r8444 in r8486
      bmp and tga image regression test
      md5 test for the image formats
      simplify
      factorize
      lzw maintainer
      simplify
      move eob_reached logic into ff_lzw_decode_tail() which simplifies the code, avoids some checks in the innermost loop and also gets rid of the controversal break while hopefully retaining the last byte in a valid bytestream, invalid bytestreams still can have very significant overread
      simplify
      check input validity, this prevents a few variables from reachin odd values which might have lead to out of array writes and thus might have been exploitable
      simplify
      factorize &
      set pict_type to I type during init so parsers which dont set it get all i frames, fixes mp3 seeking
      warning about vhooks status
      codec id addition guidlines
      require at least a score of AVPROBE_SCORE_MAX/4 or reaching of PROBE_BUF_MAX
      prevent exit missuse
      fix probe of 001.vdr-broken-6025
      analyze end condition must match r_fps selection condition
      fix demuxing of IMAG0002.ASF
      10^10l to the idiot who designed this format and didnt even realize what number he used
      argh, who commited this trash
      WMVA
      make AVCodec.init() optional
      indention
      generic rle encoder by Bartlomiej Wolowiec b.wolowiec  students mimuw edu pl
      move doxy from .c to .h
      fix indention (less work to fix it myself than to check if a indention fix patch is ok ...)
      cosmetic
      dont use *int8_t for the arguments (ive missed that in the patches ...)
      tiff encoder by (Bartlomiej Wolowiec  b.wolowiec  students mimuw edu pl)
      cosmetic
      typo
      remove compound literals gcc 2.95 doesnt support taking an address of a compound literal
      tiff & sgi regression test
      LZW encoder by Bartlomiej Wolowiec b.wolowiec  students mimuw edu pl
      TIFF-LZW encoding support by (Bartlomiej Wolowiec b.wolowiec  students mimuw edu pl)
      table[index][ch] -> table[ch][index] (might be faster ...)
      general purpose vars should be int
      simplify
      prev1/2 -> prev[2]
      simplify
      use bytestream reader instead of bitstream for THP
      sane indention (no more 3 statements on one line...)
      memleak (seems ive missed that under the obfuscated indention)
      remove inappropriate debuging av_log() alternatively they could be changed to dprintf()
      remove redundant comments
      code claims to use 32bit timestamp, lets make it also use that
      cosmetic
      use shorter names for the block type enum
      indention cleanup
      remove useless debuging av_log()
      remove redundant comments
      conform to spec
      simplify
      general purpose var should be int
      shorter variable names
      align vertically
      s/rle_num_bytes/code/
      remove c93_ prefix from static functions in c93.c
      factorize & 0x0F
      also remove c93_ prefix for static function in the c93 demuxer
      allocate 32 extra bytes at the end of the probe buffer and remove most probe buf_size checks
      factorize
      file extension based probe is max/2
      simplify
      remove code_prefix variable, no speed change
      superflouos ()
      one patch one mail
      fix overhead and bitrate statistics
      fix tiff regression test (bitexact flag was ignoredwhich caused the
      correct another possible buffer overflow where interlaced would
      fix playback of odd_height.mov
      fix playback of angels_480-mjpegcompress.avi
      convert palette to yuv instead of converting each pixel after pal->rgb
      improve horizontal chroma resolution with pal
      fix mjpeg.mov
      patch submission checklist
      cabac bytestream optimization from x264
      init packet before calling the demuxer
      fix av_seek_frame_generic() so that it doesnt fail if the
      ensure that seek_test is rebuild if any of the libs changed
      dont try to seek on single images
      dont assert(0) due to buggy rm demuxer
      seek regression test reference
      enable seektest
      patch submissions should contain a suggested commit log message
      fix exploitable buffer overflow
      simplify
      reduce number of shifts
      better wording for the alphabetical lists
      fix possibly exploitable stack overflow with num_sprite_warping_points (found by reimar)
      improve mpeg4-es detection by rejecting streams with reserved startcodes (fixes 11-i_need_your_love-daw.mp3 detected as mpeg4)
      correct last element inited check
      dont write over the end of ref_cache
      slightly more correct spliting of frames
      sanity checks for bitrate tolerance and buffer size (= fail if a single frame at the
      use AV_LOG_ERROR instead of INFO for errors
      forgotten h.261 seek regression checksum update due to
      more snow todo items
      fix "incorrect frame size" errors with small.nuv
      make decoder decode the first frame properly if theres more than just one
      fix assertion failure
      warn the user about the the mdhd problem
      fix assert() 2nd try
      sanity checks (should prevent hypothetical div by zero issue)
      move jpegls up, fixes sf bug #1629152
      check qblur > 1 (prevent assert failure / segfault)
      fix skipped blocks
      split mpeg2ts_raw mode into its own demuxer using such a flag is an incredibly dirty hack
      revert r4802 (try with auto_guess if first detection fails)
      correct implementation of "try with auto_guess if first detection fails"
      kill completely idiotic code which hides all but 2 streams
      get rid of one layer of callback obfuscation
      get rid of if(1)
      remove fake service/provider names used for PAT
      make new_service() update existing info sanely
      do not perform SDT scan at the begin (this avoids a timeconsuming search for
      passing MpegTSFilter to the filters instead of a random void *opaque
      move structs and enums to top of file so they can be used
      dont rescan for information on each pid, but rather do it just once by
      unused variables
      remove duplicate PAT scaning code and actual PAT scan
      fix indention
      harmless typo
      partially revert r2931 (Support raw mpegts streams from dvr)
      add mpegtsraw demuxer
      proper per stream pcr_pid support by (Wolfram Gloger wmglo dent.med.uni-muenchen at de )
      remove now unused "global" pcr_pid
      simplify
      dont attach huge samplesto mails
      get rid of one unneeded  #ifdef HAVE_XVMC
      ffmpegs bug/patch/feature request tracker manual
      simplify ff_h263_round_chroma()
      rtp-h263 FAQ
      remove likely wrong reseting of align_iframe
      mention rfc 4629 for rtp-h263(+)
      lcl.c doesnt exist anymore
      Avoid misdetecting acelp in ASF as MP3.
      add vst option to select video stream
      move show_pict_info() to its own function
      also support DEBUG_PICT_INFO for h263i
      remove "Advanced Prediction Mode not supported" spam
      ignore invalid packet_obj_size
      try exact match before case insensitive match in codec_get_id
      detect voxware (closes issue39)
      note about removing needs_more_info
      print max and mse
      make the reference code use double instead of float where it is easy
      support raw 4bpp avi
      warn user if max=bitrate && max > min
      prevent assert failure with negative quant_offset
      support silly PCM variant used by zork nemesis
      kill antiwrapping code, i do not think this did anything good
      deal with the case where one of pts,dts has wrapped but the other has not
      chroma_qscale seems not used anywhere for mpeg1/2 so do not waste cpu time setting it
      avoid 2 additions (1 cpu cycle) per MB
      get rid of a useless variable
      move some code which was executed for every skipped MB so it is only executed
      make grayscale only decoding checks in inner loops compiletime killable
      use ENABLE_GRAY
      remove unneeded gray="no"
      move gray so as to not introduce further misordering
      merge 2 if()
      merge/simplify 2 if()
      reorder if() so that the condition can be simplified
      merge another 2 if() to save a few cpu cycles
      change
      exchange  for() and if() order in the MT_FRAME case
      exchange if() and for() in the MT_FIELD case
      exchange the values of MV_DIR_FORWARD and MV_DIR_BACKWARD (this is more sane,
      get rid of mb_type_len
      simplify s->mv_dir init
      output last frame if seq end code is encountered
      move code which adds a AVPacket into the packet buffer into its own function
      properly implement "user specified start time offset" which was added in r3217
      remove the mailinglist still preferred note, roundup is much nicer even though
      get rid of some double timebase convertion
      URL of the webinterface
      fix block_align for mp3 stream copy
      drop non keyframes before the first keyframe for stream copy
      detect MS wav codecs
      Make MPEG-1/2 decoder work with the libmpeg2-style bitstream reader.
      indent
      be slightly more careful with copying the codec_tag in the streamcopy case
      flashsv regression test
      topic explanations
      correct initial timestamps which have AV_NOPTS_VALUE
      set a more reasonable start_time if the first timestamps are AV_NOPTS_VALUE
      set the previous tag size correctly in flv files
      //more debug info
      fix misdetection of out.ac3
      update seek regressions for r9917
      make wmv1 and wmv2 playable with M$ DMO decoder
      ffmpeg forum
      thread hijacking
      make it a little harder to find the tracker, this should hopefully
      get rid of nonsense "tolerance" and mention VBV
      vbv related questions
      ignore preroll, it is generally not what AVStream.start_time should contain
      more verbose error
      ensure that the codec_id detection in av_find_stream_info() does not
      make output buffer larger so drift compensation can work
      audio drift threshold
      make dts_delta_threshold float so values <1 can be used
      always handle negative timestamp differences as timestamp discontinuity
      trying to finally get the nut muxer back uptodate
      forgotten av_
      write stream headers
      variables needed for muxer
      make StreamContext.time_base a pointer
      write correct time_base
      move main header writing into its own function
      move stream header writing into its own function
      use a dynamic ByteIOContext instead of using the lavf buffer
      forw_ptr CRC
      check forw_ptr CRC
      remove old unused nut (de)muxer
      move syncpoint timestamp resetting code to a common file
      move crc wrapper to slightly better spot
      move lsb2full to common file
      add a universal fallback frame code for the case that no other applies
      2 FIXME comments
      100l
      put_t()
      write frames and syncpoints
      get rid of printf
      move header writing into its own function
      store headers twice (getting closer to spec compliance ...)
      fix header_checksum
      fix forw_ptr crc check
      move syncpoint cache related stuff to common file
      set back ptr properly
      muxer is not yet fully compliant so prevent people from mistakenly using it
      enable nut muxer
      more FIXME
      fix x11grab
      info packet
      header repetition
      muxer should now be spec compliant (though it likely still contains bugs!)
      disable TRACE
      nut muxer
      ensure that the aspect ratio is spec compliant
      nut regression test
      flac probe
      remove unneeded buf_size check (see AVPROBE_PADDING_SIZE)
      mention AVPROBE_PADDING_SIZE at the read_probe() doxy
      100000l (forgotten return)
      ensure that default_get_buffer() doesnt reuse images if the dimension or
      this is debug not info
      remove useless code, found by oded
      do not remove FLAG_CODED
      simplify
      cola for diego
      fix decoding of samples.mplayerhq.hu/game-formats/idroq/demoEnd.roq
      pass *fr2 chunks to decoder (Toy-Story2_better-image-quality.4xa contains them)
      add support for old flac in ogg
      add second vlc table
      support ifr2 from Toy-Story2_better-image-quality.4xa
      export 4xm video version
      replace vlc_type by version and set that from codec_tag which contains the
      use version instead of some random 32bit value to switch between 16 and
      set pix_fmt properly depending on version
      fix motion vectors for old version
      fix type 3 aka skipped blocks for old version
      use bytestream_get_le*() this also fixes a big endian bug
      do not misuse movi_end for checking chunk sizes
      remove broken movi_end skip-hack-check
      document the email interface
      elaborate a little on AVStream.start_time, it seems the comment is not clear
      fix cur_dts at the end of av_estimate_timings_from_pts()
      get rid of idiotc double subtraction
      more stupid double subtractions
      and of course the unneeded double subtractions were blindly put in the
      fixme note
      remove idiotc double subtraction from the sse2 code (untested, no sse2 here)
      slightly change horizontal lift3 so it needs 1 bit less headroom
      replace <<1 by add
      replace <<1 by add for SSE2 (untested)
      simplify, speedup and reduce needed headroom by 2 bits in the 3rd
      Simplify and speedup code, reduce needed headroom by 2 bits in the 3rd
      Simplify and optimize the 4th vertical lifting step of the MMX code
      Simplify and optimize the 4th vertical lifting step of the SSE2 code (untested)
      remove code which become unused by the previous changes
      few words about the wavelets
      intro
      Change rounding of the horizontal DWT to match the vertical one.
      describe IDWT
      typo
      LL band prediction
      more complete dc prediction description
      optimization tips for the wavelet
      simplify
      sparc asm docs
      use more bits on the encoder side
      elaborate on needs_more_info
      int->int8_t (no need for them to be int ...)
      int->uint8_t
      more sparc asm docs
      use 16bit IDWT (a SIMD implementation of it should be >2x faster then with
      we only have 2 wavelets, the 3rd was just for experimentation  ...
      cleanup (remove some old experimentation related code)
      update mmx code to latest snow changes
      prevent one overflow in the first vertical lifting step
      avoid an overflow in the 1 horizontal lifting step
      avoid overflow in the 3rd lifting step, this now needs mmx2 at minimum
      cosmetics
      simplify senselessly complex addressing
      get rid of totally senseless "m" + read in register we have enough
      typo
      optimize 1st horizontal lifting step
      optimize the first vertical lifting step, this also prevents another
      workaround gcc bug, untested as my gcc is not complaining
      print EBX/EBP availability
      typo
      fix 1fps videos
      workaround ms (lack of) intelligence in their design of dvr-ms
      code to do halfpel interpolation per frame (unfinished and under ifdef but it
      Replace constants by HTAPS (half pel interpolation filter taps) where appropriate.
      fix segfault with dracula.4xm
      fix decoding of dracula.4xm before diego complains
      add a 8 taps filter for testing
      more todo
      16bit IDWT implementation note
      export put_timestamp()
      revert 10409
      revert r10403
      cleanup mc_block()
      indention
      remove outcommented code
      store halfpel filter coefficients in the header as well as the
      rename HTAPS -> HTAPS_MAX (later is correct after the previous change)
      document MC
      clarify
      scaling obmc tables under #if 0 also by 4 (yes they where forgotten apparently)
      spatial_decomposition_count != 5 bugfix
      remove ()
      allocate slice buffer after parsing the header containing needed parameters
      unused
      unneeded and impossible during init ...
      perform init after reading the values needed for init
      allow changing of the spatial_decomposition_count and quantization tables
      update spec to reflect changes to implementation about per frame
      do not force the halfpel filter coeffs to be retransmitted on every frame
      do not ignore return of decode_header()
      make changing spatial_decomposition_count per frame work
      cosmetic
      update todo
      move 8tap filter to not important as that is purely encoder side
      high level bitstream and decoding process in ascii art
      r_frame_rate
      fix timestamps used for ratecontrol
      make intra dc prediction less contradictory
      a little bit more bitstream syntax for the residual
      fix segfault with http://samples.mplayerhq.hu/nsv/nsv_samples/ot171_novideo.nsv.bz2
      indent
      duplicate note
      some experimental iterative quantization code
      cosmetic (get rid of _ at the start of local variable names)
      remove done items
      avoid generating invalid files
      indent
      factorize code and add safety check to prevent memcpying negative amounts
      prevent infinite loop and memcpy of negative amounts
      replace (disabled by default) assert(0) by abort() if the picture buffer
      factorize stream creation
      support flv with invalid header
      fix issue130
      prefer our native vorbis decoder over libvorbis
      check av_interleaved_write_frame() return
      fix misdetection of mp3could_not_find_codec_parameters.mp3
      typo
      intrax8 decoder patch by "someone"
      factorize
      cosmetic
      remove #if 0 code
      remove debug variable
      indention
      indention
      add {} to make the else look a little more normal
      remove unneeded {}
      indention
      vertical align
      remove inline from non speed critical functions
      remove useless #ifdef
      indention
      grammar
      tring to workaround gcc 2.95 bug which causes random failures
      rename fps(frame per second) to tb(time base) to prevent
      check region depth for validity
      use sizeof in snprintf (note the changed code is all under #if 0)
      cosmetic (remove != 0 / != NULL)
      cosmetic (place { consistently)
      cosmetic (x==NULL -> !x)
      frame_size vs. frame_size*channels bug
      fix segfault with interlaced h.264
      1000l _reserved  (found by checktree.sh)
      stupid code (casting of void*) found by checktree.sh
      senseless cast and ()
      make the AVOption code work with strings instead of crash
      simplify
      Fix non-numeric AVOptions.
      Remove rc_eq-specific code, it is passed by AVOption now.
      prevent O(n^2) execution of ff_mpa_decode_header() for valid mp3 files
      remove arbitrary 4096 limit for probing
      ensure that the probe code runs in O(n)
      return a slightly larger score if we find more than 500 valid mp3 frames in a row
      make mpeg-ps probe less happy with random data
      make rc_buffer_size apply to audio as well
      correct invalid duration
      add assert to detect invalid packet_timestamp_start/end
      fix decoding of the first frame of gray.mov
      thank you for your efforts to keep my code clean, but i would like to
      prevent warnings about functions being possibly unused
      av_*_next() API for libavcodec
      avoid using first_avcodec
      av_*_next() API for libavformat
      document how to free the result of av_alloc_format_context()
      clarify FIXME
      change some normal comments to doxy ones
      ME default does not belong to the ME type enum.
      another comment ->doxy
      Remove old variable with the next major version bump.
      Document Motion_Est_IDs.
      A little more doxygenization.
      Remove unneeded avio wrapper functions.
      Allow overriding codec_ids.
      Remove *_codec_id from AVFormatParameters with the next major version bump.
      Remove dead code which disappeared with the last major version bump.
      Merge recently added and still unused play and pause functions.
      Make recently added and still unused read_seek functions return offset_t.
      Document structure change rules with relation to ABI/API.
      Document a few more structure change rules with relation to ABI/API.
      Document ByteIOContext and URLContext change rules.
      30% faster h264_chroma_mc8_c(), this also prevents a possible out of
      ~15% faster h264_chroma_mc2/4_c() these also prevent some possible out
      indent
      More consistent { placement.
      closed/needs_more_info state
      Prevent silent overwriting of files on windows.
      fix pts and keyframe flags for yuv4mpeg
      Do not force fps unless the user actually specified one.
      Fix detected fps for old divx/xvid.
      Fix encoding wmv1 with -mbd 1.
      Remove AIW support code, it is unfit for ffmpeg svn, doing usleep() and
      Revert hack which should have never been commited.
      Try to fix url_split() so that the ?foobar part is in the path and only the path.
      indent
      More verbose aspect ratio error msg.
      Document return of read_timestamp.
      Correct doxy about timestamp units of read_timestamp().
      Extract aspect ratio from ODML/vprp.
      Always set next correctly, even if a matching element is found (that is
      Move *malloc() out of tree.c, that way the code can be used with
      Support removing elements.
      Fix selftest.
      Document node removial API.
      Document O() time.
      Flip key and element so types match, not that it matters for any code
      indent
      Drop non key frames before the first key frame.
      Randomly change the amount of noise if nothing is explicitly set.
      A little proggy i wrote years ago.
      Mention damaged input testing.
      Remove extra word.
      Range coder state transition table, based on a  patch by lars punkt taeuber ät web punkt de.
      argh s/ffmpeg/snow/
      Split range coding chapter in binary and integers
      Print libavdevice version.
      unused
      Factorize some code into functions, had this in my tree and will be
      Write 0 instead of seeking forward (and leaving bytes uninitalized),
      Set pts/dts in raw (yuv,rgb,pcm) demuxers.
      Remove green clouds.
      simplify
      Print removing of nodes in the test code.
      qscale==0 is invalid
      Replace idiotic (what moron wrote that code?) "buffer overflow" message by
      Consider initial EOR in the back_ptr calculation.
      Faster ff_sqrt()
      "What speedup justifies an optimization" section
      Cosmetics to speed up finding sections of interest.
      Comment to explain how the add/remove core works.
      Fix crash for CAVLC->CABAC switch caused by writing into NULL tables.
      Select non jpeg if there are multiple substreams.
      Only export packets which belong to the existing AVStream.
      indent
      Set correct codec type and id for text subtitles.
      Add non standard id needed for dvd subtitles.
      Set subtitle codec id correctly, i hope this does not break anything.
      fixing warning
      fix "uninitalized" warning, the alternative way.
      Comment about weird warning prevention.
      Revert r11689 and r11690 (uninitalized warning fix) as its theoretically
      Fix b-libav.ul/al seek regression test.
      Fix the following using void* casts, proper casts are less readable and
      Warn about casts which remove const.
      Dont cast const away.
      Merge declaration and initialization.
      Indention, it seems i have forgotton to commit this.
      Fix const qualifer issues found by -Wcast-qual.
      The buffer passed into the decoder should be const.
      consts
      consts
      Do not cast const away.
      consts
      one "cast discards qualifiers from pointer target type" less
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      Use bytestream.
      const
      const src for bswap
      Casts to silcence "incompatible pointer type" warnings.
      const
      const
      const
      const
      const
      const
      const
      const
      some const
      const
      const
      const
      some const
      const
      const
      const
      const
      const
      const
      const
      const
      Replace get32() by AV_RL32().
      const
      some const
      const
      const
      const
      const
      const
      some const
      const
      const
      const
      forgotten const
      const
      const
      some const
      some const
      const
      const
      const
      more const
      const
      const
      const
      const
      const
      some const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      const
      Mixing declarations and statements is an error, so is any construct
      void arithmetic
      Make src const.
      const
      Make avcodec_decode_* functions take const input buffers.
      Remove incorrect casts that should have never been there.
      Make string constants const. Why is this not default?
      Missing const found by -Wwrite-strings.
      Warn if a comparison is always true or always false due to the
      const
      Missing const found by -Wwrite-strings.
      -Wwrite-strings nitpicking fix
      Fix info packet type, found by oded as well as the new pedantic const
      const
      fix 8 "initialization discards qualifiers from pointer target type"
      const
      Another const found by -Wwrite-strings.
      Remove incorrect cast found by -Wwrite-strings.
      2 more missing consts found by -Wwrite-strings.
      Correct incorrect statement about warnings.
      Correct fatal ambiguity.
      Deprecate old and inefficient per instruction asm().
      Allow the user to select which codec out of several in stsd he wants.
      optimize apply_welch_window()
      assert() and note about odd len.
      Remove pointless code.
      Remove pointless code.
      update flags
      Read match_time_delta in the framecode table.
      Read match time delta in the frame header.
      -vsync 2 (drop frames if there are too many but do not duplicate any)
      Elision header demuxing support.
      Header elision muxing support.
      It appears last_pts was not set ...
      Fix timestamps and durations if the first packets have no durations nor timestamps,
      Update ist->pts after decoding a codec frame not after decoding a container
      Use next_pts if pkt->dts is unknown. In an ideal world this would not be needed
      simplify (next_)pts code
      Set next_pts to pts if it is unknown and pkt->dts is not known either. This
      Print a warning if a decoder produces more than 1 frame of output for
      Set initial next_pts to unknown, this avoids the nonsense timestamp
      Add notes where known timestamp issues remain.
      Reduce the number of senselessly scanned bytes.
      const
      Fix  ;;
      Make timestamp interpolation work with mpeg2 field pictures.
      Get rid of a fake timestamp discontinuity at the begin of V.VOB.
      Make find_frame_end() merge 2 mpeg2 field pictures. This should make mpeg2
      Forgot this, which is needed for merging field pics too.
      revert 12156
      Disabling all SSE* code for old gcc to avoid alignment issues.
      parse_number_or_die()
      "General Tips" section
      typos
      Document avfilter_formats_ref() with some ascii art.
      Better description for merge_ref().
      Better description (merge is too vague IMHO).
      cosmetic
      Ascii art to explain what avfilter_formats_changeref() does.
      Remove apparently unneeded and buggy align.
      flv/swf do not have a big endian codec id, they only support
      Fix issue357
      realloc fifo
      Reduce initial fifo size, so as not to senselessly waste memory.
      Set indexmem to a saner default.
      Allow use of my AAN DCT under BSD, AMD has agreed to write and contribute
      Clarify which BSD license.
      Make code C conformant.
      Make diego happy.
      Fix trellis quant + AAN DCT.
      floating point AAN IDCT
      One variable less. No real change to the actual code.
      Avoid a=b+c but rather use a+=b (gcc is too stupid to do this itself),
      Get rid of another useless variable.
      Alternative and faster (on gcc x86) plane rotation.
      Remove another temporary variable with which gcc has problems.
      Stuffing bytes to not violate minimum legality rate.
      25% faster floating point AAN IDCT.
      Make gcc ignore the distinction between +0.0 and -0.0, this allows it to
      include dependencies ...
      better include
      suppress the second warning
      add cdvc
      Remove INFINITY.
      Remove INFINITY as suggested by rich.
      Store aspect ratio in avis.
      subtitles (codec tag tables still missing ...)
      factorize
      Fallback to sample rate if bit rate is 0. This avoids 0 timebases in nut.
      Print slightly more useful info.
      Use block_align as the avi spec says.
      Provide non null class name to avoid crash with ffmpeg -h.
      Ignore flags after pts/dts if there are no bytes left.
      Prevent segfault due to reading over the end of the input buffer.
      Detect and prevent reading over the end of counts_*. We pass the error
      Remove unused elements from tables.
      Turn ancient V offset numerical constants into named ones.
      Fix handling flags with the MSB set.
      sws_flags is unsigned.
      One more (forgotten) fix for fixing sws_flags.
      The last flag fix fixed mpeg4-qprd (there the flags ended up being trashed
      Subtitle support. (untested)
      simplify
      Skip over unknown extended_content_header tags.
      Clarify the intended meaning of wont_fix.
      smaller tables
      If consumed does not match nalsize, favor nalsize.
      remove unused temp
      Hardcode register to prevent aparent miscompilation.
      One non functional AVPalette chunk less, one heap overflow less.
      More complete documentation for AVFrame.reference.
      Document a few url_* functions.
      Document AVCodec.capabilities.
      Document AVCodec.flush()
      Document get_*_buffer()
      1st attempt at correcting the punktuation.
      Fix possible heap overflow caused by av_fast_realloc()
      Fix aspect ratio.
      Support wb00+dc00 chunk mix.
      1000l to myself, dereferencing uninitalized pointer.
      indent
      Fix closed gop flag disapearence.
      help gcc-2.95
      Do not use the mlib IDCT by default. We do not want videos which are only
      Add a point to the checklist about checking new demuxers/decoders
      Improve AVFilterFormats doxy.
      Factorize code a little.
      Decode stuff from extradata as well.
      Merge declaration and initialization.
      Merge declaratio and initialization.
      Disable the split function. This should end the mpeg1/2 global header issues.
      simplify
      add CODEC_ID_DIRAC
      Fix seeking in rm.
      Avoid crash with idiotic cropping. Fixes crash with CVFC1_Sony_C.jsv
      Crop parameters are unsigned, having them negative could be bad and lead
      Remove redundant &.
      Restructure if/else a little.
      indent
      Print warning if scale/rate is invalid.
      ILP64 fixes (untested)
      10l
      Stop find_stream_info() searching for further streams if 2 streams have
      revert 	r12809
      Use generic seeking code for flv. The removed code was a buggy duplicate.
      Recommit r12809
      Remove useless code.
      Identify codec in GMCMidiaASX.mov correctly.
      Fix probe for mov00003.mqv.
      Reduce probe score as it misdetects our flv file from the regression test.
      Using the same probe function in 2 formats is not a good idea.
      Do not detect a format if another has the same score.
      simplify transform
      100l lost add_bias
      normalize signs
      do the -1 outside of the transform
      commutative law based cleanup
      Avoid reverse addressing, not sure if this is faster or slower but people
      simplify
      Move scale_bias to a more sane place. I am starting to wonder how this
      Move the *-1 to a more sane place as well.
      cosmetic
      unused
      unassigned imdct & windowing TODO
      simplify
      useless variable
      senseless code
      more senseless code
      Change k limiting code, i think the code was buggy.
      Factorize decode_postfix() out.
      Factorize more code.
      indent
      remove AACAC3FrameFlag
      Change aac and ac3 parsers to use ff_combine_frame().
      Add new_frame_start and need_next_header.
      part 1 of EAC3 support
      Part 2 of EAC3 support, this is still disabled as it breaks regressions
      100l I broke AC3 in MPEG seeking.
      indent
      This fixes the code so it is correct on both sides of the if().
      Make most of E-AC-3 work without breaking regression tests.
      Ensure that the timestamp reading code used for seeking chooses a position
      Fixing memleaks.
      minor memleak
      Avoid using reserved __names.
      raw GSM demuxer (does not work yet as parser is missing)
      Support invalid mpeg created by vdr
      Fix decoding of:
      indent
      Setting i to 0 once should do.
      Move declaration a little.
      Fix memset(0) based buffer overflow.
      Heap buffer overflow.
      Output buffer overflow.
      Check url_seek() in url_open().
      Improve motion estimation metric.
      Little fix for block sizes <16
      Avoid using log2() freebsd does not support it.
      Simplify vsad_intra16_mmx()
      Simplify vsad_intra16_mmx2()
      Simplify vsad16_mmx().
      Simplify vsad16_mmx2().
      OPT_EXIT
      Do not pass program_name as argument to show_version(), this allows some
      Get rid of more senseless wrapper functions and use OPT_EXIT instead.
      Fix + - bug with non flags.
      make av_strdup(NULL) return NULL
      indent
      chapter id and timebase
      align comments
      clarify start/end timebase
      Add id to AVChapter, untested (where do i find matroska files with chapters?).
      indent
      remove useless title check
      indent
      vertical align
      s/num_chapters/nb_chapters/ all other similar variables are called nb_*.
      nb_chapters should be unsigned
      Make ff_new_chapter() return AVChapter instead of int so its consistant with
      Chapter demuxing support. (untested as I have no nuts with chapters)
      Warn user about invalid timestamps.
      Document pts/dts.
      Pass time_base as argument to new_chapter() as well.
      Anoter try to make +- behave sanely.
      Merge init_rootpow2table and init_pow2table.
      simplify init_pow2table
      Move const *pow2tab out of context.
      Do not reject odd width/height.
      Fix MB skipping with droppable frames.
      Make opt_verbose() work again.
      Make av_fifo*_read() ignore the available amount of data.
      Ensure that one can store X bytes in a fifo of size X.
      Make ac3 in avi work better.
      Make stream copy of ac3 from riff based formats work.
      print direct_spatial_mv_pred
      allow av_parser_close(NULL)
      indent
      Warn about packed B-frames. (especially useful if the file in question is
      Support non interleaved avi with phantom streams.
      Do not use the pts/dts calculation code which needs a parser, when no parser
      Do not read from prior the array, fix CID 127.
      Fix memleak, fixed CID123.
      memleak / fixes CID118
      redundant check--
      Arrays where one element too small, fixes CID114.
      Check dia size a little more. Fixes CID80.
      Tell the user if a AVI is non interleaved.
      Fixes NULL pointer dereference CID66
      redundant
      minor simplification
      useless
      Fix return type of ff_init_me().
      Null pointer check / CID26.
      Update regression checksums after dts calculation fix (r13470).
      set coded_frame
      Remove coded_frame==NULL checks for video encoders.
      Prevent 128*1<<trellis from becoming 0 and creating 0 sized arrays.
      Do not read from after the table, fixes CID108 RUN2.
      Do a little more checking.
      remove != 0
      Removing dead code, fixes CID4 RUN2.
      Dead code removal, fixes CID59 RUN2.
      memleak / CID124 RUN2
      ILP64 fix
      ENOMEM is better, yes ...
      Make init_vlc* support proper static tables instead of this broken beyond
      Little helper macro to make the use of INIT_VLC_USE_NEW_STATIC easier.
      Switch to INIT_VLC_USE_NEW_STATIC.
      Remove constant argument.
      Remove incorrect usage of av_free_static(), ffmpeg.c is not a
      Change init_2d_vlc_rl() to not use *alloc_static().
      Change init_vlc_rl() so it does not use *alloc_static() anymore.
      Move *_static to bitstream.c which is the only file left which needs
      Remove non C code.
      unused
      As *_static are not deallocated anymore except on program termination
      unused
      Prevent scalefactors from overflowing.
      Simplify parser core.
      Rename last_frame_offset to next_frame_offset.
      indent
      unused variable
      Factorize fetch_timestamp() into its own function
      Add off argument so that fetch_timestamp can search for timestamps
      Allow matching timestamps to be removed by fetch_timestamp().
      cosmetic
      Move new field to the end to avoid ABI breakage, I thought this struct
      Our MPEG-TS demuxer does not send complete PES packets but sends them
      Only store timestamps when there are timestamps.
      indent
      Do not set pts in update_initial_durations() if stream has b frames.
      Make update_initial_durations() also work if a dts occured before a
      In mpeg1/2 timestamps are associated with the access unit which
      ff_fetch_timestamp() prototype
      Fix reading an element after the array.
      Presets support.
      Search /usr(/local)/share for preset files as well.
      indent
      Support longer paths for and longer names in preset files.
      Make START/STOP_TIMER not require -v 2.
      Make timestamp debugging work nicer.
      Change MODE_* to 1<<c so several can be set at the same time.
      simplify MODE_3G2
      Calculate ftyp size instead of hardcoding it.
      Add isom to the compatible brands.
      I think we are also compliant to iso2.
      -0x8000 == 0x8000 with int16
      Remove dead code.
      get_bits() outputs exactly as many bits as requested no need to mask them.
      Support decoding of sample_g726.asf.
      Check number of bits so we do not try to use table entries which do not exist.
      Print sane error message for channels != 1.
      Correct validity checks.
      simplify
      Float11 does not need int, .o file becomes smaller and the code might
      Use smaller data types for tables.
      Remove redundant zeroing (the context is memset(0)).
      Replace i2f(0) by the actual thing done, gcc is not an optimizing compiler.
      Do not inline g726_iterate() the function is big so its inlining will
      Get rid of useless wrapper function.
      simplify
      Change iquant tables to int16.
      Remove the truncated bitstream handling from our g726 decoder.
      Get rid of the redundant AVG726Context.
      simplify
      remove unneeded tr == 0
      useless ()
      simplify
      1 abs() less
      Factorize I >> (c->tbls->bits - 1) out.
      Does not need to be int16.
      Copy 4 pointers to avid dozends of ptr dereferences.
      Get rid of G726Tables.bits.
      Factorize c->ap += (-c->ap) >> 4 out
      Do not shift F[I] twice, it is also clearer and smaller now.
      Prevent duration calculation code from generating random durations for
      Use correct timestamp value for stream copy.
      We cannot calculate the duration of vorbis packets even if we do know the
      Fill missing pts in during muxing when we know that they are equal to dts.
      Set frame_size to the gcd of what the decoder can output.
      Make av_find_stream_info() find the frame_size for vorbis.
      Prevent invalid timestamps from being stored.
      Support >=4096 dimensions, fixes issue499.
      VP3 & Theora parser to extract keyframe flags.
      Also print flags with FF_FDEBUG_TS.
      Remove start_time guessing code.
      Register intrrupt_cb() early enough so we dont get stuck with tcp input
      Fix exploitable code.
      unused var
      Remove a dozen unused variables.
      vertical align
      remove code that discards streams randomly
      Another variable that did nothing at all
      Remove debugging trash.
      Get rid of weird pre reading code.
      merge decl & init
      Do not completely bail out for invalid params, just skip the current sector.
      memleak
      Inform the user about unhandled sectors.
      indent
      Make code independent/less dependent on dsputil.
      make sum_of_squares() more generic
      More correct comment.
      We do not only calculate it but we also return it.
      All id3 parsers are buggy, 0 termination fix 1 of n (issue created by andreas
      All mp3 parsers are buggy fix 2 of n (out of array write, i suspect not exploitable)
      dont use C-asm loops and unroll once float_to_int16_3dnow()
      Fix x86-64
      1 c-asm loop less and 1x unroll of float_to_int16_sse()
      Set cur_dts to 0 only during creating new streams and not every time cur_dts happens
      10000l for myself, my last change was incomplete.
      indent
      Fix the av_set_string() free / alloc issue.
      fix IPCM_decode_error.h264
      Allow setting codecs through preset files
      indent
      float_to_int16_sse2()
      New full search ME
      Remove ancient code that hasnt worked in many years.
      Fix misdetection of MPEG-PS (AVSEQ03.DAT and AVSEQ06.DAT) as psxstr.
      simplify
      Check that we have enough output space available.
      Fix not coded channels / silence decoding.
      indent
      Pass AVPacketList instead of the whole context to add_to_pktbuf as that is
      Move set_codec_from_probe_data() up to where the other probe functions are.
      Move add_to_pktbuf() before av_read_packet(). My future work on codec identification
      command_streams are type data not unknown.
      New codec probing system try #1.
      indent
      h264_probe()
      Add mpegvideo and H.264 to the codec probe.
      Only try to probe the codec when the available data has roughly doubled
      indent
      Use strcmp() instead of strncmp() in set_codec_from_probe_data().
      Remove old hackish codec probe system.
      Use CODEC_ID_PROBE when the mpeg-ps demuxer does not know exactly which
      Use -vsync 0  unless the user explicitly asked for something
      Remove gif demuxer, the code contains a gif decoder and lzw decoder neither
      On EOF keep calling the decoders until they do not output anything
      Make our flac decoder decode all the data at EOF.
      -itsscale to scale timestamps
      Fix decoding of 320x240.ogg.
      Fix infinite loop at EOF.
      Remove common factors from timebase in av_set_pts_info().
      Move the codec_id overriding at a slightly better place.
      Make flic probe check a few more things.
      Print slice num as well.
      The funny memcpyin svq3 generally has src & dst overlapping, so it
      SVQ3 uses the JPEG variant of YUV 4:2:0.
      indent
      10000l, revert
      Try to fix FATE after my time_base simplification change.
      100l, someone used FLT_MIN where -FLT_MAX was meant at a few places in the
      Ensure that exactly one scaler algo is used.
      10l forgot SWS_BILINEAR
      simplify yuv2yuv1()
      Forgotten accurate rounding function YSCALEYUV2YV121_ACCURATE.
      indent
      Make C code in yuv2yuv1() do accurate rounding, this could be split
      If metadata has been parsed goto end instead of trying to
      Only realloc() bitstream buffer when the needed size increased,
      Make bitstream_* fields unsigned.
      Support reading large metadata.
      Correcting displayed muxing overhead and video stream size.
      Make frames non reference by default, so they are disposed of automatically
      minor simplification of qdm2_decode_fft_packets()
      Make the ugly hack which uses an unused entry in the internal buffer
      Drop out of order pics even if bitstream_restriction_flag is set.
      Remove delayed_output_pic, I do not understand what this variable was good for.
      simplify
      indent
      Do not forget marking disposed delayed pics as unused.
      Remove useless #if 0 code.
      Simplify picture output code.
      Remove apparently unneeded out!=NULL check.
      Try to fix my latest 100l fate SVQ3 breakage.
      Make the automatic has_b_frames increasing code increase it up to the
      Make strict_std_compliance available to decoders.
      Remove explicit strict_std_compliance support as it seems to work fine over
      Take the brain amputated spec literally if the user asks for it (-strict 1).
      Fix h264_loop_filter_strength_mmx2() so it works with PAFF.
      Add skip* AVOptions, which seem to have been missed.
      Fix used references indexes in the loop filter.
      Simplify spatial direct ref selection with FFMIN3()
      Cleanup, use av_freep() instead of av_free(x); x=NULL
      fix memleak
      More correct ref comparison in the loop filter.
      Simplify slice type checks.
      Remove superfluous ().
      SI/SP slice support. (not bitexact)
      Use slice_type_nos instead of slice_type where it makes no difference.
      Loop filtering with multiple slice types fix.
      Deobfusate chroma_qp_table[1] init code.
      Setting chroma_qp to 0 for PCM blocks is really an interesting minor
      qscale has a range of 0..51 we thus do not need a 256 entry table and neither need
      3rd try to get the loop filter ref/mv check working correctly.
      Reset frame_num for MMCO 5.
      Support gaps in the frame num.
      Remove redundant keyframe checks.
      Simplify cross_idr calculation.
      Remove truncated support, use an AVParser if you used that, it is the same.
      100l cross_idr simplification had a typo
      Better explanation for buf_size==0 code.
      Reset prev_frame_num for idr frames as required by the spec.
      Fix segfault with non existing input files.
      Fix loop filter with CAVLC 8x8dct.
      Fix tiny_psnr so it compares all bytes (it did skip the last block).
      init_poc() vs. idr() cleanup
      cosmetics (indent and 2 empty lines)
      Minor simplification of code in init_poc().
      Set field_poc (it was half uninitialized before) and use it
      Simplify strange obfuscated code in MMCO_LONG handling.
      Make sure mmco index is 0 (instead of uninitialized) in all error returns.
      Clear long_ref when a frame is no longer a long_ref.
      Assert that long_ref has been cleared for released pics.
      Remove redundant h->mmco_index= 0.
      Simplify mmco setup code used in IDR pictures.
      Remove another of these obfuscated special cases for field pictures
      10000l, forgotton }
      indent
      Remove strange long_ref=0.
      Merge oddly duplicated remove_long/remove_long_at_index() function.
      Make remove_long() call unreference_pic() internally.
      Removing yet another weird field pic special case,
      Make remove_short() match remove_long() in that it calls unreference_pic().
      Do not scare the user with error messages when everything is ok.
      Rewrite fill_default_ref_list(), the old code was obfuscated beyond
      indent
      Remove another stupidity, long_ref==0 check 4 lines after assert(long_ref==0).
      10l typo, fix compilation
      Remove mysterious ref->data[0] check.
      Print an error if a reference picture is missing.
      Remove another 2 incorrect checks.
      Revert r14404
      chroma_format_idc=0 aka grayscale support.
      indent
      Detect IDR less H.264 correctly.
      Make the selftesting code use and compileable.
      Change benchmarking code so it returns comparable values.
      Fix misdetection of H.264 in mpegts as mpeg1video in mpegts.
      Simple lagged fibonacci PRNG.
      Warning about MTs speed.
      Document av_lfg_get().
      Simplify intra PCM MB handling.
      Add a multiplicative LFG for those thinking the additive is not good
      1000l typo my new PRNG converged to -1.
      Revert r14497
      Recommit r14497
      I found a maintainer for all parts that had none listed :)
      Picture.ref_count/ref_poc have to be stored per field (actually also per
      mb_field_decoding_flag was not initialized
      Remove useless mb_field_decoding_flag setting code that was at the wrong spot.
      Use the correct collocated field in pred_direct_motion() when it differs in parity.
      Fix dist_scale_factor for long term reference pictures.
      Use ref_poc and ref_count from the correct field.
      Fill ref_count/ref_poc for both fields of frames.
      Remove redundant code.
      The code under FRAME_MBAFF for temporal direct mode MBs is needed for PAFF as well.
      Port col_parity code from spatial mode to temporal mode,
      *2 /2 is wrong for PAFF references, so do not do it.
      Use frame_num and reference instead of poc for matching frames for direct
      Port b4/b8_stride code from spatial direct mode to temporal.
      Factorize some code between temporal and spatial direct mode.
      indent
      simplify y_shift/ref_shift code
      Simplify l1mv/l1ref calculation.
      indent
      Use local variabes for *stride, where local variables exist.
      Prevent low delay from being used with mpeg1.
      Remove odd pairwise deblock filtering applied to MBAFF MBs.
      Remove more code that has become useless after the pairless MBAFF loop
      Do not execute the part of fill_caches() for the loopfilter that overwrites
      Use the correct reference tables for interlaced MBAFF MBs in the loop filter.
      Remove a few more lines of unneeded code in fill_caches() for MBAFF.
      Mask bits of reference indexes properly to avoid negative numbers entangling
      Support MBAFF + constrained intra prediction.
      Simplify constrained_intra_pred code in fill_caches().
      The deblock filter does not need all the intra availability stuff so do
      Add needed include, make it compile without -D_BSD_SOURCE.
      Move -D_ISOC9X_SOURCE to configure so its available to all tests.
      10000l, revert mistakely commited and unrelated hunk.
      Change _ISOC9X_SOURCE to _ISOC99_SOURCE.
      Remove unneeded whitespace and quotes from -D_ISOC99_SOURCE.
      Do not fill the pts reordering buffer with guessed dts.
      Add -D_POSIX_C_SOURCE=200112 to cflags.
      Add -fasm and -std=c99 when supported.
      Simplify by using FFMAX.
      Add flag to mark demuxers that can output discontinuous timestamps.
      Revert 	r14260
      Fix rc_eq mem leak.
      Principal component analysis
      fix includes
      Testing code uses random().
      put testing code under #ifdef TEST
      Do not mix declarations and statements (by ramiro).
      Make ff_pca_init() allocate it struct instead of letting the user provide
      Add prototypes to header (based on code by ramiro)
      Move context struct to c file.
      Add multiple inclusion guards for consistency and to avoid them being
      Add PCA to TESTS.
      restore alphabetical order of TESTS
      Copy and paste LGPL from tree.h, the previous one referred to a non-existing
      Workaround bug in types.h that has typedef __u64 under #ifndef __STRICT_ANSI__.
      Provide a simpler way for the user to reorder her timestamps.
      Simplify timestamp reordering by using the new API.
      Remove some irrelevant comments.
      Document code, do not document difference to a irrelevant reference implementation.
      Make doxygen comments consistent with the rest of FFmpeg.
      Cleanup comment for ff_acelp_interp_filter.
      Mention the fixed-point scale of the FIR filter coefficients.
      Right half is clearer than positive half for specifying which part of
      Correct incorrect length parameter doxy for ff_acelp_interpolate().
      Clarify how generic ff_acelp_interpolate() is.
      Try to improve parameter doxy of ff_acelp_interpolate().
      rename pitch_delay_frac in ff_acelp_interpolate()
      Rename subframe_size to a name that is not specific to some specific use
      Remove ff_acelp_weighted_filter() which is just a function wrapping a 1
      vertical align
      Remove mathops.h dependancy.
      Mention cut-off freq of ff_acelp_high_pass_filter() instead of some
      Doxygen syntax consistency.
      Make comments doxygen compatible.
      doxygen consistency
      Make equation in doxy of ff_acelp_weighted_vector_sum() more readable by
      Remove redundant FFMIN().
      Ensure that SDL_AddTimer() is never called with a delay of 0 because
      Replace cliping in ff_acelp_interpolate() by a check&av_log, this should be the
      Noone seems to want to maintain MPEG-PS so i will.
      vertical align
      Cosmetics (format ff_nelly_dequantization_table nicer)
      Fix mid/side stereo buggy output zeroing, fixes issue264 part 2.
      Do not truncate timestamps before the muxer as it makes simple
      Simplify (drop useless intermediate sum and diff arrays).
      Move t1 and t2 declaration from the QMF code closer to where they are used.
      Half the size of subband_fir_noidea and get rid of memmove & memset of it.
      Get rid of the 512 sample memmove().
      Replace obfuscated mdct in qmf_32_subbands() by ff_imdct_half().
      Remove the cos_mod table that has become unused after replacing the imdct
      Fix lossless jpeg encoder to comply to spec and store full redundant
      Remove debuging junk that probably hasnt been used by anyone since years.
      Optimize vorbis_residue_decode() so that vr->type is a constant.
      Warn the user about me_method values that are not supported.
      1000l Fix snow regression test that i broke in r15135.
      Check ff_init_me()s return value.
      Fix detection of MPEG4 in MPEG-PS.
      Try to support invalid h264 that does not mark IDR with POC=0.
      Fix round to even for aspect ratio correction.
      Try to clarify the semantics of AVPacket.duration.
      Add AVPacket.convergence_duration.
      Clarify sample_aspect_ratio.
      Ignore the spec, and calculate aspect ratio on width/height.
      Fix artifacts of MPEG4-720I-DN-SV8K.trp caused by lack of IDR pictures and
      Add fflush to prevent stdout & stderr from being mixed.
      Fix rgb15/16 vs. bgr part1.
      Fix rgb15/16 vs. bgr part2.
      rgb vs bgr fix for the unscaled converters.
      Fix 4 of the unscaled rgb15/16 converters, each of these contained
      Remove workaround for rgb/bgr mess.
      Fix typo.
      Fix 4 and 8 bit RGB/BGR input.
      Support PIX_FMT_RGB32_1 and PIX_FMT_BGR32_1.
      Fix SWS_FAST_BILINEAR and SWS_POINT with some unscaled rgb<->bgr converters.
      Make 16bit grayscale output work.
      Make fast bilinear scaler work again.
      Fix accurate rounding mode on x86_64.
      Fix typo in comment.
      Fix jpeg yuv.
      Try to fix wma regression.
      Remove stuff under #if LIBAVCODEC_VERSION_INT.
      Drop deprecated SAMPLE_FMT_S24.
      Remove DEFAULT_FRAME_RATE_BASE.
      Rename error_resilience to error_recognition.
      avcodec_find_best_pix_fmt() needs a 64bit mask as there are more than 32 pix_fmts.
      Remove deprecated avcodec_decode_audio() this one was not under #ifdef.
      Prevent overflows during mpeg->jpeg yuv.
      More correct rounding for the rgb/bgr->yuv converters.
      Add table of rgb->yuv convertion coefficients.
      Rewrite bgr24->yuv mmx code, the new code is cleaner, more accurate,
      More accurate rounding for 8bit inputs.
      Make SWS_FULL_CHR_H_INP work.
      Factorize rgb/bgr15/16/32->Y by using the preprocessor.
      Factorize RGB/BGR15/16/32->UV by using the preprocessor.
      Clarify convergence_duration.
      Add bgr/rgb15/16/32->UV-half to the macro so theres less code duplication
      Change RGB2YUV_SHIFT from 16 to 15 to make it able to work
      Fix build failure due to %%eip on x86_64.
      Fix overflow.
      Make the 2point linear interpolation coefficients correct even for the
      Correct normalization constant for the vertical filter, iam not
      Fix chroma yuv->rgb tables for jpeg style yuv, this was missed as it
      Set rgb2yuv constants more accurately, makes no real difference though.
      Factorize yuv2packedXinC().
      Do not do unneeded cliping in YSCALE_YUV_2_PACKEDX_C.
      Implement full horizontal chroma for rgb/bgr24/32 output.
      Remove mistakely commited code i used for testing.
      Fix typo that lead to averaging of the same pixel in rgb24ToUV_half().
      100000000000000l, forgot to commit header change for r27580.
      Do not use the unscaled yuv->rgb converters if SWS_ACCURATE_RND is set,
      rgb24toyv12 isnt accuratly rounding, so disable it as well when the
      Add support for PIX_FMT_MONOWHITE as output format.
      Support mono as input format.
      10000l PIX_FMT_MONOWHITE check was really a || 1.
      Add support for PIX_FMT_YUV440P.
      Fix another 1000l bug in the mono input code.
      The yuv->rgb tables are too small for cliping to be avoidable,
      Add bitexact flag.
      Make the horizontal C scaler code clip only against INT16_MAX not 0,
      yvu9toyv12Wrapper is not bitexact so disable it when the user wants
      Ensure that additional filter coeffs that exist due to alignment
      Make horizontal mmx scaling code match C code.
      Disable mmx routines that are not bitexact when the user wants
      Switch regression tests to swscale.
      Rename yuv2rgb variables to avoid name clashes with the ones used by bfin asm.
      Fix indention.
      Add sparc maintainer.
      The official guide to swscale for confused developers.
      Reword a few sentences to make them clearer.
      Add , to clarify sentence.
      Move dither tables from yuv2rgb to swscale, they have been written by me and
      Use av_mallocz() instead of for() =0;
      Avoid some explicit types in sizeof().
      Avoid using floating point for calculating filter coefficients.
      Correct the threshold of get_ur_golomb_jpegls() at which the optimized
      Fix generated md5, it was wrong for some input lengths.
      Make decoder more robust by default against broken encoders.
      Fix regression test failure caused by golomb limit not being considered
      Only print "invalid new backstep" when it is really invalid.
      Fix segfault with rgb24 and full_internal_chrome due to non existing alpha
      Ensure that extradata is not parsed twice.
      Make pps/sps scaling_matrix initialization closer to the spec.
      Support comments in preset files.
      Ensure that muxer and encoder aspect ratios match. So no
      Avoid undefined behavior for removing elements that were not in the tree.
      Remove check and error message for "MBAFF + !direct_8x8_inference is not implemented"
      Remove fixme that is based on the assumption of !direct_8x8_inference && MBAFF.
      Set poc for field references in MBAFF.
      Move direct_dist_scale_factor() and direct_ref_list_init() after fill_mbaff_ref_list()
      Make MBAFF temporal direct mode closer to the spec.
      Simplify initialization of a few direct mode related variables.
      Correct comment in the direct mode code.
      indent
      Fix ref_shift so that it is correct for more/all? MBAFF/PAFF mixes.
      Fix all the recent problems Justin had run into with the AC3 parser,
      Another try to fix temporal direct mode references.
      A few empty lines for readability.
      Factorize lpc code slightly.
      Merge the 2 lpc loops.
      ASF stands for "advanced stream format" not "adaptive ...", changing it to "ASF".
      Avoid segfault due to not set AVCodec.
      Field -> MBAFF direct mode can use data from fields that are not
      Fix CAVLC+8x8+MBAFF loop filter.
      Print which slices are IDR not just the slice type.
      Initialize next/last_picture* as the generic code does not always
      Do not use the generic "alloc missing references" code for h.264 as it does
      Do not free sps & pps on width/height changes.
      Ensure that no non-existing frame is used.
      Avoid out of picture accesses when CODEC_FLAG_EMU_EDGE is set.
      Prevent image2 from complaining about non monotone timestamps as
      2 in 1, an ASS and SSA demuxer.
      Do not lose the '\n'.
      ASS/SSA muxer.
      Forgotten changelog update for ASS/SSA (de)muxer.
      Check the 4th plane too when selecting the pixfmt.
      Simplify pix_fmt_id instead of listing a subset of non-simplified ones.
      Replace apparently always true condition by assert().
      Support loading presets from random paths.
      Rename opt_default() so it will not conflict with the one from ffmpeg.c
      Move opt_default() and set_context_opts() to cmdutils so it can be used from
      AVOptions support for ffplay.
      Guess picture type from picture coding extension when the main header is damaged.
      Fix infinite loop with spline, bug was introduced in r27612 by me.
      Print all cases that are tested, not just the ones that are bad.
      Revert mistakely commited hunk.
      Do not attempt to perform error concealment in lowres mode as this is
      Replace second (and wrong) call to avcodec_align_dimensions() by adjusting
      Allocate 1 line more in the chroma plane for H.264, this avoids some
      Move nal unit types enum and EXTENDED_SAR #define from h264data.h to h264.h.
      ff_parse_close() is not the correct function for H264Context.
      Fix indention of NAL type enum.
      100l forgot return type.
      Factorize stream id parsing.
      Favor even chunk sizes in ambiguous cases.
      Make gcc 2.95 happy.
      Limit discontinuity handling to containers that allow discontinuities.
      Add variable to hold the svq3 watermarking key,
      Export svq1_packet_checksum() that is needed for svq3 watermarks,
      Include zlib.h as it is needed for watermark support,
      Assume mono if no other information for raw.
      Append read data onto the buffer instead of overwriting, this ensures
      Move code setting delay and presentation_delayed a little up so
      Check for pts==dts on I/P frames in the presence of non low delay decoding and
      Use av_freep() in ff_parse_close().
      Port read_time() that works on x86_32 and 64 from noe.
      Do not attempt to use the unscaled yuv2rgb converter when height is odd because
      Remove rgb2rgb.h dependancy.
      rgb2rgb.h was not really intended to be a public header, thus remove it.
      Replace literally hardcoded max slice number by named constant.
      Check sub_packet_size against 0 to avoid div by zero later.
      Reset state after some frame header is found.
      Add udp.c maintainer.
      Fix segfault due to deallocated entries in delayed_pic after size change.
      Allocate parameter sets sanely instead of using alloc_parameter_set().
      Get rid of now unneeded pps/sps correctness code as pps/sps will not
      Fix regression test failure with pthreads on multiprocessor systems.
      ADTS AAC probe.
      add GEOX fourcc for mpeg4
      Implement complexity estimation parsing and try to detect an incorrectly set
      Fix FF_OPT_TYPE_INT64 support of defaults.
      Check that cpb_count is within the allowed range.
      remove _t for POSIX compatibility.
      Remove comment suggesting using variables ending in _t as this would break POSIX.
      s/qp_t/qp_tt/ to avoid bad examples that could lead to issues with reserved
      More POSIX _t namespace cleanup.
      Remove silly FIXME comment.
      Fix files with fps between 0.5 and 0.1.
      Fix non interleaved variable fps ODML avis.
      10l (forgot sample_size)
      Enable filter_mb_fast if CODEC_FLAG2_FAST is set.
      Do not skip frames until the first keyframe when stream copying but
      Fix bug where the avi demxuer lost the first frame of malformed
      Horizontal and vertical prediction functions needed for new lossless mode
      Support the new lossless mode.
      Do not recalculate constant part of is_complex for every MB.
      Move ENABLE_SMALL back to the per MB check, as otherwise gcc wont remove
      Factorize 3 multiplications out, code becomes 3 cpu cycles faster.
      Fix indention, also do a little vertical alignment of changed lines.
      Split filter_mb_dir() out of filter_mb().
      Replace != 0 || check by |
      Fix svq3 decoding, is_complex was not initialized.
      Skip chroma handling when there is no coded chroma.
      Skip non intra luma code when there is no coded luma.
      H.264 idct functions that include the chroma, inter luma and intra16 luma loops
      Use the new idct functions (except chroma as it was slower in benchmarks)
      indent
      Remove redundant nnz variable.
      Do not calculate idct_dc_add/idct_add when the variables are unused.
      10l, svq3 didnt set h->cbp, this broke decoding a little.
      Remove useless IS_8x8DCT check i forgot, spotted by dark shikari.
      Remove unreachable else clause, found by dark shikari.
      Remove if() surrounding decode_cabac_mb_type() that can never be true.
      Faster CAVLC decoding of trailing_ones. Based on a patch by dark shikari.
      s/h->cbp_table[mb_xy]/h->cbp/
      Reorder ifs in chroma hl_decode_mb to avoid a duplicate transform_bypass
      indent
      Move idct_(dc)add closer to where it is needed.
      Check ref values in CABAC H.264 for validity.
      Remove unacceptable NULL pointer hack from mc code.
      Optimize ctx calculation in decode_cabac_mb_mvd(), code by dark shikari.
      Simplify ctx calculation in decode_cabac_mb_dqp()
      Simplify ctx update in decode_cabac_mb_dqp().
      Replace /2 by >>1 in decode_cabac_mb_dqp()
      Fix decoding with the plain C idcts of
      Document new idct requirements.
      Optimize get_dct8x8_allowed().
      Only execute clear_blocks() when needed.
      Remove useless code.
      Fix detection of audio codec in K70707-ARIA229.flv.
      Move CODEC_FLAG_LOW_DELAY into ff_mpeg4_decode_picture_header().
      Fix PTS returned by the demuxers for mpeg4 without a decoder.
      Change mpeg audio parser so it only sets frame_size, channels and bit_rate
      Simplify if(), 3 cpu cycles faster in pentium dual.
      Simplify ifs(), 8 cpu cycles faster on pentium dual
      Negate a few variables, this simplifies the code and makes it 5 cycles faster
      Simplify if/else, no speed change
      Negate 2 more variables, 1 cpu cycle faster on pentium dual.
      inline decode_cabac_mb_type for I & P frames, 9 cycles faster on pentium dual.
      Indent
      Simplify decode_cabac_mb_ref() a little bit, 2 cpu cycles faster on
      Optimize 0 0 0-3 search, 45% faster on pentium dual.
      100l, I broke H.264 again, forgot one hunk.
      Make h264 parser 50% faster.
      unified CAVLC level decoding LUT.
      Indent
      Optimize esc removal code.
      Use the new VLC table for the first non trailing coeff too.
      indent
      Add get_ue_golomb_31()
      Update golomb decoding tables so that get_*_golomb_31() return some
      Use get_ue_golomb_31() where possible, almost all are just in headers
      Integrate get_te0_golomb() calls into the code, this allows some checks
      Indent
      Optimize pred_pskip_motion()
      Allow decoders to set AVCodecContext.refs.
      export refs
      CCITT Fax Group 3 and 4 decompression by kostya.
      Fix (rewrite) find_group3_syncmarker().
      Factorize mode switching in put_line().
      Avoid negation in put_line().
      Simplify zero run handling in put_line().
      indent
      10l != vs. == (yes, my fault not kostya's).
      Ensure that an invalid run terminates the line.
      Merge ff_ccitt_unpack_1d and 2d.
      minor simplification
      Close gaping sechole. That is, a series of run=0 allows arbitrary data to
      cosmetic (moving pix_left update)
      Factorize "avctx->width + 2" out to avoid someone mistakenly changing
      Get rid of pix_left in the 2d code, it is simpler that way.
      Get rid of check for condition that is always true (run_off < avctx->width).
      Move sync line pointers code to the end of the loop, it is simpler that way.
      Remove unused run_start variable.
      Minor simplification.
      Fix another sechole.
      Fix yet another sechole.
      And another sechole ...
      Simplify vertical mode
      indent
      Enable faxcompr.
      Add a list of GnuPG fingerprints.
      Explain the lack of +3/7
      Add spaces into the gpg fingerprint for better readability.
      Clarify that the list of gpg fingerprints is not just for maintainers.
      All non zero returns of rm_assemble_video_frame() are errors, check things
      Add a few error checks to rm_assemble_video_frame()
      Simplify the top of rm_assemble_video_frame().
      Merge videobuf and audiobuf.
      Replace buffer by AVPacket and avoid a memcpy() for video when the number
      Fix indention which was off by 1 space.
      10l set AVPacket.size to the true size of the returned data instead of
      Add enum AVSubtitleType
      Display subtitles within the display area.
      Change AVSubtitle.rects to an array of pointers so ABI does not break
      Future doxy for a few fields of AVSubtitleRect.
      Forgot to update (I likely missed it due to its similar name to dvb...)
      Replace some fixed width variable from AVSubtitle* by normal int.
      Replace AVSubtitleRect.rgba_palette and bitmap by AVPicture.
      Add type, text and ass to AVSubtitleRect.
      Bump LIBAVCODEC_VERSION_MINOR due to AVSubtitle changes.
      Generic metadata API.
      Move metadata API from avcodec to avformat.
      10l format include path for metadata.h
      Document the need of buf=NULL buf_size=0 at the end to obtain all frames.
      Do not try to duplicate packets that have data==NULL.
      Inject NULL packets into the video packet que at EOF to force all delayed
      All negative returns are errors not just -1.
      Check the return of audio and video encoders.
      All negative values are errors not just -1.
      fix indention
      Increase BUFFER_SIZE.
      Add CODEC_ID_MP1.
      Add dummy mp1_decoder to complement the existing dummy mp2/mp3 decoders.
      Let the mpeg audio parser correct the codec_id.
      Remove hack to print mp2/mp1 with codec_id of CODEC_ID_MP3.
      Avoid using sub_id in mpeg1/2video.
      Handle av_read_frame() EAGAIN return.
      Handle av_read_frame() EAGAIN return.
      Use EAGAIN return, primarely intended as example of EAGAIN useage.
      Move av_read_frame/parser related buffer variables from AVFormatContext
      Workaround ffms behavior of returning EAGAIN at EOF.
      size < 4 is invalid
      better header_size check
      Fix memleak of header in error returns.
      vertical align
      Fix memleak of fourxm->tracks on error return.
      Move current_track variable closer to where it is used.
      remove pointless {}
      Export all tracks (each is a different language) instead of just one.
      Add {} between if/else
      Prefer av_freep() over av_free() for variables in the context for safety.
      Fix remotely exploitable arbitrary code execution vulnerability.
      Fix unknown header 10001000 / http://samples.mplayerhq.hu/real/VC-RV10/thankyou.rm / issue809
      Fix probe & demuxing of Topfield 58XXPVR and 68XXPVR (topfield.rec)
      Reformat rdt_demuxer AVCodec struct sanely.
      Add size_in_bits to PutBitContext
      Ensure that the palette is set in data[1] for all 8bit formats.
      Fix list of input pix_fmts supported by GIF. This makes it work with swscale.
      Patch check script.
      Factorize grep -v cases.
      Fix reserved identifer detection so it doesnt detect __asm and such.
      avformat_seek_file() prototype
      Clarify that avformat_seek_file() is not part of the stable API yet.
      Clarify has_b_frames doxy.
      Move doxy for public resample2 functions to the public header.
      Try to doxyfy av_resample_init(). I hope I did not misremember anything as I
      Check for duplicate const/static/inline.
      Try to clarify doxy of avformat_seek_file().
      av_uninit() to suppress false uninitialized warnings from gcc without deoptimizing code.
      av_flatten to make the similarly named attribute available.
      Fix: libavcodec/msmpeg4.c:1612: warning: ‘dc_pred_dir’ may be used uninitialized in this function
      Let rgb32 pass for ljpeg.
      Silence "assdec.c:146: warning: passing argument 4 of ‘qsort’ from incompatible pointer type"
      Fix @param[in/out] handling
      Clarify get_ue_golomb_31() behavior with >31.
      A try to fix the regressions.
      Fix raw rgb/bgr vertical flip in avi based on info from http://www.fourcc.org/fccbihgt.php.
      indent
      Be less noisy with invalid pts/dts.
      Handle chunks with the MSB in size set correctly that is unsigned.
      Check for EOF during index reading.
      Check for alloc failures.
      Print all timebases (idea from netgem)
      Waste less space for printing timebases.
      Add note about big vs. little endian.
      Search for things like "int *src" without const.
      Compact repeated messages to "Last message repeated x times".
      Fix vc1 split().
      Allow av_find_stream_info() to be aborted.
      Approved hunks for VAAPI / our new shiny hwaccel API
      Pixel Format descriptor, try #1.
      AVPixFmtDescriptor doxy
      Add CODEC_ID_H264 to tb_unreliable(), it belongs there for the same
      Add AVFMT_VARIABLE_FPS to specify which muxers do not need duplicated frames.
      Do not duplicate frames if the muxer claims that it is not needed.
      Add and use ff_pixfmt_list_420.
      Another AVhwaccel hunk from Gwenole Beauchesne.
      av_register_hwaccel() from Gwenole Beauchesne.
      av_hwaccel_next() by Gwenole Beauchesne.
      ff_find_hwaccel()
      add ff_find_hwaccel() by Gwenole Beauchesne
      1000l for me, wrong pix_fmt.
      AVInputFormat.reed_seek2()
      untypo
      More AVhwaccel hunks from Gwenole Beauchesne.
      Add hwaccel->decode_slice() call for avhwaccel by Gwenole Beauchesne.
      Fix next_pts computation to consider repeat_pict.
      Half user provided time_base once during the first frame when there is no
      Add ticks_per_frame, this should hopefully fix the regressions caused
      Fix SVQ3 regression since ticks_per_frame=2.
      Update commented fprintf().
      Make sure -vsync 2 does not produce timestamps equal to the previous,
      Fix misdetection of #else in  'missing } prior to else'.
      Improve detection of non doxy comments.
      Detect static prototypes.
      Adjust time_base exactly when changing ticks_per_frame to 2 this is safer
      Do not set low_delay if has_b_frames has been set before init()
      Disable MPEG-1/2 style timestamp calculation for H264. It still randomizes
      Change the timebase of the raw demuxer to one that can represent the ts of fields.

Michael T. Dean (1):
      Missing LDFLAGS in libavcodec/Makefile patch by ("Michael T. Dean" <mtdean at thirdcontact dot com>)

Michel Bardiaux (48):
      configure on SGI IRIX patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      av_log() patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      cvsignore patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      seek by mouse in ffplay patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      suppress PTS in packets when not needed (slightly smaller files), fixed PTS generation in some cases, added provision for DTS generation, slightly better SCR generation (initial patch by Michel Bardiaux)
      av_log() patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      compilation of mov.c broken patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      Help debugging by keeping symbols ans line numbers patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      Help debugging by keeping symbols ans line numbers patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      Doc improvement for -target patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      config options log patch by (Michel Bardiaux <mbardiaux at peaktime dot be>)
      better error message patch by (Michel Bardiaux <mbardiaux peaktime be>)
      Remove stray non-existent -o option from an example.
      fix some pointer to intger without cast warnings (patch by Michel Bardiaux)
      Fix some #ifdef to allow building amr_wb alone.
      Add -Wdeclaration-after-statement to CFLAGS on GCCs that support it.
      Use != instead of -ne in test invocations for greater portability.
      Update the FAQ with the current status of image formats.
      Clarified API for numbered sequences, patch by Michel Bardiaux % mbardiaux A mediaxim P be %
      Fall back on grepping for the revision number if no svn client is installed.
      Fix wrong comment, patch by Michel Bardiaux  mbardiaux mediaxim be.
      Segregate code common to BMP decoder and future encoder
      Fix segfault in bmp decoder. Patch by Michel Bardiaux mbardiaux mediaxim dot be.
      Make BMP decoder use bytestream. Patch by Michel Bardiaux
      Add some context to av_log calls.
      Fix memcpy out-of-bounds.
      Activate guards in avcodec_default_get_buffer. Patch by Michel Bardiaux,
      Add decode_end method to bmp decoder. Patch by Michel Bardiaux,
      Force stereo in target vcd.
      Add FACT chunk for non-PCM wav formats
      Regression changes after addition of fact chunk to non-PCM wav
      Change some dprintf to av_log
      Fix one warning
      Spelling fix
      Spelling fix
      Remove some old disabled debugging output
      Add support for MS-GSM codec
      Remove code duplicated between adpcm.c and riff.c
      Supply context to tprintf
      Rename WMADecodeContext to WMACodecContext
      Corrections so that builds with DEBUG work
      Give context to dprintf
      Clarify comments about PIX_FMT_MONOWHITE/MONOBLACK
      Print pixel and display aspect ratios
      Suppress runaway index on streamed input
      Clarify the error message when video width, height, or framerate are not
      Allow bitrates zero and 13200 (needed for decoding mov and aiff)
      Make sure some value is always returned via data_size

Mike Brown (1):
      changes to ignore command media embedded in MS WMV files patch by ("Brown, Mike": mikeb, vibephone com)

Mike Frysinger (4):
      fall back mmap() call using MAP_PRIVATE for working on no-mmu systems
      The Blackfin toolchains will automatically select FDPIC according to the
      FLAT objects cannot have multiple sections, so using the L1 attributes breaks
      FLAT objects cannot have multiple sections, so using the L1 attributes breaks

Mike Melanson (249):
      register cyuv with allcodecs.c
      hook this up to the CYUV decoder
      native Indeo3 decoder implementation
      code redundancy reduction, courtesy of suxen_drol -at- hotmail.com
      first pass at a new VP3 video decoder
      fix decoder so that ffmpeg does not crash, at least not right away
      fixed buffer allocation logic (hopefully) so that decoder does not crash
      looking better all the time! motion compensation is starting to work
      correct the custom coding mode alphabet, add some validation on the
      change the way the ImageDescription is passed to the decoder
      dump the shady binary search logic (the part that binary searches
      squashed a bunch of subtle array indexing bugs, fixed block mapping
      get those permutations straight
      fix motion vector decoding bug and reinstate interframes
      fix image buffer leak on keyframes, add more error condition checks
      fix subtle logic problem in block unpacker that leads to incorrect token
      first pass at an experimental 4xm demuxer
      add 4xm to the family
      rework SVQ1 decoder to use more intuitive VLC tables as well as ffmpeg's
      allocate enough bytes
      placeholder IDs for 4xm A/V codecs
      dispatch video as well as audio
      get the video dispatch straight
      overdue format updates
      revising and fixing motion vectors, squished block unpacking bug that
      fix identifier to fix compilation
      added support for B-frames and multiple slices
      4, not 2
      don't step on the context
      Fix third pixel motion compensation
      - fix PLANE_PRED8x8 prediction (H/V are swapped, this is correct!)
      added the official VP3 IDCT (C implementation) as well as a grayscale
      endian-conscious fix, courtesy of Sebastian Jedruszkiewicz
      reverse earlier patch
      naughty, forbidden variable declarations
      skip preamble bytes
      first pass at PSX STR demuxer; does not yet interact correctly with MDEC
      initial commit for Id RoQ and Interplay MVE multimedia subsystems
      doc update
      it seems that ffmpeg can also mux QT/MP4 now
      New fringe codecs: WC3/Xan video, Xan DPCM, DK3 & DK4 ADPCM
      demuxer for custom Wing Commander III movie (.mve) files
      support for Duck DK3 and DK4 ADPCM
      hacks to support Xan AVI files
      fringe feature update
      Tim
      forgot to apply the palette component lookup table
      xan_wc3 decoder now works correctly; added a bunch of output
      copy & paste strikes again
      deploy palette control API
      final word on RGBA byte order
      change pixel identifier conflict with AltiVec headers (patch courtesy of
      Ministry of English Composition fixes (courtesy of Diego Biurrun
      adapt to use interim palette control API; do away with multi-chunk data
      initial Interplay video decoder
      wrong, wrong, wrong...coded bits are read in the opposite direction
      video looks beautiful now, many thanks to Alexander Belyakov
      provisions for raw PCM audio
      do away with all those naughty global variables
      MACRO-ize a bunch of redundant code blocks; fix 16-bit RGB modes (it's
      fix Interplay DPCM (frames are intracoded, predictors do not carry
      iterate through the appropriate range of palette entries
      clean up 4xm demuxer; make valgrind just a little happier
      send the chunk preamble bytes to the respective decoders; Id RoQ demuxer
      New demuxers: Sega FILM/CPK, Westwood VQA & AUD; new decoders: MS RLE &
      the correct data type is offset_t
      properly demux silent files; implemented precise framerate calculation
      experimental half-pel motion compensation for C planes, courtesy of Dr.
      first pass at a VQA video decoder
      implemented Id Quake II CIN support
      instructions for using the decoder
      handle a wider variety of VQA files, including silent files
      added solid color vectors; basic PAL8, 4x2-vector video (as in
      This patch marks frames used in RPZA decoder as reference frames, from
      support a few more types of VQA files
      optimised versions of the get_pixels, diff_pixels, and bswap_buf
      disable encoders where appropriate (patch courtesy of BERO
      added CRI ADX and XA ADPCM audio modules, courtesy of BERO
      remove experimental reference
      use the correct file modes
      quiet
      this should be part of adpcm.c
      replace file to remove nasty DOS CRs and hard tabs
      activate the XA and ADX ADPCM codecs
      hook up the XA ADPCM decoder
      added XVID fourcc (ISO MPEG-4?) and AC3 fourcc
      support Cinepak files with funky (not divisible by 4) resolutions
      revised palette API, courtesy of Roberto Togni (rtogni at freemail.it)
      added Apple Graphics (SMC) decoder
      added palette support to the QT demuxer; registered SMC video decoder
      yep, FLI support
      use the proper file framerate (specified by a float); account the pts
      set frame rate information, for good measure
      off by 1 in the palette
      initial support for Duck TrueMotion v1 (think of it as On2 VP1); only
      DUCK codec
      this should make the decoder safe for big-endian platforms
      initial predictors are not sent to the output in QT IMA; fix stereo QT
      SVQ3 uses unrestricted motion vectors
      remove vile lie (MS ADPCM encoding is not supported yet)
      off-by-1 error in the never-before-tested embedded string code
      first pass at Sierra VMD A/V decoders; video looks great, audio is not
      initial commit for Sierra VMD file demuxer
      VMD support
      handle 0-length buffers
      handle 0-length buffers
      watch those hard tabs
      handle invalid buffer sizes
      these functions (get_bit_count and align_put_bits) are referenced even
      vsad16_mmx2 only applies if encoders are turned on
      added support for EA ADPCM and SMJPEG IMA ADPCM
      initial commit for Quicktime Animation (RLE) video decoder; bit depths
      seperated out the C-based VP3 DSP functions into a different file; also
      cut over to using new VP3 DSP functions and remove the old ones; bring
      use optimized VP3 functions where appropriate
      Matroska credits
      added SGI image format, encoding and decoding, courtesy of Todd Kirby
      credit update
      smarten up the SVQ3 extradata decoder without changing the external API
      be smarter about the variable size of the extra SVQ3 data
      remove numerous definitions of BE_*/LE_* macros; convert FOURCC_TAG ->
      minor VMD system update; still not perfect, but should not crash either
      don't fetch the size of SEQH unless SEQH was found
      update the API for the idct function
      added QT faststart remuxing utility to ffmpeg distribution per, ahem,
      patch courtesy of Todd Kirby:
      add the minimal documentation to make this decoder useful
      correct MMX-optimized variant of VP3 IDCT, with comments (thank you
      SSE2-optimized variant of VP3 IDCT
      add selected SSE2 ASM macros
      hook up support for SSE2-optimized VP3 IDCT
      hush...
      move __align16 some place where non-MMX machines can see it
      reorganize and simplify the VP3 IDCT stuff
      yes, that's nice...
      separate out put_signed_pixels_clamped() into its own function and
      move the 0x80 vector outside of the function, thus saving the compiler
      support SVQ1 in AVI files, just for fun
      first pass at a Sorenson Video 1 (SVQ1) encoder
      how to encode SVQ1
      no longer true
      rpza resolution fix, courtesy of Todd Kirby (doubleshot at pacbell.net)
      attempt to create some separation in the FLAC system with respect to
      fix initialization bug in which execution overwrites essential data
      the ongoing struggle for encoding conditionals
      warning vigilance
      cut-n-paste error slipped through the cracks... thanks to DJ Hartman for
      sweeping change from -EIO -> AVERROR_IO
      cvs diff before cvs commit
      add some expanded information when printing the banner, and always print
      use a more portable version macro
      yep, tyop
      well that does not need to be there anymore
      TechSmith Camtasia (TSCC) video decoder, courtesy of Konstantin Shishkov
      build++
      8-bit mode corrections
      added support for the elusive AVI palette change chunk, courtesy of
      IBM Ultimotion video decoder, courtesy of Konstantin Shishkov
      if audio is PCM, push it through (sync is not perfect yet)
      support for Sierra Online audio files and Apple QuickDraw codec,
      FAQ update; entry on C++
      Creative ADPCM decoder, format 0x200, courtesy of Konstantin Shishkov
      FAQ entry for manually changing fourcc
      Electronic Arts Game Multimedia format demuxer (WVE/UV2/etc.)
      Miro VideoXL (VIXL) decoder, courtesy of Konstantin Shishkov
      H.261 stuff
      credited with H.261 encoder
      native QPEG video decoder, courtesy of Konstantin Shishkov
      new NSV demuxer
      use correct types
      load the 'avcC' atom needed for 'avc1' decoding
      doc update for Shorten codec
      go LOCO, courtesy of Kostya Shishkov
      VP3 post-processing loop filter; disabled until the correct final step
      first pass at ALAC decoder from David Hammerton; while David's original
      do not use a variable before proper initialization
      decoder works fine now, when fed properly-sized chunks by the demuxer;
      get rid of nasty exit()'s so the decoder will only warn the user before
      replace custom bit-reading functions with FFmpeg's internal function;
      Winnov WNV1 video decoder, courtesy of Konstantin Shishkov
      Westwood SND1 decoder, courtesy of Kostya
      free that last frame
      Autodesk Animator Studio Codec (AASC) video decoder, courtesy of
      remove superfluous CONFIG_ENCODERS (entire function is already wrapped)
      Ministry of English Composition, reporting for duty (and the word is "skipped", not "skiped"; "skiped" would rhyme with "hyped")
      Fraps FPS1 video decoder (v1 & v2), courtesy of Roine Gustafsson <roine
      signature has an 'x'
      use VLCs for in place of get_fragment_run_length(), get_mode_code(), and
      replace get_superblock_run_length() with a VLC table
      fix 4MV MC and reinstate new MC VLC
      replace unpack_token() with a series of lookup tables
      read Huffman tables from Theora header (patch courtesy of
      another bitstream extraction to optimize: fixed-length motion vectors
      clean up some warnings and dead code
      render by the slice rather than the plane
      apply the loop filter to fragments as they are rendered into the final
      fix bugs in new loop filter code; also, refrain from filtering against
      read loop filter limit values from Theora header, courtesy of Matthieu
      this file got skipped during the API update
      tinfoil patch: make sure that pixel pointer does not go out of bounds
      tinfoil patch: accept no widths that are no multiples of 4
      tinfoil patch: make sure pixel_ptr never goes negative
      tinfoil patch: validate motion vectors and do not free frame on exit if
      tinfoil patch: be more diligent about checking array boundaries before
      removed a bunch of unused, sub-optimal code
      tinfoil patch: no array is written to in bulk before counts are
      automatically detect Cinepak data from Sega FILM/CPK files
      Duck TrueMotion 2 video decoder, courtesy of Konstantin Shishkov
      support for FLX and DTA extensions in the FLIC format, courtesy of
      documentation update
      gcc 2.95 fix, courtesy of Luca Abeni
      TM2 fixes by Kostya
      chroma decoding fix by Kostya
      improvements by Reimar Döffinger; 24-bit decoding is not perfect, only
      correctly handle very large Cinepak frames (courtesy of John Koleszar
      correctly deal with the alpha channel in 32-bit QT RLE (courtesy of John
      ugly, but necessary
      complete American Laser Games MM playback system, courtesy of Peter Ross
      dosbox native ZMBV decoder, courtesy of Kostya
      make TSCC endian-safe, PPC testing courtesy of Diego B.
      add another set of zlib guards
      complete AVS playback system (from Creature Shock computer game),
      printf -> av_log, only occurs in DEBUG cases
      Simplify the Sega FILM/CPK demuxer to not modify the bastardized Cinepak stream.
      Another hack to allow the Cinepak decoder to detect both types of deviant Cinepak
      Fix issue 123 (Support Magic-carpet fli version,
      simplify PTS handling
      Check sanity in the palette loading operation. The addresses a potential security risk in
      adjust indentation on latest patch
      Error checking: make sure that there are 3 parameters and that the
      Ministry of English Composition edits for SNOW spec-- introduction (many more to come)
      Ministry of English Composition edits
      clarify previous revision on optimization justification
      Use a consistent capitalization for 'Snow' as discussed on the list
      minor English corrections
      fix the id CIN demuxer by using the pts API correctly
      fix wc3movie demuxer to conform to the PTS API
      fix the WC3 Xan decoder by correcting the frame accounting logic
      recognize the CLJR FourCC (Cirrus Logic AccuPak)
      There is no reason to assign a variable to itself
      These video decoders do not need to include and initialize the DSP
      Ministry of English Composition edits
      Ministry of English Composition edits, courtesy of the Wanderer
      Tighten up the Westwood AUD detection. Probability of random detections
      correct build documentation for qt-faststart utility
      add 'ACDV' FourCC (which is just complete JPEG files in individual AVI
      The POSIX namespace shall be held sacrosanct. To that end,
      undoing revision 16118 for truemotion1.c; commit was premature
      Ministry of English Composition treatment for the new metadata API

Milan Cutka (3):
      Quantizer noise shaping crashes on SSE2 machines patch by (Milan Cutka <cutka at szm dot sk>)
      When bswap_32 is a macro, png images fail to decode properly, patch by (Milan Cutka <cutka>at<szm>dot<sk>)
      Fixed decoding of 410p ffv1 files patch by (Milan Cutka <cutka szm sk>)

Morten Hustveit (2):
      Define rc override string as const.
      Fix prototype and return value of opt_bitrate and opt_bsf.

Måns Rullgård (799):
      c syntax update patch by (mru at users dot sourceforge dot net (Måns Rullgård))
      -vcodec copy fix patch by (mru at users dot sourceforge dot net (Måns Rullgård))
      fixes the frame number counter for -vcodec copy patch by (mru at users dot sourceforge.net (Måns Rullgård))
      c standard style / warnings patch by (mru at users dot sourceforge dot net (Måns Rullgård))
      C99 initializers and kill warnings patch by (mru at users dot sourceforge dot net (Måns Rullgård))
      makes configure create the libpostproc directory and
      Write correct MPEG2-PS streams patch by (mru at users dot sourceforge dot net (Måns Rullgård))
      support more codecs in MPEG-TS patch by (mru at kth dot se (Måns Rullgård))
      Get H.264 frame rate from SPS/VUI patch by (Måns Rullgård <mru at kth dot se>)
      untested AAC in WAV/AVI patch by (Måns Rullgård <mru at mru dot ath dot cx>)
      credits patch by (Måns Rullgård <mru at mru dot ath dot cx>)
      kill warnings patch by (Måns Rullgård <mru inprovide com>)
      improved detection of "AVC1" style H.264 patch by (Måns Rullgård <mru inprovide com>)
      H.264 and Vorbis support in matroska patch by (Måns Rullgård |mru inprovide com)
      double free patch by (Måns Rullgård <mru inprovide com>)
      Oops patch by (Måns Rullgård <mru inprovide com>)
      uuid patch by (Måns Rullgård <mru inprovide com>)
      set constant QP from AVCodecContext.global_quality.
      argument name for -vdt
      B frames and CABAC/CAVLC selection by Erik Slagter (erik at slagter dot name)
      indentation fix
      PSM support in MPEG-PS demuxer.
      Ogg demuxer ported from tcvp by Luca Barbato <lu_zero at gentoo dot org>,
      pre-c99 compatibility
      s/u_char/uint8_t/
      s/u_char/uint8_t/
      fix seeking bug
      s/u_char/uint8_t/
      Fix codec list when running configure outside the source directory.
      support theora in ogg, plus required ogg core changes
      theora decoding using libtheora
      flac in ogg support
      check theora version
      allow various x86 models with --tune
      change extradata format for vorbis
      set correct source path when running configure using relative path
      remove/replace non-ascii characters
      work with latest svn of x264
      add VP30 fourcc
      fix shared libpostproc build
      fix decoding of multi-channel vorbis streams
      accept old extradata format
      don't try to find file length if streaming
      enable x264 multi-threading
      add -gray command line flag
      fix segfault if flush is called before the first frame
      remove mixed code/declarations
      fix building in separate directory
      kill a bunch of compiler warnings
      remove junk characters
      Kill some compiler warnings.  Compiled code verified identical after changes.
      utf8
      no mixing of code and declarations
      revert incorrect change
      use $CFLAGS and $LDFLAGS in dlopen() check
      BMP image decoder
      ogg format probe function by Ivo <ivop at euronet nl>
      alpha needs -fPIC for shared libs
      use utf-8 encoding
      remove unused variables
      fix const warning
      use PRIxN, %zd, %td formats where needed
      interpret H264 VUI timing info correctly
      place SPS/PPS in extradata if GLOBAL_HEADER flag is set
      native ac3 parser
      remove unused function get_longterm_datarate()
      decode adts aac streams
      adts aac parser
      read/write adts aac
      create adts stream if CODEC_FLAG_GLOBAL_HEADER not set
      bump version numbers and update docs for adts aac support
      use get_bits1/skip_bits[1] in ac3 and aac parsers
      generate adts frame headers if required
      add adtsenc.c with myself as maintainer
      move common parts of makefiles into common.mak
      add quotes around $(libdir)
      Create installation directories in the common install-* targets.
      kill some warnings
      SRCS must be set using := or things break when ASM_OBJS is added to OBJS
      can't have PES headers in MPEG video elementary streams so fail probe
      since Michael prefers to keep dsputil.h private to libavcodec
      make install target depend on vhooks
      be nice to syntax highlighting that doesn't understand here-docs
      fix memory leak on mpegts close
      This patch simply adds demuxing support for AC-3 streams in DVB TS
      #define [U]INT32_{MIN,MAX} if missing
      s/SAMPLT_FMT_NONE/SAMPLE_FMT_NONE/
      add newline at end of file
      Create shared libraries using versioned filenames and add -rpath-link
      ogm demuxing
      add unaligned16() and unaligned64()
      set stream time_base properly
      make configure --whatever --help work as expected
      major cleanup of compiler/library tests
      10l to Diego
      remove redundant make variable SUBDIR
      apply $cross_prefix before $cc is used
      indentation
      fix seek related bugs
      probe for mpeg audio
      use ifeq(...,yes) instead of ifdef
      don't use xlib in ffplay, and bump required sdl version
      10l to myself for being overzealous with $cross_prefix
      MIPS doesn't like unaligned accesses
      check for SDL_VideoInfo.current_[wh] availability in configure, and
      free codec extradata in av_close_input_file
      test availability of various compiler flags, and verify that requested
      The AAC frame header uses 13 bits for the frame size, so the buffer should
      use standard codec tag if the specified tag is out of range and would be
      improved logging tests
      put local -L flags first in link commands
      add myself as build system maintainer
      libfaac uses pascal calling convention on win32 so the generic tests
      allow packets with non-zero PES_scrambling_control
      increase search limit in av_find_stream_info
      allow recent libfaad versions with new function names
      enable logging in configure by default, and print some advice on failure
      fix a silly typo (1l to myself)
      -logg needs to be after -lvorbis
      cleaned up buffer management
      10l: don't allocate a new buffer quite so often
      s/u_char/uint8_t/ accidentally copied from tcvp
      get svn version when using separate build tree
      detect MPEG PES streams as MPEG PS; the PS demuxer will cope
      do not fail av_find_stream_info() if params for all streams could not be found
      detect audio-only program streams and broken files made by VDR
      extradata may be used as bitstream reader input so allocate with padding
      detect availability of -rdynamic and -export-dynamic
      send output of exec checks to log file
      add --cross-compile option to assume cross-compilation without cross-prefix
      compiler sanity check
      fill_mbaff_ref_list is missing a return statement, its return value
      get rid of veccpy(), replace with simple assignment
      remove unneeded casts
      kill warnings
      kill warnings
      1l: put that int i back
      kill warning
      remove redundant macro definitions
      attempt to detect Chinese AVS video
      solaris sh doesn't allow functions and variables with the same name
      Try to find a better shell if the one configure was started under is
      floorf() is not used, and causes warnings on Solaris
      set flags needed for shared libs on solaris
      #define SBUTTERFLY outside CONFIG_ENCODERS
      set CONFIG_ENCODERS/CONFIG_DECODERS only if any encoders/decoders are enabled
      use $(LIBOBJFLAGS) when compiling sparc/*.c
      build mpegaudio.c and ac3enc.c only if needed
      simplify
      remove -fforce-addr as it to generate slightly slower code
      Sun mlib works fine.  Remove comment stating otherwise.
      rewrite kconfig-style
      cosmetic: sort codec list alphabetically
      enable/disable pcm and adpcm codecs
      make the adx adpcm codec available
      forgot to remove adx.o from objs always built
      only build dpcm.o if its codecs are enabled
      10l: make disabling all pcm/adpcm codecs work
      --enable/disable parsers. Warning: some combinations are broken.
      make some tables const
      give AVInput/OutputFormat structs consistent names
      allow individual selection of muxers and demuxers
      10l to Diego for disabling all codecs
      find all codecs, even the ones with numbers
      log stderr of test programs in check_exec()
      another round of cola, please
      tidy up option parsing
      simplify list processing
      move common stuff from avienc.c and wav.c to new file riff.c
      nut needs crc.o
      split compile and link in check_ld
      static libvorbis needs -logg
      rename AMR config variables to CONFIG_AMR*
      add lost raw pcm input/output formats
      move adler32 to libavutil
      cleanup suggested by Michael Niedermayer
      replace -D_GNU_SOURCE with -D_ISOC9X_SOURCE to avoid accidental use
      remove redundant #include <math.h>
      remove old msvc cruft
      remove impossible #ifdef case
      simplify
      simplify
      indent
      set defaults before showing help
      cosmetics
      can't trust the compiler exit status, check for messages instead
      ignore explicit #warnings in compiler tests
      remove libtheora wrapper
      checking compiler output was a bad idea, revert to checking exit status
      simplify command logging
      detect more broken shells
      remove redundant declarations
      inline align_get_bits()
      fix out of tree builds
      add our -L flags before any other linker flags
      ffmpeg, ffplay, ffserver need version.h
      clean up #ifdef HAVE_IPP
      remove junk characters in comment
      fix MUL16 macro
      ARMv5TE optimized IDCT
      remove useless memset()
      fix some signedness warnings
      add some #ifdef CONFIG_ENCODERS/DECODERS
      try to stop people blindly using --cross-compile
      put ARM feature tests next to other CPU tests
      make sure we can create and execute files in $TMPDIR
      return 0 from encode_picture() and estimate_qp() on success
      dc_val should be signed
      remove extraneous #ifdef CONFIG_ENCODERS/DECODERS
      f*cking windows confuses itself by adding .exe to filenames
      libdts and lame sometimes need -lm
      coded line size is a multiple of 4
      CODEC_CAP_DELAY support
      indentation
      allow spaces in source and build directory names
      rename HAVE_VHOOK/BUILD_VHOOK to CONFIG_VHOOK
      rename CONFIG_SDL -> HAVE_SDL
      check for byteswap.h
      disable -Winline spam, --enable-extra-warnings to enable
      make two functions static
      make some symbols static
      make some symbols static
      fix make dep for out of tree builds
      make vp6_pcr_tree static
      adding more static keywords
      make some tables static
      #define ENABLE_*_CODEC etc as 0 or 1 in config.h
      fix make dep with some gcc versions
      rename inverse -> ff_inverse
      rename cropTbl -> ff_cropTbl
      rename squareTbl -> ff_squareTbl
      make more tables static
      zigzag_direct_noperm doesn't exist, remove declaration
      remove filter_out() function, it's unused and not POSIX compliant
      rename CONFIG_AC3 to CONFIG_A52
      rename shell variable netserver to beos_netserver for consistency
      rename shell variable vorbis to libvorbis
      remove unused option --disable-simple_idct
      rename CONFIG_HAVE_DLFCN and CONFIG_HAVE_DLOPEN to HAVE_DLFCN/HAVE_DLOPEN
      rename shell variable _malloc_h to malloc_h
      rename shell variable _memalign to memalign
      rename shell variable sunmlib to mlib for consistency with CONFIG_MLIB
      rename shell variable _altivec_h to altivec_h
      rename shell variables v4l[2] to video4linux[2]
      factorize config.h/config.mak writing
      rename MEMALIGN_HACK to CONFIG_MEMALIGN_HACK, use common code in configure
      fix HAVE_LRINTF setting
      break CONFIG_LIST and HAVE_LIST onto multiple lines
      move AMR notices next to other messages
      clean up setting of HAVE_*THREADS
      beautify
      move makefile linking after config.* creation
      simplify
      move comment where it makes sense
      move some CFLAGS settings away from config.* writing section
      revert bad checkin
      move some CFLAGS settings away from config.* writing section
      indentation
      merge CONFIG_WINCE sections
      simplify
      check for inet_aton in configure
      clean up inttypes.h and int_fastxx_t types detection
      remove workaround for missing inttypes.h
      move comment
      improve broken shell detection, and give a better error message
      rename HAVE_DLFCN to HAVE_DLFCN_H
      set status variables automatically in check_header and check_func
      gcc always #defines __MINGW32__ on mingw targets
      set CONFIG_OS2 using new method
      add copyright notices for myself and Diego
      simplify [TARGET_]ARCH_* setting
      TUNECPU is never used
      improved pthreads test
      remove 'restrict' definition, it is always #defined in config.h
      move some __attribute__ macros to internal.h
      move more macros to internal.h
      use the standard INT64_C() macro for 64-bit constants
      use standard INT64_MAX instead of MAXINT64 (and MIN)
      make shell test work with Solaris /bin/sh
      do not use non-standard test -nt
      fix mingw shared build, and get rid of FF_IMPORT_ATTR
      merge #ifdef HAVE_AV_CONFIG_H sections
      BUILD_SHARED_AV is no longer used
      rename always_inline to av_always_inline and move to common.h
      lavc is at major version 51, delete old stuff
      simplify test for memalign emulation requirement
      OpenBSD settings by Nikns Siankin <nikns at secure lv>
      fix typo in Darwin shared lib naming
      improved stack misalignment warning
      replace bashism with POSIX construct
      don't use non-standard sed -i option
      fix libavutil.pc
      make version.h depend on .svn/entries
      allow --cpu=core2 for Intel Core2 tuning with gcc 4.3 snapshots
      merge fragmented mpeg4adv section in regression.sh
      save/restore number of streams
      check for sane values in ogg_get_length(), fix lol-ffplay.ogm fuzz test
      update swf regression checksum
      check dependencies before setting CONFIG_*/ENABLE_*
      remove no longer needed #ifdefs
      remove no longer needed ifeq tests
      reorder alphabetically
      it's called libgsm, not just gsm
      remove useless #ifdef CONFIG_ZLIB
      remove useless #ifdef CONFIG_ZLIB
      remove the last #ifdef from allcodecs.c
      remove unused files
      add -Wno-pointer-sign to CFLAGS
      check for ARMv6 instruction set support
      simplify setting of TARGET_* variables
      rename CMOV_IS_FAST to HAVE_FAST_CMOV and simplify configure
      nobody uses TARGET_CMOV
      ARMv6 SIMD IDCT
      theoretically save one cycle
      typo fix by Dmitry Antipov dmantipov =a= yandex =d= ru
      optimize IDCT of rows with mostly zero coefficients
      remove redundant loads
      save one cycle
      revert accidental checkin
      missed a redundant load
      cosmetic: indent sensibly so code can be read at all
      cosmetics: don't needlessly open new blocks
      remove preprocessor cruft
      move static variables to private context struct
      fix multichannel decoding
      oops, revert accidental checkin
      reorder setting of defaults by category
      indent
      simplify video4linux2 test
      test for video4linux
      remove extraneous per-OS disabling of video4linux[2]
      test for OSS audio
      #include detected soundcard.h instead of testing for OpenBSD
      disable bktr if required headers not found, enable by default
      prettify texi2html check
      check for gnu make
      simplify altivec warning
      rewrite some one-line if/then/fi using && and ||
      do nothing in restore_flags() if flags are not saved
      simplify freetype check
      simplify imlib2 check
      don't skip freetype2 check on beos
      simplify SDL check
      tidy up ffserver deps
      let check_deps handle ffplay
      recursive feature dependency checking
      handle network dependency on protocols with check_deps()
      finish feature calculations before printing result and creating files
      rename HAVE_GPROF to CONFIG_GPROF
      rename some variables to match corresponding --enable-* options
      move pkg_requires setting to end
      add armv6 to TARGET_LIST
      clean up amr option handling
      simplify --enable/disable-* option handling
      10l to Diego for breaking A52 decoder
      fix --enable-*threads
      rename some --enable options for external libs
      rename POWERPC_PERFORMANCE_REPORT to CONFIG_POWERPC_PERF
      add powerpc_perf to CONFIG_LIST, rename --powerpc-perf-enable option
      simplify ever so slightly
      combine handling of --enable/disable-encoder/decoder/muxer/demuxer/parser
      rename video4linux[2] variables to v4l[2], consistent with --disable options
      simplify
      restore lost --enable-amr-if2 option
      add --target-os option, and move OS specific settings after command line
      put mingw section together with other OS specifics
      move arch processing after command line loop
      mention --target-os in --help
      rearrange --help text slightly
      10l: I thought I'd tested all changes...
      case insensitive OS name
      remove --enable-mingw32 option, --target-os=mingw32 should be used
      TARGET_OS in config.mak is unused
      fix for x86-64
      simplify check_lib
      revert last change, it breaks on brain damaged distros
      feature tests for functions requiring one or more headers
      generic foo-config --cflags/libs test
      libgsm_ms codec depends on libgsm
      remove [U]INT64_C definition
      simplify
      simplify conditional targets
      remove useless OBJS variable
      write DTS, based on patch by Luca Abeni, lucabe72 email it
      include only files actually built in SRCS
      fix init_vlc() with nonzero flags
      run ebp register check in subshell so we can redirect crash message
      trap signals in tests that can crash
      add .type foo, %function directives for the benefit of debuggers
      replace TARGET_ARCH_* to ARCH_* for consistency
      add x86 to ARCH_LIST simplifying config file generation
      replace TARGET_* with HAVE_* simplifying lists in configure script
      merge print_config_enable() and print_config() as only the former is used
      test whether linker supports --warn-common
      move dsputil_init_foo() out from #ifdef
      kill a warning on 64-bit machines
      kill some "defined but not used" warnings
      correct type for {put,avg}_h264_chroma_mc2_mmx2
      fix some printf format specifiers
      replace endian detection hack with #ifdef WORDS_BIGENDIAN
      fix indentation
      correct type for h263_parse()
      correct type of qsort() comparison callback
      move definition of MB_TYPE_H261_FIL to h261.h so h261.c doesn't
      move a couple of macros and structs to mpegaudio.h
      include all prerequisites in header files
      add multiple inclusion guards to headers
      add "checkheaders" target
      simplify simd extension checking
      indent
      remove unneeded variable initializations
      simplify ppc64 handling
      fix report at end of configure
      fix thread support selection
      extract revision from svn-1.4 .svn/entries files
      kill error messages from failed svn revision extraction commands
      move lrintf() check before external lib checks to prevent runtime
      place -l flags after object files in check_ld()
      simplify strstart()
      simplify pstrcpy()
      cosmetic: combine declaration and initialisation
      simplify stristart()
      remove useless #ifndef CONFIG_NOCUTILS
      add some string functions
      use new string functions
      remove old unused string functions
      replace the uses of old string functions that Reimar missed
      replace strncpy with av_strlcpy
      use more Vorbis metadata tags
      use AV_RL32
      whitespace cosmetics
      warnings--
      remove duplicate #include
      use av_freep() in free_vlc() and failed init_vlc()
      update my email address
      handle --enable-small same way as other CONFIG_* options (almost)
      remove unused functions
      kill uninitialised variable warning in ac3_probe()
      kill uninitialised variable warning in amr_read_packet()
      add 64-bit bytestream read/write functions
      whitespace cosmetics
      use bytestream_get_* and AV_RLxx
      remove redundant codec_get_bmp_tag() call in avi_write_header()
      nutdec: don't use deprecated codec_get_{bmp,wav}_id()
      ogm: don't use deprecated codec_get_{bmp,wav}_id()
      remove deprecated codec_get_{bmp,wav}_{id,tag}()
      bethsoftvid: use correct type for size passed to av_fast_realloc()
      kill harmless 64-bit pointer cast warning
      kill uninitialised variable warning in mp3_read_probe()
      remove redundant declaration of mpegts_muxer
      update SectionCallback and PESCallback typedefs to match functions
      avoid uninitialised variable warning and clarify code
      undef forbidden names before we #define them in case they are macros in libc
      fix build after breakage in r9544
      trivial warning fixes
      make arguments to ssd_int8_vs_int16() const
      fix warning
      kill warnings
      trivial warning fixes
      set sample rate in more logical place, kill warning
      indent
      use correct type for palette
      add fork to HAVE_LIST
      clean up vhook variables
      whitespace cosmetics
      prettify conditional targets
      avoid deprecation warnings for ImgReSampleContext from avcodec.h
      avoid deprecation warnings for AVFrac from avformat.h
      set locale to C in seek test
      10l: fix av_str[i]start()
      remove OS/2 support
      move ff_h263_round_chroma() to mpegvideo.h as static inline
      split lcl.c into lcldec.c, lclenc.c, lcl.h
      forgot to rm lcl.c
      remove useless LclDecContext.avctx field
      indent
      enable $ARCH_EXT_LIST before parsing command line
      libavtest and seektest no longer need gpl
      create and populate "tools" directory
      avoid division by zero in av_find_stream_info() when no time base is set
      move #include of system headers to top of file
      separate en/decoder specific parts from png.c
      hardly anything in PNGContext is shared; split it
      remove superfluous #include
      remove useless #ifdef CONFIG_ZLIB from zmbv decoder
      replace some "disabled" calls with "! enabled"
      fix endian check when cross-compiling
      fix endian detection on MacOS
      use POSIX "strings -a" rather than "strings -"
      do not use "strings" in endian check
      create and use check_asm() function
      set -mcpu for ARM
      fix -lm check
      use pr -n -t instead of non-standard cat -n
      remove duplicate Vorbis comment tag handling
      fix buffer overread with invalid Vorbis header
      use bytestream_get_* in vorbis_header()
      check values more thoroughly in vorbis_header()
      use bytestream_get_* in vorbis_comment()
      simply buffer checks in vorbis_comment()
      Improve Sofdec file detection
      set PKT_FLAG_KEY for Theora and OGM streams
      use -Bsymbolic linker flag if supported
      do not create copy relocations when linking to shared libs
      use av_log_get/set_level()
      Revert r11202: do not create copy relocations when linking to shared libs
      refer users to ffmpeg-user list in case of errors
      fix dynamic loading of libfaad 2.6
      cosmetic: libfaad.c indentation
      clean up dynamic loading of libfaad
      remove some ifeq ($(CONFIG_FOO),yes)
      whitespace cosmetics
      rename dv50 regression test output file to dv50.dv
      remove duplicate, disabled nut regression test
      get seek regression test filenames from reference file
      split regression tests allowing make to run them in parallel
      split lavf regression tests
      remove useless/incorrect comments in regression.sh
      remove single-frame images from seektest
      do not run ldconfig after installing shared libraries
      clean up definition of DECLARE_ALIGNED_8 and STRIDE_ALIGN
      prettify encoder/decoder/muxer/... list extraction
      prettify lib* version number extraction
      remove duplicate AV_STRINGIFY() definition
      remove duplicate AV_STRINGIFY() definition
      Clean up lib* version definitions
      Clean up lib* version definitions
      #include avutil.h in postprocess.h
      use diff -u -w in codec regression test
      fix memory leaks in vp3 decoder
      Install headers in $prefix/include/$libname
      #include "libavutil/avutil.h" in swscale.h
      less preprocessor magic in version number macros
      less preprocessor magic in version number macros
      simplify version number extraction
      print enabled things in columns
      fix libavfilter version extraction
      fix versions in .pc files
      fix version in libavfilter.pc
      simplify lzo-test rule
      check for inline asm support
      remove unnecessary backslashes
      remove invalid free() forgotten in r12300
      Fix endian test with grep not handling non-text files
      simplify processing of some --disable-* options
      allow vpath search only for source files
      consolidate CFLAGS, LDFLAGS, EXTRALIBS assignment
      consolidate CFLAGS, LDFLAGS, EXTRALIBS assignment
      fix AVClass for avfilter
      Remove mp4psp regression test
      simplify library version handling
      Simplify do_video_encoding() in regression.sh
      replace -f pgmyuv with -f image2 -vcodec pgmyuv
      simplify library version handling
      check for x86 bswap instruction
      hopefully portable endian test
      new attempt at portable endian test
      rename LIBSRC to SRC_DIR, add quotes where SRC_DIR was used
      set SRC_DIR in top-leve Makefile
      fix DEPEND_CMD for out of tree builds
      improve precision in mdct.c using double for some temporaries
      fix typo ARCH_PPC -> ARCH_POWERPC
      typo: add missing \ in multi-line macro
      mpegvideo_armv5te.o belongs to OBJS, not ASM_OBJS
      clean up FFT SIMD selection
      indentation
      cosmetics: remove useless \-newline escapes
      use HAVE_FAST_64BIT
      add REGd definition
      get register names from x86_cpu.h
      simplify ALIGN_MASK definition
      check for prefix on extern symbols in configure
      fix extern prefix detection on mingw
      cosmetics: sanitise asm() statements in bswap.h
      32/64-bit agnostic x86 bswap
      add missing #include "common.h" to libavutil headers
      Rename configure variable BUILDSUF to build_suffix
      Rename configure variable PREFIX to prefix
      combine setting of various configure command line options
      fix ff*_g -> ff* rules
      replace some $@ with $(SLIBNAME_WITH_MAJOR) in SHFLAGS
      clean up DEP_LIBS setting
      simplify vhook/*.o rule
      simplify LDFLAGS and EXTRALIBS handling
      non-recursive makefiles
      non-recursive makefiles
      collect vhook-related makefile lines in one place
      append TESTS-yes to TESTS; one ifeq less
      replace some ifeq with prettier ifdef
      generate separate dependency files per object file
      fix build with --build-suffix
      move DEP_LIBS to common.mak
      save DEP_LIBS before processing subdirs
      correct soname for vhooks
      remove unused MAKE variable
      ffplay.d needs SDL_CFLAGS
      clear TESTS variable between subdirs
      fix building of test programs (make tests)
      fix make tests in subdirs
      check for ARM VFP support
      Revert "Adjust pkg-config files to include files residing in separate subdirectories."
      add necessary #includes in headers
      fix checkheaders when libdirac not installed
      set temporary filenames before using them
      acelp_filters.h needs stdint.h
      set $logfile before use
      Revert "Add version information to DLLs."
      fix servertest in out of tree builds
      improve installation path selection
      alloc_size attribute is new to gcc 4.3; don't use it with lesser versions
      accept --cpu=cortex*
      check for ARM NEON assembler support
      enable version.h to be created from subdirs
      fix dependency generation
      add libmpeg2 permutation to dct-test
      Revert "add libmpeg2 permutation to dct-test"
      regenerate dependencies when building object files
      configure: allow !foo as argument to enabled/disabled and *_deps lists
      configure: simplify negative dependencies
      Bitstream: use AV_RB16() in LIBMPEG2_BITSTREAM_READER
      mathops: add MAC64, MLS64, and MLS16
      mpegaudio: use MAC/MLS macros in ff_mpa_synth_filter
      ASF: use dprintf(), allow compilation with -DDEBUG
      RTP: use dprintf(), allow compilation with -DDEBUG
      dvdsubdec: allow compilation with -DDEBUG
      H264: fix index range assert() in remove_short_at_index()
      Remove dead code in bitstream.h
      ARM: ARMv6 optimised bswap_16/32
      ARM: fix ARCH_ARM -> ARCH_ARMV4L test in bswap_32()
      ARM: disable use of system byteswap.h
      ARM: ARMv6 optimised FASTDIV
      ARM: ARMv6 optimised MULH
      ARM: optimised MUL64
      ARM: optimised MAC64 and MLS64
      Rearrange AV_[RW][BL]*() macros
      MPEGTS: Improve probe function
      intreadwrite: support DEC compiler __unaligned type qualifier
      fix ARMv6 FASTDIV for divisor <=2
      Maintain pointer to end of AVFormatContext.packet_buffer list
      Add myself as ARM maintainer
      ARM: set STRIDE_ALIGN and DECLARE_ALIGNED_8 to 16 for NEON
      ARM: add prefetch function using ARMv5 PLD instruction
      aac_parser: fix strict aliasing violation
      configure: simplify PPC dcbzl check
      Revert "Replace sed trickery in the gcc dependency generation command"
      Enable FAST_UNALIGNED on PPC
      Enable FAST_UNALIGNED on ARM variants that support it
      fix configure with libx264 disabled
      ARM: Kill warning in bswap_32()
      configure: fix handling of option arguments containing = sign
      Do not add $cross_prefix to compiler name given with --cc option
      Allow target name on every line in dependency postprocessing
      ARM: allow --cpu=armvN in configure
      split bswap.h into per-arch files
      configure: fix disabled_any to return false for empty list
      configure: add foo_select facility
      configure: cosmetics
      Auto-enable CONFIG_FFT for codecs that need it
      configure: allow recursive foo_select options
      configure: add enable_weak and disable_weak
      configure: add foo_suggest facility
      DCA decoder needs FFT
      Build MMX/SSE FFT only if CONFIG_FFT is set
      AC3 encoder doesn't need FFT
      Auto-enable CONFIG_MDCT when needed
      Add CONFIG_GOLOMB and auto-select when needed
      Remove unused SRCS make variable
      Use OBJS variable for all object files
      ARM: faster ARMv6 FASTDIV()
      Merge OBJS-$(FOO) lists with same FOO
      cosmetic: align backslashes in makefiles
      ARM: add some assembler convenience macros
      ARM: use new macros for assembler function labels
      ARM: move prefetch_arm to dsputil_arm_s.S
      ARM: move VFP DSP functions to dsputils_vfp.S
      ARM: move add_pixels_clamped_ARM() to dsputil_arm_s.S
      Remove forgotten ASM_OBJS in libswscale
      ARM: mathops.h whitespace cosmetics
      ARM: change MULL() macro to inline function
      ARM: change MULH() macro to inline function
      ARM: prettify MAC16() macro
      ARM: change MUL16() macro to inline function
      OGG: untypedef demuxer structs
      OGG: correct PTS with old theora streams
      ARM: fix j_rev_dct_ARM
      SH4: rename functions with conflicting names
      SH4: replace missed rnd_avg2 -> rnd_avg32
      SH4: remove unused functions
      SH4: fix compilation of memzero_align8()
      SH4: do not mix code and declarations
      SH4: fix IDCT
      SH4: remove some commented out blocks of code
      SH4: add macros for setting and restoring FPSCR.PR bit
      SH4: fix memzero_align8()
      SH4: fix IDCT
      configure: simplify set_default function
      configure: set HOSTCC to native compiler
      configure: set TARGET_EXEC and TARGET_PATH
      Build regression test data generators using HOSTCC
      Run regression tests on target system.
      ARM: move dct_unquantize_h263_*_armv5te asm to separate file
      Add shift argument to MULL() macro
      flashsv: use skip_bits_long() where required
      svq1enc: move scratch buffer from stack to context to ensure alignment
      snow: move scratch buffer from stack to malloced buffer in context
      ARM: convert VFP code to UAL syntax
      ARM: remove some useless comments
      AAC: move scratch array to context to ensure 16-byte alignment
      configure: clean up .align detection
      mpegts: replace Program_t typedef with "struct Program"
      Remove other shared lib names from linker command line for shared libs
      configure: fix $nm
      oggflac: fix bitstream reader usage
      ARM: NEON optimised put_pixels functions
      ARM: NEON optimised simple_idct
      ARM: NEON optimised {put,avg}_h264_chroma_mc[48]
      ARM: NEON optimised H.264 loop filter
      ARM: NEON optimised H.264 8x8 and 16x16 qpel MC
      ARM: NEON optimised h264_idct_add
      ARM: NEON optimised h264_idct_dc_add
      dct-test: add PARTTRANS permutation
      dct-test: add ARM IDCTs
      ARM: replace "armv4l" with "arm"
      AAC: fix strict aliasing violation in parser
      AC3: fix strict aliasing violation in parser
      ARM: make FASTDIV() an inline function
      ARM: add new h264 idct functions
      ARM: NEON optimised float_to_int16
      simple_idct: use MUL16 and MAC16 from mathops.h
      Replace UNALIGNED_STORES_ARE_BAD with !HAVE_FAST_UNALIGNED
      configure: simplify TMPDIR setting
      configure: accept $TMP as location for temporary files
      ARM: rename coefficient table in NEON IDCT
      configure: check for ARMv6T2 support
      Make shared/static full CONFIG options
      ARM: work around linker bug with movw/movt relocations in shared libs
      SH4: fix typo in asm constraint
      configure: fix PIC activation for x86_64 shared libs
      AVI: add bswap.h include needed on big endian
      ARM: use rX register names in simple_idct_armv6.S
      ARM: clean up pc-relative references in simple_idct_armv6.S
      ARM: simple_idct_armv6.S whitespace cosmetics
      ARM: use push/pop pseudo-instructions in simple_idct_armv6.S
      configure: add Alpha CPU types
      Fix mpeg audio regression test failure
      Alpha: fix dct_unquantize_h263_inter/intra_axp()
      Alpha: factor common parts from dct_unquantize_h263_intra/inter_axp
      Alpha: add some const, kill some warnings
      Alpha: proper IDCT selection
      Alpha: indentation
      Alpha: add ff_ prefix to idct functions
      dct-test: add Alpha idct
      configure: add fastdiv to CONFIG_LIST
      Alpha: fix pix_abs16
      Alpha: add -mieee to CFLAGS
      ARM: simplify ff_put/avg_h264_chroma_mc4/8_neon definitions, no code change
      configure: add help for fastdiv
      ARM: NEON optimised H.264 biweighted prediction
      ARM: NEON optimised H.264 weighted prediction
      configure: add datadir variable and write to config.{h,mak}
      Fix version.h dependency on version.sh
      configure: exit with error if OS is unknown
      ARM: allow register operands for shifts in MULL()
      ARM: optimised mid_pred()
      ARM: change alignment of loops in put_pixels*_arm to 32
      ARM: replace explicit literal loads with ldr Rd, =lit
      ARM: replace jump tables with conditional branches
      ARM: reorder some instructions in put_pixels*_arm for speed gains
      ARM: remove some unused macro arguments
      ARM: NEON optimised vector_fmul
      ARM: NEON optimised vector_fmul_window
      shorten: trivial simplification of signature check
      configure: move basic sanity tests before other tests
      Remove CONFIG_H264_ENCODER references
      configure: set EXESUF before it is used
      configure: add command flag to disable yasm use
      avidec: fix error introduced in r17382
      Alpha: fix inline asm with DEC/Compaq/HP compiler
      configure: sort cases in arch section
      configure: add (empty) avr32 arch case
      Correct, portable definition of INT_BIT
      Declare version.h dependencies explicitly
      Remove special treatment of version.h in DEPEND_CMD
      configure: move detection of compiler type to single location
      Add --sysroot and --sysinclude configure options
      Pass sysroot argument to gcc
      configure: indent
      configure: detect ccc compiler (Alpha) and set required flags
      Avoid warnings from AV_GCC_VERSION_AT_LEAST with some non-gcc compilers
      Fix get_ur_golomb_jpegls() with A32_BITSTREAM_READER
      configure: remove redundant alsa/asoundlib.h check
      ARM: fix corner-case overflow in H.264 weighted prediction
      Revert "Improve frame rate guessing for streams with two fields per frame."
      configure: {enable,disable}_safe functions that sanitise arguments
      configure: check_func_headers should mark headers as found
      ARM: fix missing MUL16() return type
      SVQ3: Fix decoding with A32_BITSTREAM_READER
      configure: fix return value from check_header
      RA144: work around gcc bug/oddity on ARM

NVIDIA Corporation (6):
      Add VDPAU hardware accelerated decoding for H264 which can be used by
      Add VDPAU hardware accelerated decoding for MPEG1 and MPEG2 which can
      Add VDPAU hardware accelerated decoding for MPEG1 and MPEG2 which will
      Add VDPAU hardware accelerated decoding for WMV3 and VC1 which can
      The field rangered in VDPAU's VdpPictureInfoVC1 now also contains
      Do not set h264.is_reference for pictures that are delayed, but not

Nathan Kurz (5):
      * seek in raw DV patch by Nathan Kurz (nate at verse dot com)
      ffmpeg not selecting first available stream, patch by (Nathan Kurz <nate at verse dot com>)
      default stream timebase docs patch by (Nathan Kurz <nate at verse dot com>)
      set keyframe flag for au/wav patch by (Nathan Kurz <nate at verse dot com>)
      bug in libavformat av_update_cur_dts(), patch by (Nathan Kurz <nate at verse dot com>)

Nazo (1):
      FLV seeking patch by Nazo. lovesyao at aol com.

Neil Brown (7):
      mention that start_time maybe AV_NOPTS_VALUE (from patch by neilb suse de)
      more verbose doxy for duration (from patch by neilb suse de)
      There is no guarantee that every stream will
      When looking for the last packet in each
      Do not butcher start_time in av_estimate_timings_from_bit_rate().
      If a stream has no start time, but the first packet has a 'pts', use that
      Make sure that if a stream has a duration but the container doesn't

Nicholas Tung (3):
      Remove superfluous setting of has_b_frames in codecs without B-frames.
      Bethsoft VID demuxer and video decoder
      Get rid of unnecessary pointer casts.

Nick Kurshev (73):
      Sync with mplayer's stuff
      Remove unnecessary record
      Adding fastmemcpy stuff to speedup mplayer project
      Fixed one lack
      Suppressing external gas stuff to improve portability to Win32
      Portability and testing issues
      Aligned malloc. Another 10% of speedup.
      memalign autodetection
      Independed compilation of SUBDIRS
      Shared objects support
      Shared objects support
      headers installation
      missing header
      vo_vesa: rgb2rgb support
      vo_vesa: more rgb2rgb support
      vo_vesa: more rgb2rgb support
      Prev ver could work only on x86
      forgot, sorry
      mmx optimized 24to32
      mmx, mmx2, 3dnow optimized 24to32
      More accurate prefetching
      sfence
      preprocessor cleanup
      More elegant solution
      32to24: MMX, MMX2, 3DNOW optimization
      Finish mmx2, 3dnow optimiz. 15to16 should be tested. Better fix of can't compile
      No longer required
      cosmetic
      using const modifier
      More standards compilance
      Well - old algorithms and new stuff rgb24to16(15)
      Fixed rgb32(24)to16 stuff, rgb32(24)to15 is still broken
      Fixed rgb32(24)to15 stuff
      Forgot
      Move yuv2rgb to postprocess
      rgb24to32 now is faster
      optimized rgb24to16 stuff
      MMX2, 3DNOW, MMX optimized rgb32(24)to16(15) stuff
      Minor speedup of rgb32to24. (performance is not successful)
      rgb32tobgr32 now is documented ;)
      fixed gcc-3.0.x compilation (by Michael Niedermayer)
      Let loops will be aligned
      Cygwin's mangling by Felix Buenemann <atmosfear at users.sourceforge.net>
      Alpha optimizations by Falk Hueffner <falk.hueffner at student.uni-tuebingen.de>
      Alpha optimizations by Falk Hueffner <falk.hueffner at student.uni-tuebingen.de>
      Small simple idct improvement for Alpha by Falk Hueffner <falk.hueffner at student.uni-tuebingen.de>
      version info
      version info for so too
      I was too lazy to be a hero ;)
      build info for ABI
      build num ++
      untested generic bswap64
      new field for communicating with external postprocessing
      use qstride instead of fixed constants
      new codec: Sorenson v1
      * prefixication (due Kabi's request)
      Applied patch of Felix Buenemann
      Change licence to LGPL since there are no objections from side of original author
      Add Arpi to list of porters.
      10L
      1001L
      fixed debug compilation
      reverse first hunk
      10l
      Function renaming. Solves some namespace conflicts
      reverse
      sync with mplayerxp. Fixes some 405.avi related flaws
      binary compatibility fix (workaround for r.147, r.148)
      RealAudio 1.0 (14_4) and 2.0 (28_8) native decoders. Author unknown, code from a mplayerhq (originally from public domain player for Amiga
      some useful warning
      finally works now
      to be more precise
      optimization: merge phases 0 and 1

Nico Sabbi (21):
      according to the standard the object_type for H264 is 33, not 241
      added support for DVHS (192) packet size
      try with auto_guess if first detection fails
      install dsputil.h, that is used by many filters in mplayer
      Etsi identifies DTS audio with stream_private+descriptor 0x7b (in the PMT).
      initial support for AVOption in AVFormatContext
      added option probesize; at the moment only used by mpegts.c
      scan s->probesize packets instead of MAX_SCAN_PACKETS (that is still the default value)
      install opt.h
      demux VC1 in mpegts
      Added definition and utility functions to handle  AVProgram(s) in AVFormatContext
      replaced custom tracking of programs/services with calls to with av_new_program()
      added structures and code to keep track of pids<->programs mapping: it's needed
      cosmetics: moved to a separate function the code to print the characteristics of an AVStream
      API extension: AVProgram now has an array (stream_index) containing the indexes of the streams in AVFormatContext
      associate to each AVProgram the indexes of the AVStreams contained in it
      print newline after program id even when the name is missing
      added option programid to select the program to use and relative code to handle it
      reindented after previous commit
      removed unused variables accidentally introduced in stream_index_from_inputs()
      treat stream_type==0x82 as AUDIO_DTS when the program has a REGISTRATION_DESCRIPTOR with the value HDMV; approved by Mans

Nicolas Boos (3):
      Install only manpage if binary is present patch by (Nicolas Boos <nicolas.boos at wanadoo dot fr>)
      autobuild doc if possible patch by (Nicolas Boos <nicolas.boos at wanadoo dot fr>)
      remove redundant entry from manpage patch by (Nicolas Boos <nicolas.boos at wanadoo dot fr>)

Nicolas George (9):
      add the possibility to choose aac profile
      make help message go to standard output instead of standard error
      More detailed documentation
      Use the codec name from the command line options instead of the codec id.
      Resolve hosts and bind sockets even when the local_port is not set (0)
      Use compression level to set mp3lame quality option.
      Add a flag to avoid calling vorbis_analysis_wrote twice with an empty buffer.
      Add ALSA support in libavdevice.
      libvorbis: Use 0-10 range for audio quality.

Nicolas Martin (1):
      NC camera demuxer

Nicolas Plourde (3):
      fix palette8tobgr32/palette8torgb32 on big endian
      Should fix altivec detection for g3 system.
      add support for intel mac. mp3lib is not fixed yet.

Nigel Pearson (2):
      Use add_cflags/add_ldflags/add_extralibs instead of manually manipulating
      Fix MacIntel build: "/nop" is illegal on Apple's older version of GAS

Niki W. Waibel (1):
      compiling using nonstd include/library locations / add LDFLAGS to a few places patch by (Niki W. Waibel - nikiwaibel)

Nikolai Zhubr (1):
      fixing overflow in 16->8 bit conversion, patch by (Nikolai Zhubr <s001 at hotbox dot ru>)

Nilesh Bansal (1):
      01-makefile_fix_updated.patch

Oded Shimon (153):
      10l: Only set *compose97i *add_yblock to dsputils context if we are building with Snow enabled
      Disable lavf NUT until it is synced to spec
      fix to vorbis decoder for book==0
      vorbis.h -> vorbis_data.c
      make ready_floor1_list() shared for vorbis_enc.c
      make some more functions global so they can be shared with vorbis_enc.c
      100l, broke ffmpeg compilation, ABS() and uint* types not defined in vorbis_data.c
      prevent warning in vorbis_data.c
      remove render_line from vorbis.h, add ff_vorbis_floor1_render_list instead
      skeleton for vorbis_enc.c
      Original Commit: r3 | ods15 | 2006-09-16 11:49:07 +0300 (Sat, 16 Sep 2006) | 2 lines
      Original Commit: r4 | ods15 | 2006-09-16 15:36:03 +0300 (Sat, 16 Sep 2006) | 3 lines
      Original Commit: r5 | ods15 | 2006-09-16 17:29:04 +0300 (Sat, 16 Sep 2006) | 4 lines
      Original Commit: r6 | ods15 | 2006-09-16 20:36:31 +0300 (Sat, 16 Sep 2006) | 2 lines
      Original Commit: r7 | ods15 | 2006-09-17 08:50:47 +0300 (Sun, 17 Sep 2006) | 2 lines
      Original Commit: r8 | ods15 | 2006-09-17 09:43:38 +0300 (Sun, 17 Sep 2006) | 2 lines
      Original Commit: r9 | ods15 | 2006-09-17 18:56:41 +0300 (Sun, 17 Sep 2006) | 2 lines
      Original Commit: r10 | ods15 | 2006-09-17 19:37:17 +0300 (Sun, 17 Sep 2006) | 2 lines
      Original Commit: r11 | ods15 | 2006-09-17 19:41:28 +0300 (Sun, 17 Sep 2006) | 4 lines
      Original Commit: r12 | ods15 | 2006-09-22 09:28:30 +0300 (Fri, 22 Sep 2006) | 5 lines
      Original Commit: r13 | ods15 | 2006-09-22 09:29:38 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r14 | ods15 | 2006-09-22 12:26:17 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r15 | ods15 | 2006-09-22 12:26:42 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r16 | ods15 | 2006-09-22 12:27:17 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r17 | ods15 | 2006-09-22 12:28:28 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r18 | ods15 | 2006-09-22 12:29:26 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r19 | ods15 | 2006-09-22 12:54:18 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r20 | ods15 | 2006-09-22 13:12:48 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r21 | ods15 | 2006-09-22 13:40:31 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r22 | ods15 | 2006-09-22 13:49:56 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r23 | ods15 | 2006-09-22 13:55:48 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r24 | ods15 | 2006-09-22 13:57:09 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r25 | ods15 | 2006-09-22 14:06:04 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r26 | ods15 | 2006-09-22 14:18:22 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r27 | ods15 | 2006-09-22 14:19:51 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r30 | ods15 | 2006-09-22 18:24:04 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r31 | ods15 | 2006-09-22 18:33:17 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r32 | ods15 | 2006-09-22 18:37:50 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r34 | ods15 | 2006-09-22 18:46:57 +0300 (Fri, 22 Sep 2006) | 2 lines
      Original Commit: r35 | ods15 | 2006-09-22 18:53:22 +0300 (Fri, 22 Sep 2006) | 4 lines
      Original Commit: r36 | ods15 | 2006-09-23 08:43:09 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r37 | ods15 | 2006-09-23 08:43:43 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r38 | ods15 | 2006-09-23 09:02:34 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r39 | ods15 | 2006-09-23 09:36:11 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r40 | ods15 | 2006-09-23 09:36:30 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r41 | ods15 | 2006-09-23 10:52:34 +0300 (Sat, 23 Sep 2006) | 4 lines
      Original Commit: r42 | ods15 | 2006-09-23 10:53:16 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r43 | ods15 | 2006-09-23 10:53:46 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r44 | ods15 | 2006-09-23 10:54:43 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r46 | ods15 | 2006-09-23 11:36:32 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r47 | ods15 | 2006-09-23 12:34:48 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r48 | ods15 | 2006-09-23 12:45:25 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r49 | ods15 | 2006-09-23 12:52:21 +0300 (Sat, 23 Sep 2006) | 3 lines
      Original Commit: r50 | ods15 | 2006-09-23 16:24:51 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r51 | ods15 | 2006-09-23 16:32:30 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r52 | ods15 | 2006-09-23 17:38:43 +0300 (Sat, 23 Sep 2006) | 2 lines
      Original Commit: r54 | ods15 | 2006-09-24 13:54:12 +0300 (Sun, 24 Sep 2006) | 2 lines
      Original Commit: r55 | ods15 | 2006-09-24 13:54:43 +0300 (Sun, 24 Sep 2006) | 2 lines
      Original Commit: r56 | ods15 | 2006-09-24 13:55:01 +0300 (Sun, 24 Sep 2006) | 2 lines
      Original Commit: r58 | ods15 | 2006-09-24 17:35:28 +0300 (Sun, 24 Sep 2006) | 2 lines
      Original Commit: r59 | ods15 | 2006-09-24 17:40:40 +0300 (Sun, 24 Sep 2006) | 2 lines
      Original Commit: r60 | ods15 | 2006-09-25 12:46:30 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r61 | ods15 | 2006-09-25 13:10:56 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r62 | ods15 | 2006-09-25 13:12:07 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r63 | ods15 | 2006-09-25 13:51:44 +0300 (Mon, 25 Sep 2006) | 3 lines
      Original Commit: r64 | ods15 | 2006-09-25 18:39:30 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r65 | ods15 | 2006-09-25 18:40:19 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r66 | ods15 | 2006-09-25 18:42:13 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r67 | ods15 | 2006-09-25 20:33:06 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r68 | ods15 | 2006-09-25 21:15:09 +0300 (Mon, 25 Sep 2006) | 2 lines
      Original Commit: r69 | ods15 | 2006-09-27 20:27:50 +0300 (Wed, 27 Sep 2006) | 2 lines
      Original Commit: r70 | ods15 | 2006-09-28 19:07:36 +0300 (Thu, 28 Sep 2006) | 2 lines
      Original Commit: r71 | ods15 | 2006-09-28 21:00:24 +0300 (Thu, 28 Sep 2006) | 2 lines
      Original Commit: r72 | ods15 | 2006-09-28 21:00:47 +0300 (Thu, 28 Sep 2006) | 2 lines
      Original Commit: r73 | ods15 | 2006-09-28 21:01:59 +0300 (Thu, 28 Sep 2006) | 2 lines
      Original Commit: r74 | ods15 | 2006-09-29 13:03:31 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r75 | ods15 | 2006-09-29 13:03:51 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r76 | ods15 | 2006-09-29 13:16:37 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r77 | ods15 | 2006-09-29 15:12:56 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r78 | ods15 | 2006-09-29 15:14:28 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r79 | ods15 | 2006-09-29 15:15:26 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r80 | ods15 | 2006-09-29 15:44:22 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r81 | ods15 | 2006-09-29 21:02:09 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r82 | ods15 | 2006-09-29 21:03:43 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r83 | ods15 | 2006-09-29 21:06:13 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r84 | ods15 | 2006-09-29 21:07:07 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r85 | ods15 | 2006-09-29 21:07:58 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r86 | ods15 | 2006-09-29 21:08:42 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r87 | ods15 | 2006-09-29 21:10:05 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r88 | ods15 | 2006-09-29 21:10:36 +0300 (Fri, 29 Sep 2006) | 2 lines
      Original Commit: r90 | ods15 | 2006-09-29 22:14:40 +0300 (Fri, 29 Sep 2006) | 3 lines
      Original Commit: r91 | ods15 | 2006-09-30 10:05:16 +0300 (Sat, 30 Sep 2006) | 4 lines
      Original Commit: r92 | ods15 | 2006-09-30 12:13:47 +0300 (Sat, 30 Sep 2006) | 3 lines
      Original Commit: r93 | ods15 | 2006-09-30 22:02:21 +0300 (Sat, 30 Sep 2006) | 2 lines
      Original Commit: r94 | ods15 | 2006-09-30 22:09:33 +0300 (Sat, 30 Sep 2006) | 2 lines
      Original Commit: r95 | ods15 | 2006-09-30 22:11:08 +0300 (Sat, 30 Sep 2006) | 3 lines
      Original Commit: r96 | ods15 | 2006-09-30 23:25:18 +0300 (Sat, 30 Sep 2006) | 2 lines
      Original Commit: r97 | ods15 | 2006-09-30 23:32:19 +0300 (Sat, 30 Sep 2006) | 2 lines
      Original Commit: r98 | ods15 | 2006-10-01 10:15:33 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r99 | ods15 | 2006-10-01 11:12:44 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r100 | ods15 | 2006-10-01 14:01:27 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r101 | ods15 | 2006-10-01 14:15:00 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r102 | ods15 | 2006-10-01 17:58:38 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r103 | ods15 | 2006-10-01 18:00:43 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r104 | ods15 | 2006-10-01 18:04:08 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r105 | ods15 | 2006-10-01 18:09:55 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r106 | ods15 | 2006-10-01 18:11:55 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r107 | ods15 | 2006-10-01 18:13:12 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r108 | ods15 | 2006-10-01 18:46:06 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r109 | ods15 | 2006-10-01 19:14:52 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r110 | ods15 | 2006-10-01 19:17:32 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r111 | ods15 | 2006-10-01 19:40:50 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r112 | ods15 | 2006-10-01 20:48:39 +0200 (Sun, 01 Oct 2006) | 3 lines
      Original Commit: r113 | ods15 | 2006-10-01 21:35:47 +0200 (Sun, 01 Oct 2006) | 2 lines
      Original Commit: r114 | ods15 | 2006-10-01 21:38:37 +0200 (Sun, 01 Oct 2006) | 2 lines
      Add vorbis encoder to ffmpeg
      add vorbis_enc_data.h, remove codebook data from vorbis_enc.c
      add vorbis encoder to changelog/MAINTAINERS
      10l, wrong order in freeing
      more forgotten free's.
      new optimized eval method, by seperating parsing and runtime
      update ratecontrol to new ff_eval API
      possible bug of 'gte' being read as 'gt', same with 'lte'
      shut gcc warning, also makes sense for NAN to be returned if the loop was never executed
      add missing 'e->value * '
      ignore whitespace in ff_eval
      I hope noone minds, adding myself to eval.c copyright...
      fix for disabling vorbis decoder but enabling vorbis encoder
      ministry of english grammar :) then->than
      Add libnut support until ffmpeg gets an independant and complete de/muxer for NUT
      register nut muxer if libnut is used
      change include "nut.h" to <nut.h> to make gcc pick the correct nut.h for libnut de/muxer..
      update to libnut API, rename nut.h->libnut.h
      update to libnut, add cache_syncpoints
      update to libnut API, don't free the streams
      I suppose I can add myself, at least for my vorbis encoder...
      Update to libnut API, non-negative errors
      update to libnut API, changes to frame_table_input
      update to libnut, use framecode table generator
      change mpeg4 fourcc from "mp4s" to "mp4v"
      sync to libnut, nom->num
      Add doxygen file header to libnut.c
      remove usage of deprecated functions in libnut.c (codec_get_id/tag)
      fix mjpeg decoding with broken huffman table headers
      ff_nut_reset_ts() expected to get 'ts*time_base_count', but muxer only
      missing " && j<syncpoint_count" protection in the index parsing, as the
      Remove old workaround in nutdec.c for libnut bug
      nut->max_distance was not set at all anywhere in the NUT muxer
      Fix intented algo in syncpoint writing before keyframes in NUT muxer
      back_ptr calculation used wrong timebase when searching for back syncpoint
      Update seektest regression as well to previous lavf/nut commits
      Update lavf/libnut.c to _tt type names in libnut API
      Rename all vorbis encoder related typedefs to not use _t for POSIX compatibility

Oliver Pfister (2):
      fix mem leak
      fix mem leak

Palle Lyckegaard (1):
      Add -D__EXTENSIONS__ to CFLAGS on Solaris to make ip_mreq visible.

Panagiotis Issaris (220):
      Fixes GCC 3.3.2 warnings patch by (Panagiotis Issaris <takis at lumumba dot luc dot ac dot be>)
      GCC 3.3.2 warnings patch by (Panagiotis Issaris <takis at lumumba dot luc dot ac dot be>)
      When using the -target option, although PAL or NTSC is specified,
      target DV patch by (Panagiotis Issaris <takis at lumumba..luc..ac..be>)
      vbv size for ffserver.conf patch by (Panagiotis Issaris <takis [] lumumba*luc*ac*be>)
      target VCD GOP size patch by (Panagiotis Issaris <takis >< lumumba dot luc.ac.be>)
      imgresample test cleanup patch by (Panagiotis Issaris <takis )( lumumba d0t luc d0t ac.be>)
      clip_uint8 should return an uint8_t instead of an int (patch by Panagiotis Issaris < takis.issaris _at_ uhasselt.be >)
      Correct wrong memset invocation.
      Fix linking when the FFM muxer/demuxer is disabled.
      Better tag support:
      adds support for setting that album tag.
      Remove unused defines, patch by takis P issaris A uhasselt P be
      Allow to store the the track number though the IPRT (part) tag in AVI.
      Allow to get the the track number in the IPRT (part) tag in AVI.
      makes avformat_get_context_defaults() static
      enable setting the year and track number using AVOption
      enables AVOption for muxing.
      Add -lav* to the front of LDFLAGS, fixes linking on Linux.
      undefines a few temporary defines which were not undefined after their usage.
      make AVOptions default value field work.
      adds doxygen docs to av_opt_set_defaults. Patch by Takis
      Use and set sane AVOption defaults
      Remove nonexistent option -hq from the documentation.
      adds descriptions to the libavcodec AVOptions array, taken from the
      corrects the type of the AVOption parameter "bugs", patch by Takis
      Allow parameter values (AVOptions) to use the 'k', 'M', 'G'
      Remove a redundant memset from libavformat.
      Make OFFSET() reuse offsetof
      Synchronize the defaults specified in the AVOptions[] array to the defaults
      Rename the AVOption's names to the shorter OptionDef names.
      Remove the "b" OptionDef option from the ffmpeg.c file. From now on the unit of
      Remove the "maxrate" and "minrate" OptionDef options from ffmpeg.c.  This
      Remove the GOP "g" OptionDef option from ffmpeg.c .
      Remove the "preme" OptionDef option from ffmpeg.c
      Remove the "ibias" OptionDef option from ffmpeg.c. The equally named "ibias"
      Remove the "pbias" OptionDef option from ffmpeg.c. The equally named "pbias"
      Remove the "qblur" OptionDef option from ffmpeg.c. The equally named "qblur"
      Remove the "qcomp" OptionDef option from ffmpeg.c. The equally named "qcomp"
      Fix usage of b_quant_factor. It was not possible to actually use the b_qfactor
      Remove the "b_qfactor" OptionDef option from ffmpeg.c. The equally named
      Remove the "b_qoffset" OptionDef option from ffmpeg.c. The equally named
      Remove the "i_qoffset" OptionDef option from ffmpeg.c. The equally named
      Split the DV demuxer and muxer into separate files (as suggested by Diego
      Remove the "me_range" OptionDef option from ffmpeg.c. The equally named
      Correct the i_qfactor minimum, which was current set higher then the default
      Remove the "i_qfactor" OptionDef option from ffmpeg.c. The equally named
      Remove the "bt" OptionDef option from ffmpeg.c. The equally named
      Remove the "rc_init_cplx" OptionDef option from ffmpeg.c. The equally named
      Add support for SI (k, M, ...) and IEC/IEEE (Ki, Mi, ...) units.
      Remove the "bufsize" OptionDef option from ffmpeg.c. The equally named "bufsize"
      Migrate the "muxrate" OptionDef option to become an AVOption.
      Migrate the "packetsize" OptionDef option to become an AVOption.
      Correct the constraints of some AVOptions. These had a default value of zero,
      Inform the user that a certain AVOption is out of range.
      Remove the "bf" OptionDef parameter so the equally named AVOption will take
      Remove the "qmin" OptionDef.
      Remove the "qmax" OptionDef.
      Remove the "qsquish" OptionDef. The AVOption min/max value is changed to match
      Remove the "bug" OptionDef used for setting the workaround_bugs variable.
      Remove the "sc_threshold" OptionDef variable.
      The "video_rc_buffer_aggressivity" variable is not being used anymore in
      Remove the "mb_threshold" OptionDef. The equally named AVOption takes over its
      Remove the "skip_threshold" OptionDef. The equally named AVOption takes over its
      Remove the "skip_factor" OptionDef. The equally named AVOption takes over its
      Remove the "skip_exp" OptionDef. The equally named AVOption takes over its
      Remove the "mepc" OptionDef. The equally named AVOption takes over its function.
      Remove the "error" OptionDef. The equally named AVOption takes over its
      Define NAN -if not already defined- as 0.0/0.0
      Missing extern declaration for av_strtod.
      Handle possible failure of ff_eval.
      Introduce ff_eval2 which is equivalent to ff_eval but does not log anything.
      Fix compilation by postponing deprecation of ff_eval() until the next version
      Make ratecontrol use ff_eval2().
      Make AVOption parsign code use ff_eval2()
      Break compatibility only when first part of version number changes, in this
      Remove the "lmax" OptionDef. The equally named AVOption takes over its function.
      Remove the "lmin" OptionDef. The equally named AVOption takes over its function.
      Remove the "mblmin" OptionDef. The equally named AVOption takes over its
      Remove the "mblmax" OptionDef. The equally named AVOption takes over its
      Pull out the ff_eval* from the mpegvideo header, as it doesn't belong there and
      Remove the override of lmin and lmax, so the actual option value provided will
      Reformat the output of the list of available AVOptions, by indenting the
      Move the ratecontrol related code from mpegvideo.h to a separate header file.
      Add forgotten eval.h...
      Remove the H264Context parameter from decode_init_vlc() as it is not being used.
      Change NAL unit type defines to an enum and move to a header file for usage
      Add comment denoting group of H.264 functions
      Removing unused code
      Export symbols of several H.264 intra prediction functions
      Add more descriptions to the AVOption array taken from the avcodec.h AVCodec structure member comments.
      Add "mixed_refs" description suggested by Loren Merritt and update accompanying
      Removing unused code
      Making rem6 and div6 globally visible and thus adding prefixes.
      Add codec ID for a native H.264 encoder.
      Added entry for H.264 DCT to the DSP context.
      Adding function which enables writing H.264 NAL units.
      Adding H.264 DCT.
      Allow provision of displaynumber, screennumber, x-offset and y-offset parameters
      fix segfault with http://sam.zoy.org/zzuf/lol-ffplay.ogm and
      Change the Snow encoder to always use the available MotionEstContext pointer.
      Brings down the number of snow.c warnings from 33 to 27 by putting parentheses
      Bring down the number of snow.c warnings from 27 to 17 by using the const
      Silence GCC when incorrectly complaining that the "line" variable could be used
      Typo
      Add the const specifier as needed to reduce the number of warnings.
      Remove the getbe16 functions and use the AV_RB16 macro instead. Patch by Ian
      Make x11grab output its warning that it couldn't find the mouse pointer only
      Use a more correct value for Pi.
      Remove a non-C dependency (being M_PI) from the libavcodec example.
      The --enable-a52 option was renamed to --enable-liba52, fix the FAQ regarding
      Provide Doxygen documentation.
      Adds Doxygen docs for the av_log function.
      Add some more Doxygen documentation to libavcodec/utils.c.
      Correct typo.
      Fix typo.
      Stray comment, probably copy & pasted from audio_encode_example().
      The memsetting of the end of buffer is not needed for audio.
      Fix spelling.
      Add more Doxygen documentation.
      Move the Doxygen function documentation unaltered from libavcodec/utils.c to
      Doxygenize the comments for the avpicture_free() function.
      Doxygenize the comments for the av_get_pict_type_char() function.
      Doxygenize the comments for the av_get_bits_per_sample() function.
      Doxygenize the comments for the av_realloc_static() function.
      Doxygenize the comments for the av_mallocz_static() function.
      Doxygenize the comments for the av_free_static() function.
      Doxygenize the comments for the av_fast_realloc() function.
      Doxygenize the existing amr.c documentation.
      Move the av_realloc_static() declaration out of avcodec.h into bitstream.c.
      av_realloc_static() is an internal function and therefore should use the ff_ prefix instead of av_.
      Move the unaltered documentation of the functions which are part of the public
      Add documentation to some avcodec.h's pix_fmt related functions and defines.
      Move the libavformat public API comments to avformat.h. The comments are
      Correct wrong avformat documentation.
      Add the prefix "av_" to img_crop(), img_copy() and img_pad(), and rename "img"
      Move the unaltered Doxygen docs from the log source code to the accompanying
      Move av_mod_i()'s comment to the header file. The moved comment is unmodified.
      Move av_div_i()'s comments to the header file. The comments are unaltered.
      Move av_shr_i()'s unaltered comments to the header file.
      Move av_cmp_i()'s unaltered comment to the header file.
      Move av_int2i()'s unaltered comments to the header file.
      Move av_i2int()'s unaltered comments to the header file.
      Move av_log2_i()'s unaltered comments to the header file.
      Move rtp_get_payload_type()'s unaltered comments to the header file.
      Move rtp_check_and_send_back_rr()'s unaltered comments to the header file.
      Doxygenize the existing comments in avformat.h. The contents of the comments
      Doxygenize the existing comments without altering the actual contents of the
      Doxygenize the unaltered comments for the RTP_MAX_SDES definition.
      Doxygenize the unaltered comments for the rtp_payload_data_s struct.
      Doxygenize the unaltered comments for the AUHeaders structure.
      Move unaltered av_malloc() comments to the header file.
      Move unaltered av_realloc() comments to the header file.
      Move unaltered av_free() comments to the header file.
      Move unaltered av_freep() comments to the header file.
      Convert macro+enum into a simple enum: Currently the enum was filled using a
      Doxygenize comments in rtsp.h
      Move unaltered rtsp.c comments to the header file.
      Make rtsp_set_callback() comments available to Doxygen
      Move aviobuf.c comments to avio.h. By moving the aviobuf.c comments to avio.h,
      Move avio.h's url_get_max_packet_size() comment from source to header file.
      Convert all the comment sections into Doxygen compatible comments, same for
      Fix a typo in a doxygen tag.
      Cleanup doxygen comment.
      Clarify doxygen description.
      Update the documentation to be consistent with the new unit for the "ab"
      Fix typo in Doxygen comments.
      Introduce two new logging functions av_hex_dump_log() and av_pkt_dump_log()
      Move the memory related functions out of common.h into their own header file
      Install mem.h on "make install" as:
      Fix compilation when using the --disable-opts parameter. This to help those
      Typos
      Fix Vorbis encoder Xiph dependency by making sure the references to
      Make vp3dsp*.c compilation optional.
      Split the h261.c file into separate files for the encoder, decoder and common stuff.
      Only compile cabac.c when needed.
      Fix compilation when the Sonic encoder and decoder were disabled and only the Sonic
      Only compile golomb.c when needed.
      Make the compilation of h264idct.c optional.
      Reindent
      Export the four remaining H.264 decoder intra prediction functions for reuse in
      Remove the unnecessary masking when extracting the start bit in the H.264 RTP
      Remove the unnecessary masking when reconstructing the NAL unit header in the
      Remove the unnecessary masking when counting received packet types in the H.264
      Remove an unreferenced variable from qpeg_decode_inter().
      Correct signature of just_return().
      Remove unneeded semicolon.
      Fix the compilation of dct-test, fft-test and motion-test which are using
      Fix the self tests which are contained in some codecs and are using random().
      Replace all occurrences of AVERROR_NOMEM with AVERROR(ENOMEM).
      Replace all occurrences of AVERROR_IO with AVERROR(EIO).
      Replace the only occurrence of AVERROR_NOENT with AVERROR(ENOENT).
      Correct one Doxygen entry, replacing NOMEM with AVERROR(ENOMEM).
      Replace all occurrences of AVERROR_NOTSUPP with AVERROR(ENOSYS).
      Replace ENOSYS by ENOTSUP as in this case the problem is not really a function
      Return AVERROR(ENOMEM) on memory allocation failure of avcodec_open.
      Return AVERROR(EINVAL) when invalid width and/or height are specified to
      Add a section about C++ and extern "C" in the FAQ.
      Add the possibility to show all configurable encoders, decoders, parsers,
      Replace a return of -1 with ENOMEM.
      Rename h264dsp.c to h264dspenc.c
      Take care of some renames (Doxygen and function name) after the previous pure rename patch.
      Handle malloc failure
      Fix typo
      Remove redundant test.
      cosmetics: typos
      Typo
      Attribute for specifying direction should come right after @param.
      Fix Doxygen parameter name
      Fix Doxygen parameter name
      Fix Doxygen parameter
      Add a --disable-devices flag which disables all muxers and demuxers of
      check mov_read_default return value where appropriate, patch by takis, fix issue 285
      Verify long_ref_count!=0. Fix issue 281.
      Show which bitstreamfilters were compiled in, similar to the protocols etc
      Make av_class a pointer to a const AVClass. Addresses one warning in
      Make av_class a pointer to const.
      Update the regression tests reference values after the changes to the Matroska
      Show the "--disable-mmx2" and "--disable-ssse3" options when using "./configure
      Check whether the memory allocation for extradata succeeded. Fixes issue 472.

Pascal Massimino (3):
      Additional checks for strange num_val in FLV metadata
      in TRUNCATED mode, use residual fragments in ParseContext
      fix for >2GB flv files

Patric Stout (1):
      Fix streaming to XBox360.

Patrice Bensoussan (14):
      Option to copy meta data from specified file patch by (Patrice Bensoussan <patrice.bensoussan at free dot fr>)
      Fix bug causing conversion to stop when output_packet() fails patch by (Patrice Bensoussan <patrice.bensoussan at free dot fr>)
      iTunes metadata encoding support patch by (Patrice Bensoussan <patrice.bensoussan at free dot fr>)
      Fix linking failure on Mac OS X <10.4.
      Fix lrintf detection when extra libs are not installed in a standard location.
      Fix asf header decoding issue, patch by < patrice -- bensoussan AH gmail -- com >
      Decode genre on asf/wma files.
      Do not hardcode cc on Darwin, fixing --cc option in the process.
      Add proper #idef to dlsymm call, also fixes an unresolved symbol on OS X 10.2.
      change atoll call to equivalent strtoll to fix compilation on Mac OS X 10.2
      Add support for track numbers when they are set via strings
      Improve track number decoding for some WMA files
      Add support for ID3v2 year tag
      Add support for reading WM/Year tag

Patrik (1):
      Replace snprintf() with av_strlcpy().

Patrik Kullman (14):
      Fix (with av_uninit()) false positive warning about uninitialized variable:
      Add a default (error) for the switch in case of an unsupported PIX_FMT.
      Use av_strlcpy instead of snprintf without formatting.
      Change ffm_write_write_index to return int, and return if error occured.
      Check return value of ffm_write_write_index and ffm_read_write_index.
      lpc: Document the use_lpc parameter to ff_lpc_calc_coefs().
      Assert that ff_lpc_calc_coefs() is called with a valid LPC method.
      lpc: Remove false positive compilation warning about weight being
      Add libavcodec/internal.h include to avoid warning: implicit declaration of
      Add av_uninit to variable declarations to fix the false positive warnings:
      flacdec: Silence false positive warning about uninitialized variables in
      Use av_strdup to allocate string, prevents the warning:
      Check return value of posix_memalign and explicitly set pointer to NULL if it
      Check the return values of a few functions to avoid the corresponding warnings.

Paul Curtis (1):
      AVOutputFormat mpeg2dvd_mux and int is_dvd from the dvd patch by (Paul Curtis <pfc at terrapin dot com>)

Paul Egan (1):
      new avm2 (flash 9) muxer, patch by Paul Egan, paulegan at mail dot com

Paul Kelly (2):
      user specifyable maximum amount of memory to use for the index.
      Add 'reservoir' flag to lavc's flags2 controlling whether or not codes may

Paul Kendall (3):
      Simple refactoring to resolve a FIXME.  Patch by Paul Kendall.
      left_block intialisation optimisation.
      Use #define instead of a constant.  Patch by Paul Kendall.

Peter Ross (96):
      EA ADPCM R1, R2 and R3 decoder
      EA IMA SEAD decoder
      EA IMA EACS decoder
      This header just needs stdint.h, not inttypes.h, which is a superset of
      Disable vhook on Interix, patch by Pete Ross, pross xvid org.
      Compiling ffmpeg.c fails when when the .ffm demuxer is enabled, but the .ffm
      flvenc.o depends on avc.o (for H.264 support functions).
      Add missing entry for ADPCM_EA_MAXIS_XA_DECODER.
      Add missing entry for ADPCM_IMA_QT_ENCODER.
      Add missing mpegvideo.o to list of vc1.o dependencies.
      vp6: ensure that huffman decoding table is sorted with descending symbol order
      Only define AltiVec vector syntax macro when AltiVec is enabled.
      Remove AVPaletteControl from ALG MM demuxer/decoder
      Electronic Arts CMV decoder
      Electronic Arts CMV demuxer
      Electronic Arts CMV latent changelog entry
      Electronic Arts CMV latent dox entry
      increment minor version for commit r14110 (Electronic Arts CMV decoder)
      add CODEC_ID_PCM_F32BE (32-bit floating point PCM big endian decoder)
      Support 32-bit floating point audio samples in Sun AU demuxer
      Include 32-bit PCM description
      add sample_fmts infrastructure: AVCodec->sample_fmts and SAMPLE_FMT_NB enum.
      increment minor version for commit r14412 (add sample_fmts infrastructure)
      give additional use case for version bumping; remove no longer true statement about version strings
      EA Demuxer: make rev2 flag optional
      Modify all codecs to report their supported input and output sample format(s).
      Add sample format support functions: avcodec_get_sample_fmt_name(), avcodec_get_sample_fmt(), avcodec_sample_fmt_string()
      Write sample format description within avcodec_string()
      Add opt handler for pending "-sample_fmt" option
      Revert r14484 hunk that deleted the 'include avcodec.h' statement.
      Modify av_audio_convert() to use AVAudioConvert context struct; add av_audio_convert_alloc() and av_audio_convert_free() support functions.
      Simplify AVCodecContext usage within FFplay audio_decode_frame()
      Add sample format converter to FFplay.
      Add sample format converter to FFmpeg (adds -sample_fmt option)
      Add graceful error handling to avcodec_get_pix_fmt_name() and avcodec_get_pix_fmt().
      Add D-Cinema audio muxer
      Make avcodec_string() and av_get_bits_per_sample() report the sample size for CODEC_ID_PCM_ZORK
      Simplify PCM codec; use sample_fmts field to set the avctx->sample_fmt field.
      Simplify PCM codec; replace switch() statements with av_get_bits_per_sample().
      Support muxing of Zork PCM audio into RIFF formats.
      Add av_memcpy_backptr(): deliberately overlapping memcpy variant.
      Electronic Arts TGV decoder
      Add Electronic Arts TGV chunk tags to EA demuxer.
      Prevent DAUD PCM encoder from fetching values outside of ff_reverse[] array bounds when input sample values are < 0.
      Simplify PCM codec; change 'n' in pcm_decode_frame() to equal "total number of samples".
      Add simpler/cleaner/faster F32BE encoding/decoding.
      PCM regression tests.
      Add MDEC tags to Electronic Arts demuxer.
      Apply PCM ENCODE/DECODE() macros to the S/U,8/24/32,LE/BE PCM codecs.
      Support Electronic Arts files containing MPEG2VIDEO.
      Electronic Arts demuxer: ignore PTxx platform ID values.
      Electronic Arts demuxer: perform identification of ADPCM EA R2 codec using revision2 flag.
      Fix CODEC_ID_PCM_U8 decoder output size calculation to support odd-number of samples.
      Electronic Arts demuxer: support SxEN audio tags found in 'Need for Speed: Pro Street' computer game.
      Use memcpy() for PCM S16/S32 codecs when codec byte-order matches machine byte-order.
      Mark SAMPLE_FMT_S24 as deprecated.
      Support Electronic Arts files containing MP3 audio.
      Fix PCM_S16LE_PLANAR channel-address calculation bug introduced in r14659.
      Add SAMPLE_FMT_DBL.
      Add F64BE, F64LE and F64BE PCM codecs.
      Make AU demuxer handle S24BE, S32BE and F64BE PCM audio.
      Make RIFF-based demuxers handle F32LE and F64LE PCM audio.
      Make MOV demuxer handle F32BE, F32LE, F64BE and F64LE PCM audio.
      Make AIFF demuxer handle F32BE and F64BE PCM audio.
      Add regression tests for F32LE, F64BE and F64LE PCM codecs.
      Fix PCM DVD divide by zero bug introduced in r14659. Patch supplied by lars dot taeuber at gmx dot net.
      Include floating point PCM in changelog and documentation.
      Add raw muxers/demuxers for F64/F32/S32/S24/U32/U24 PCM audio.
      Upgrade 20/24-bit PCM DVD decoder use SAMPLE_FMT_S32. Patch supplied by lars dot taeuber at gmx dot net.
      Change AVCodecContext.bits_per_sample to bits_per_coded_sample.
      Add AVCodecContext.bits_per_raw_sample field.
      Bump libavcodec minor version number for r15017.
      Remove 'currently unused' warning about SAMPLE_FMT_xxx.
      Display error message when user attempts to resample non-S16 audio formats.
      Remove further 1sample=2byte assumptions within FFmpeg to allow
      Add audio channel layout API to libavcodec.
      Decode audio channel layout from RIFF/WAV file format.
      Promote inv_aanscales array to global scope (ff_inv_aanscales)
      Electronic Arts TGQ video decoder
      Electronic Arts TGQ/TQI/MAD IDCT algorithm
      Electronic Arts demuxer: support TGQ video tags
      convert mpeg1_decode_block_intra() to inline global
      don't assume quant_matrix[0]==8 in ff_mpeg1_decode_block_intra(). this is required for the Electronic Arts TQI decoder.
      Electronic Arts TQI decoder
      Electronic Arts demuxer: support TQI video tag
      Use consistent indent style in oggdec.
      Prevent heap corruption when resampling 8-bit audio.
      Add AVERROR_EOF.
      AVI demuxer: return AVERROR_EOF upon detection of end of file.
      FLV demuxer: return AVERROR_EOF upon detection of end of file.
      FFplay: Stop processing input file when AVERROR_EOF is returned from av_read_frame()
      Add myself as maintainer for American Laser Games and Electronic Arts decoders.
      Add PIX_FMT_RGB48BE and PIX_FMT_RGB48LE.
      Support 48-bit RGB PPM image.
      Increase ffmpeg intermediate conversion buffer size to accomodate PPM/PNM header.
      Increase ffmpeg intermediate conversion buffer size to accomodate 48-bit RGB rawvideo

Peter Schlaile (1):
      Make sure strides have positive values before converting.

Petr Doubek (1):
      HTTP Authentication Patch by (Petr Doubek <doubek at vision dot ee dot ethz dot ch>)

Philip Gladstone (151):
      * patch by Philip Gladstone
      Move the EXTRALIBS to the end of the link line where they actually get used!
      * Fix nasty problem with unitialized memory in the file_table
      * Fix a nasty problem with output buffering not have enough (or large enough)
      * These are my notes on streaming
      * Updated so that some streams actually work!
      * Many, many changes to make it work (mostly). Information snagged from
      * Make it work with sound cards (like mine) that can only capture in stereo.
      * Add a duration field to the AVPacket
      * Add default cases to switch statements.
      * Change extern inline to static inline. This improves compiling without optimization.
      * Add the gop_size to the video parameters. Also the audio framesize.
      Intialize the codec type to video.
      * Add a sleep when doing the post to ffserver. Yes, this is the wrong
      * Add defaults to switch statements
      * Try and fill out some more of the RV header fields correctly
      * Change extern inline to static inline so that it will compile without optimization
      * Make put_wav_header return the number of bytes that it put.
      * Start using enumerated types (makes debugging much easier)
      * Every frame is a key_frame
      * Fix a problem with the first sample when down sampling.
      * Don't allocate 0 bytes of memory. It upsets electricFence!
      * Add the 'flags' field to the list of codec fields passed from ffserver to
      * Add prototype for strlcpy
      * Add implementation of strlcpy
      * Improve logging -- now actually gives you the number of bytes sent
      * Mea culpa: it seems that I broke encoding to 8-bit pcm files. This fixes it.
      Pass over the bit_rate_tolerance field to ffmpeg
      * Add code to configure the following:
      * Updated with new stream configuration items
      * Change the default behaviour to start streaming as soon as possible (i.e.
      Added use of StartSendOnKey
      * Added a note that says that WMP cannot stream .mpg files.
      Fix the WAIT_FEED problem. It turns out that when you open up an FFM
      * If a stream gets stuck in WAIT_FEED, then disconnecting the other end
      Fix a segfault when handling errors or .asx or .ram files. Silly bug
      Add an av_abort macro that aborts, but also prints out the location of the abort.
      * Add support for time-limiting a live stream. (Some guy streamed one of
      * Fix a problem caused by a duplicate free of priv_data -- now uses av_freep
      * Added yet more output to the status page to try and track down a problem
      * Add a probe function for FFM files
      Change order of extensions and read_probe checking. This gives the read_probe
      * Add first cut of code to handle Windows Media Player rate switching
      * Make video grabbing work again (if you are grabbing multiple streams
      Make sure that the http_proxy environment variable starts with http://
      * Go faster stripes: don't check to see if the user presses 'q' if stdin
      * New stuff and bug fixes:
      Added Launch, FaviconURL and Redirect examples
      * Bad buffer management bug. Forgot to take care of the case
      Remove some calls to av_free that tured out to be duplicate. This improves
      * Give the max stream time configuration item a chance of working
      * Add the comment block if any of the comment fields are present.
      Add the copyright/comment/author/title keywords
      * Get the buffer size right now we have the buffer inside the FFStream block
      * Add code to restart ffmpeg if it crashes
      Change the way that the ?date= stuff is parsed. It now takes a time (with optional date)
      Add ticker_abs function that returns the number of ticks to get the
      Change ticker_tick to use ticker_abs as we want the absolute number
      Add another asf output format type -- asf_stream. The convention is that
      Add prototype of guess_stream_format
      * Add code for ffserver to get stream output formats: guess_stream_format
      Add strptime into the source pool for those systems that don't have it.
      Add logic to figure out whether strptime is present on the system. If not, then
      Add conditional build of strptime
      * Make asf streaming work again. Now uses the output format asf_stream.
      Add some examples of using ACL
      Fix a potential endless loop (nb (1 * 3) / 2 == 1)
      Add conversions to and from RGBA32 and BGRA32.
      Add notes on the ?date= format.
      * Removed a debug printf
      * Now we do not delete the feed file if it was created with the same
      Add the transfer of the new parameters from ffmpeg to ffserver and vice-versa
      * Add code to pace sending a stream when it is being sent from a file.
      Change calls to abort() to be calls to av_abort()
      Change abort() calls to av_abort() calls.
      Added VideoBitRateRange and VideoBitRAteTolerance settings
      * Made encoding to ASF streams work again. It turns out that the default
      Get rid of a warning from asfcheck. Not sure that it really helps with WMP though!
      Set the pkt->pts field so that we have a chance of doing frame rate reduction
      Set the pkt->pts field correctly so that frame rate reduction has a chance
      Initial versions of code to all per-frame video processing in ffmpeg.
      Compile framehook.c
      Build and install the vhook subdirectory if enabled (not by default).
      * Add frame rate conversion when there is no audio channel. This makes
      The problems of doing cvs add foo.*
      Added three sample video hooks. See the (rudimentary) documentation on what
      Updated to reflect the current state of ffserver (as I understand it).
      Applied changes to make it compile on darwin
      Try to get this to build on more platforms (especially Darwin)
      Add some rudimentary support for sparc64
      More fixes to compile and build on more platforms.
      More changes to make things compile on more platforms
      I goofed on the darwin LDFLAGS front. Till sourceforge install gcc3
      Add more tests for the presence of dlfcn.h and dlopen
      Moved hook documentation into the doc directory
      Simplify an expression and eliminate a compile warning
      Added support for a realease function to eliminate the context blocks used
      Added a release function to free up the context. Also removed the dummy
      Try to make this compile on platforms with dlfcn.h
      Fixed problem with frame rate reduction when capturing from a source
      Fix a crash caused by a null coded_picture pointer in ffserver.
      * Try and fix the crashes in ffserver associated with the coded_frame stuff.
      * Add code to allow the img reader to read files at the nominal frame rate.
      * Extend the syntax of a filename for the img reader to allow looping. Thus
      Add the -em_rate option to make the img reader run at the nominal frame rate.
      Initial version of test configuration file for ffserver that encodes the
      First shot at the ffserver regression tests. All feedback received with
      Make asf files work again -- the match against the URL was being done incorrectly.
      Added my TODO list
      Applied Mark Hill's patches.
      Make sure that if you abort the server regression tests, then it doesn't say OK
      Changed thru to through and fixed several other typos
      This fixes the jerky video from vob files. However, I am convinced that this
      Updated as a result of changes to the rate change code in ffmpeg
      Fix a crash in ffserver when streaming to WMP from a file. However, it
      Improve bind failure message to indicate the port that failed
      AUtomatic dependency generation. Runs on Linux (at least). Probably should
      When doing -vstats and also changing the frame rate, then don't count (and
      Fix the 'hard cpu loop' problem when capturing audio from /dev/dsp. This
      Fix a bug in the conversion of rgba32->yuv420p. This resulted in garbage images
      prevent segfault when passed no arguments
      Add checking for the existience of freetype2
      Add Makefile entry for the drawtext filter
      Initial checkin for a cheaper text drawing function than imlib2.c.
      Updated with many new features.
      Make ACLs work correctly for streams. Try comparing the addresses in
      This should fix the negative bandwidth in use in ffserver. Turned out that
      Try to eliminate memory leak on stream termination. It doesn't crash, but I'm
      * Add Video4MotionVector in a stream description to behave like the -4mv flag
      Applied Roumen Petrov patch to fix crash/memory leak
      Apply patch to remove -em_rate
      Fix a horrible bug with the launch keyword not handling arguments of length 1!
      Clear out an unused field in the created AVI file. This helps make
      Add a file serving test as well
      Only reference the images that exist
      These are my values and seem pretty stable. YMMV
      * Removed ffmpeg output during encoding runs. Now only print out errors
      * Actually make the sort function do some sorting (data needs to be on own lines)
      Made it work again -- it now detects fish! (This involved clearing out
      Have the test.asf stream enabled by default (thanks to htz at kde.org)
      Get all the stride stuff right so that we actually get the colours
      Minor bugfix to prevent segfault if a malloc fails. I have no idea what
      Fix a very nasty problem with extra bytes appearing in TCP data streams.
      Changed a bunch of calls to sprintf to snprintf to protect against buffer
      Add more resilience in reading ffm files. In particular, don't assume
      Remove .mpg streams due to a standards compliance issue
      ffserver now runs again (sort of).
      Add in many fields that have been added to the Codec structure. This means
      Fill out some mandatory fields in the Codec structure. Also add fields in
      Turn on the seeking support for ffm streams. Now means that streams are
      Flush the ffm packet to the wire (or file) whenever we flush the ffm packet.

Philipp Matthias Hahn (1):
      Out-of-directory build patch by (Philipp Matthias Hahn <pmhahn titan.lahn de>)

Philippe De Muyter (1):
      add support for grayscale rawvideo avi

Piero Bugoni (2):
      small updates taken from a patch by Piero Bugoni, crboca32 yahoo com
      Add watermark and drawtext descriptions, plus some random fixes.

Pierre Lombard (4):
      a few warning fixes (missing #include's)
      added missing #include <malloc.h>
      printf() requires stdio.h
      reversed for cause of code obsolescence

Quoc Cuong Pham (1):
      Detect "tif" suffix as TIFF.

R. Brian Anderson (1):
      Fix segfault

Rafaël Carré (2):
      Read asf file header fields in the right order
      Fix embarassing __GNU__ vs. __GNUC__ typo in preprocessor condition.

Ralf Terdic (1):
      Fix mem leak.

Ramiro Polla (257):
      Install into $PROGRAMFILES instead of a hardcoded path on MinGW.
      The FFLDFLAGS for MinGW are only meant for the shared libraries and not
      Use av_malloc instead of plain malloc where appropriate
      The fail test needs to be outside of the GPL test, it's referenced elsewhere.
      MSys bash 2.04 is broken.
      Have xvid encoder wrapper honnor -vtag option
      add interactive support for MinGW
      Remove MinGW install path special-casing.
      Remove useless MinGW #ifdefs.
      Install generated .lib files in shlibdir on MinGW.
      support files >2GB on MinGW
      Disable threading support by default on all platforms, add individual
      Segmentation fault fix when no video device is specified.
      BeOS cleanup
      Remove installstrip in favor of explicit strip; better for cross-compilation.
      Change all straight 'test $foo = "yes"' to 'enabled foo'.
      Change one occurence of 'test $foo = yes -o test $bar = yes' to
      Change occurences of 'test $foo != "no"' to 'enabled foo' for
      Add disabled() function and make use of it.
      Move libfaad2 check out of the GPL check.
      Simplify GPl check.
      move networking #includes into separate file
      Simplify libvorbis/libtheora check for libogg dependency.
      implement new grabbing interface, as described here:
      remove fallback definitions of PRIxNN macros,
      use FFMIN
      Remove superfluous HAVE_THREADS #ifdef.
      Add first and second output to earlyclobbers in COPY3_IF_LT macro.
      Cosmetics, patch by Ramiro Polla.
      Remove unused vars, patch by Ramiro Polla.
      Remove unused vars, patch by Ramiro Polla.
      Simplify, patch by Ramiro Polla.
      "fast unaligned" bytestream functions
      make 2 functions "return x;" to simplify next patch
      reorder bswap functions into bit-depth, special-casing inside the functions.
      Install MinGW shared libraries (dlls) to bindir
      Fix logic for return value of init_vlc
      Add another tmpbias variable, as bias' value will be used later
      Indentation for previous patch
      Add bias instead of -32 or 0, as is done in g.722.1
      Use 1 array with double the size instead of 2 arrays with normal size
      Factorize usum
      Use AV_XX16 macros
      Use bytestream
      Simplify ptr[0]; ptr++; to *ptr++
      Fix r8963
      Revert r8979 (Ugly fix for r8963)
      Remove duplicate bytestream functions
      Use AV_xx throughout libavcodec
      Remove get_byte wrapper
      Export av_init_packet
      Doxygenize av_init_packet
      Typos
      Remove unused declarations
      Print static vhook warning for Cygwin only when both are enabled
      Print static vhook warning for Cygwin and MinGW
      Make ppm vhook depend on fork()
      Vhooks require special cflags and ldflags for Windows builds
      Allow user to specify rgb.txt for imlib2 vhook
      Indentation for previous commit
      Typo
      Add expression evaluation parameters for colors in imlib2 vhook
      Indentation
      Add alpha channel support for imlib2 vhook
      tests/Makefile has been removed
      Set error parameter for ff_parse
      Remove localtime_r implementation for systems that don't have it.
      Check for O_BINARY instead of a list of systems that need it
      Add myself to Windows MinGW maintainers
      Add note about dlfcn-win32 for MinGW vhooks
      Remove OS/2 support
      Remove MinGW lrint hack.
      Remove unnecessary MINGW ifdef.
      Proper check for mkstemp().
      Proper check for interactive support (termios and conio).
      Check for termios before conio
      Remove underscores before fcntl flags
      Check for getrusage
      Collapse getutime into one declaration
      Check for GetProcessTimes and use it in getutime
      MinGW has gettimeofday() since mingw-runtime-3.10
      Move av_gettime() back to lavf/utils.c
      MinGW has those include files.
      Remove EXTRADEPS from common.mak since it is no longer used
      Add new imlib2 documentation from vhook/imlib2.c to doc/hooks.texi.
      Indent
      Simplify configure's output
      Indent
      Indent
      Correct misleading comment and typo
      Remove duplication of imlib2 vhook documentation.
      MinGW returns EAGAIN instead of EINPROGRESS
      Check for winsock2.h instead of __MINGW32__
      Add initialization and cleanup functions for Winsock
      Make the pipe URLProtocol share read and write functions with the file URLProtocol
      10l: removed pipe_close instead of pipe_read
      Enable swscale in MinGW vhooks only if swscale was configured.
      Also check for closesocket in winsock2.h
      Add attribute that forces alignment of stack to functions that need it.
      Introduce AVERROR_PATCHWELCOME
      Use AVERROR_PATCHWELCOME instead of AVERROR(ENOTSUP)
      CONFIG_7REGS -> HAVE_7REGS
      Set UDP receive buffer to 64k
      Enable MinGW networking support
      Group network checks together
      Remove --disable-dv1394 option
      CONFIG_7REGS has been renamed to HAVE_7REGS
      Treat dc1394 as a library
      Remove video_grab_ prefix from video_grab_{bktr,v4l}
      Give better filenames to bktr and v4l grabbers
      Remove --disable-{v4l,v4l2} options from configure
      Prefix library format names with 'lib' in libavformat
      Rename dc1394.c to libdc1394.c
      Separate audio_(de)muxer into oss_(de)muxer and audio_beos_(de)muxer
      Remove --disable-audio-beos option from configure.
      Remove --disable-audio-oss option from configure.
      Rename the format name audio_device (used in both oss and audio_beos) to oss and audio_beos.
      Bump micro version number for r10429
      Typo
      Split av_exit() out of main()
      Move av_exit() to before main()
      Remove --disable-bktr option from configure
      Indent
      Remove forced -pthread for freebsd.
      Properly check for -lm
      Use enable and disable macros where appropriate
      gpl for the X11 grabber has already been checked for
      Add extralibs to the dependency checker
      Replace one-lined if;elses to && ||
      Let the dependency checker decide about x11_grab_device_demuxer
      Proper check for --help and -h parameters
      Organize header files and system functions checks
      Let the dependency checker deal with winsock2's extralibs.
      Fix avisynth compilation after allformats.h was removed
      ffserver needs ffm_muxer specifically, and rtsp_demuxer for rtsp_parse_line()
      Fix memleak
      Remove gcc import libraries from Win32 builds.
      Add entry about pr not being present in standard MSYS install.
      Grammar.
      Append BUILDSUF directly to NAME instead of prepending it to (S)LIBSUF.
      Install MSVC import libraries without version numbering in the filename.
      Updated and rewritten Windows-related documentation.
      The Windows Help Forum is not ffmpeg-user for Windows
      Warn user if bitrate parameter is too low
      Speed up show_list
      List enabled code in configure output
      Express opinion in a less aggressive manner.
      os_support.h is also needed for usleep and lseek on MinGW.
      Use "disable" helper function
      Force DOS users to enable memalign hack when needed,
      __stdcall -> WINAPI
      Group network-related code together
      Put network-related code under #ifdef
      Include windows.h instead of trying to redeclare Sleep
      Define WIN32_LEAN_AND_MEAN before including windows.h to exclude
      Introduce max_picture_buffer variable in AVFormatContext,
      VFW capture support
      Typo + grammar
      Split libavdevice maintainers out of libavformat
      dc1394.c has been renamed to libdc1394.c
      add myself as vfwcap.c maintainer
      Correctly handle case where buffer is 100% full
      VFW grabber
      Remove windows installer. I'm not aware of anyone that uses it.
      Typo
      Mimic decoder.
      MSN TCP Webcam stream demuxer.
      Remove useless variable since r8456.
      Add Mimic and MSN TCP Webcam streams to documentation.
      Simplify returning errors by using goto
      Simplify more errors by using goto
      Cosmetics: remove {}s over one-lined ifs.
      Support biCompression I420
      Add #if 0 code to test for yet unsupported compressions
      typo
      Move av_exit() so it can be used by more functions
      Use av_exit() instead of exit() for a minimal cleanup
      make wininstaller has been removed in r12416
      Add note about using latest MSYS make for MinGW.
      Rename vlc1 to vlc.
      Cosmetics and style.
      Do not pass dct_block to vlc_decode_block().
      Fix bug introduced by r12827.
      Use bytestream functions for reading frame header.
      The code now is pretty self explanatory about the header structure.
      Fix regression introduced by r12929.
      get_vlc2() only gets up to three levels in the tables. The last codes weren't being read properly.
      Typo: distoration -> distortion.
      Ignore whitespace differences in seek test.
      Do not duplicate -y parameter on regression.sh.
      Do not read out of array bounds.
      Indent.
      Fix configure for cross-compilation and --arch=amd64.
      Prevent targaenc.c from outputting trash byte.
      Check for sys/resource.h and include it conditionally.
      Compile network-related code conditionally.
      Do not misuse long as the size of a register in x86.
      Also install .lib files with versioned filename.
      Move arch check before OS check.
      Add more support for Win64:
      Forgot to commit changes to subdir.mak in last revision.
      raw mlp demuxer
      Merge declaration and initialization.
      Remove unneeded NULL at end of struct.
      Remove useless img_read_close().
      output_example.c now uses proper include paths.
      Add note about not including MinGW's include directory from within MSVC.
      Remove wrappers of pcm_read_seek().
      Remove wrappers of pcm_read_seek().
      Make it a little easier to spot that the code is not dealing only with
      Seek the file forwards instead of backwards when sync is lost.
      Make ff_mlp_read_major_sync() take a GetBitContext instead of buffers.
      Clarify comment about parity nibble.
      MLP/TrueHD decoder.
      Update Changelog and documentation for MVI demuxer and decoder.
      Add a checklist about new codecs or formats to the documentation.
      Prevent users from setting -vol and -acodec copy together.
      bitstream: move put_sbits() from flacenc.c to bitstream.h and use it
      mlpdec: Check for bits left before each read of End-of-Stream indicator and
      mlpdec: Split filter parameters from context into their own struct.
      mlpdec: whitespace and line-breaking cosmetics.
      mlpdec: Split channel parameters from context into their own struct.
      mlpdec: Put some doxy comments to the right of vars instead of interleaved.
      mlp_parser: Initialize crc data in its own function.
      mlp: Split common code from parser and decoder to be used by encoder.
      mlp: Cosmetics: Close comments in an extra line and
      mlp: split simple inline function that xors 4 bytes into one.
      mlp: Define End-of-Stream code in common header file and use it in decoder.
      flacenc: Allow more flexible shift calculation in LPC.
      flacenc, lpc: Move LPC code from flacenc.c to new lpc.[ch] files.
      lpc: cosmetics: vertically align declarations and definitions.
      Import ok'd parts of ALAC encoder from GSoC repo.
      alacenc: NULL_IF_CONFIG_SMALL long_name.
      Import more ok'd parts of ALAC encoder from GSoC repo.
      Import more OKed parts of ALAC encoder from GSoC repo.
      alacenc: Use user-specified min and max prediction order.
      Update link to FFmpeg Windows Help website.
      Spelling...
      Specify minimum version for some MinGW packages.
      MinGW runtime 3.15 has a wrapper for usleep().
      Only special-case absolute DOS paths on systems that support them.
      nellymoserenc: flush PutBitContext after use.
      Support absolute DOS paths for preset parameters
      configure: Use check_lib2 for bzlib.
      configure: vertical alignment cosmetics.
      mlp: support bit-depths greater than 16 by default.
      mlp: cosmetics: remove {}s over one-lined if.
      mingw: update w32api to 3.13
      documentation: update documentation for minimal w32api version.
      Use posix_memalign() if available.
      configure: allow to disable sse code.
      Use setmode() if it exists in <io.h>, and not based on O_BINARY.
      Pass subarch instead of arch to YASMFLAGS
      mingw: don't check for minimum mingw runtime and w32api versions for mingw-w64
      indentation
      Do not misuse unsigned long to store pointers.
      Revert part of r16593.

Reimar Döffinger (306):
      width instead of chromWidth causing segfault in some cases
      100l, fix broken AMD64 patch. To whoever applied it: Did you actually _try_
      Compile fix on non-x86
      fix for MAP_ANON vs. MAP_ANONYMOUS fix...
      100l, forgot to change an ifdef on last commit
      Another REG_d -> REG_D fix.
      reset restart_count when restart_interval is set and fill the
      Add missing +1 in bounds check.
      Support de-/encoding of 24 and 32 bit PCM (from and to internal 16 bit).
      Support D-Cinema audio demuxing.
      Document "us" parameter for PCM conversion functions.
      Fix compile without CONFIG_GPL, misplaced #endif caused a missing }.
      check return value of decode_cabac_mb_dqp, it is INT_MIN on error.
      Change unsigned->signed and int->long, this fits the asm code better on 64
      clear the whole vert_pred buffer for 24 bit decoding.
      Fix wma2 audio decoder
      CamStudio decoder, only 32 bit lzo mode is tested
      Our own LZO (1X) implementation, under LGPL and optimized for readability.
      Wrong state handling causing decompression errors in some cases
      get_byte may not return -1 on error, since it can lead to a negative backptr.
      10l, get_byte returning 0 on error can cause a hang. So let's try with 1 instead...
      faster copy functions for lzo decoder that also need padding
      NuppelVideo/MythTVVideo support, including rtjpeg decoder
      make "size" variable in avpicture_alloc signed, since avpicture_get_size
      Avoid crash if there is a rtjpeg quant header but no video stream in file
      change the type of esp in SwsContext to uint64_t so it can hold a full register
      fix variable type used via "m" asm constraint to match size used in asm.
      read_packet return value must be < 0 when no packet is returned, including EOF.
      Ignore overrun in spectral RLE decoding, fixes
      Do not read full byte when less than 8 bits are still to be read.
      Fix initialization of vc1_decoder (the same as wmv3_decoder).
      add GXF demuxer
      Split common.h in two parts, purely internal stuff (internal.h) and things
      do not include bswap.h in common.h for external programs, since the former
      Missing part of the malloc -> av_malloc patch: memalign must be replaced as well!
      Try to find out correct start time to make seeking faster and add some
      Do not crash when the first frame is not a keyframe (and thus none of the
      Set duration to AV_NOPTS_VALUE instead of 0 when unknown.
      Fix typo in comment
      Support for seeking, both with and without index and correct timestamps
      Minor resync optimization
      cc -> $cc typo fix, caused amd64 to be misdetected as x86 when cc not available
      Fix get_bits_long to work with ALT_BITSTREAM_READER_LE.
      Changes needed due to show_bits_long behaviour changes/fixes
      Get rid of get_bits_long_le, get_bits_long was fixed and now behaves the same way.
      remove get_le16 and get_le32, get_bits and get_bits_long can just be used directly.
      Avoid void *-arithmetic
      CABAC assembler optimizations ported to AMD64
      Reenable AMD64 optimizations for cabac accidentially disabled in r6852
      Move idct block array variable onto rtjpeg_decode_frame_yuv420 stack since
      Avoid void*-arithmetic, and uint8_t* is more correct anyway.
      Fix typo in comment
      Correct GET/PUT_UTF8 comment: the get/put functions might be called up
      Set st->need_parsing = 2 for MPEG-content.
      Set CODEC_TYPE_DATA for timecode tracks.
      Change "p" asm constraints to "g", since "p" was a no longer necessary hack to
      Move BE_*/LE_*/ST*/LD* macros to a common place. Some further
      Typo in error message
      Fix 16 bit cscd samples, 16 bit raw means RGB555 on Windows, and the original
      Fix JPEG-LS encoder 0xff-escaping writing too much or uninitialized data.
      Check mb_pos is big enough before trying to access data left or above.
      Fix error message, it's package ref that can not be found, not track in this case
      Extend mxf_resolve_strong_ref by a type parameter, to avoid modify something
      Use macros to eliminate code duplication in MXF metadata read functions.
      128bit-only AES implementation under development.
      Simplify klv_decode_ber_length
      10l, > vs. >= typo, caused crashes on last mpc frame
      Buffer for get_bits must be padded because readers may overread.
      Make seeking work in files that contain encrypted tracks.
      Avoid calling decode_slice when context_initialized is not set.
      Demuxers with AVFMT_NOFILE will open a (possibly different) file themselves,
      Attempt to better document AVFMT_NOFILE.
      Fix crash when pred_order greater s->blocksize >> rice_order.
      Add lzo test code
      Fix buffer end checks in lzo copy code to work in all cases.
      Fix _LZO_H vs. LZO_H typo
      LZO optimization: check input buffer bounds less frequently if padded
      Add code to testcode to ease comparing with liblzo.
      Add two extra needed bounds checks
      Optimize LZO copy operations
      Documentation fix: Copy functions should "work" fine for cnt == 0
      Remove boundary checks that are actually done "well enough"
      Simplify checks, use that we know that cnt will not be < 0
      Make sure we do not accidentially "fix" cnt to something < 0
      Move lzo decompression to libavutil
      Fix incorrect use of ff_get_fourcc that made mjpgb not play on big endian.
      10l, forgot to skip payload description in UMF packet parsing
      set pts_wrap_bits to 32 since timestamps in packets are 32 bit
      Use av_set_pts_info and set some arbitrary timebase fallback
      special-case AES-128 decrypter is no longer necessary, the more general
      Minor AMD64 compilation fix
      add key field to AVFormatContext for de/encryption support, patch by Reimar
      mxf aes decryption support, patch by Reimar, simplified to only look for first crypto context, will be extended once we get files with multiple cryptocontext, and hope that they won't have broken container ul
      Add av_ prefix to clip functions
      Add av_ prefix to clip functions
      Fix a possible crash on 64 bit systems when the lumSrcPtr or chrSrcPtr
      Check dimensions to avoid division by zero due to invalid new dimensions.
      Fix the most obvious typos in the development policy doc section
      Fix parts missed in clip -> av_clip rename
      Fix two small typos in comment
      Do not use return with argument for a function "returning" void.
      Change some leftover __attribute__((unused)) and __attribute__((used)) to
      Use DECLARE_ALIGNED to ease porting
      Use DECLARE_ALIGNED for alignment in libpostprocess
      Move attribute_unused to variable declarations in macros instead of
      Part of MPlayer patch:
      Fix yuv2rgb C functions to work for widths divisible by 4 but not by 8
      Copile fix: missing libswscale part of ffmpeg r9322 TARGET_ARCH -> ARCH change.
      Missing libswscale part of TARGET_ -> HAVE_ change
      Fix @file in matroskadec.c, it was still matroska.c
      DV muxer should not only fail for more that 3 streams altogether
      Cosmetics: use consistent indentation by four spaces
      more av_strl* adjustments
      intreadwrite.h needs bswap.h if HAVE_FAST_UNALIGNED is set, so include it.
      Some more minor spelling improvements, some grammar improvement might be
      ffmpeg-issues subscription URL
      Document aes init code writing on purpose beyond round_key array into state array.
      First version of xsub decoder, not yet tested
      Add forgotten xsub timecode parsing
      100l, wrong argument to av_freep
      Hack to support XSUB subtitles in AVI
      10l, runbits order was reversed
      get rid of xsubdec array and calculate value instead
      get rid of av_rescale_q, it does not work as intended.
      Colours except background should not be transparent
      Deinterlace xsub subtitles
      Fix deinterlacing for odd height
      Remove AVCodecContext parameter from parse_timecode function
      Add proper license header
      Fix blend_subrect to respect the boundaries of the destination image
      Scale subtitle position
      Revert r10021, it was committed by accident and is a bit ugly anyway.
      100l in YUVA_IN macro: missing () caused horrible artefacts in blend_subrect
      Change imprecise "Unknown codec" message to say "encoder" or "decoder"
      Use defines instead of raw hex numbers to specify CRC polynomials
      Additional documentation for CRC functions
      Add request_channels member to AVCodecContext so we now have a proper
      Replace complicated and currently broken manual alignment code by
      Copy nuv header only when the video codec is nuv
      Simplify nuv: factor out LZO decompression
      First ugly and slow attempt to fix nuv files with extra frameheader
      10l, add check forgotten in last commit
      Fix extreme stupidity: buffer size check must of course be before buffer use!
      Make sure rtjpeg quality is a valid value
      Do not set codec_tag when the container does not specify it.
      Properly handle nuv file with changing resolution
      dsputil must be initialized before calling rtjpeg init.
      Set codec_id correctly for RJPG codec_tag
      Apply skip_loop_filter before checking if we can parallelize with the selected
      Remove uses of SIGILL for CPU extension detection, that method is not acceptable
      Hopefully fix sparc compilation again, add incorrectly removed variable declaration
      Allow to override zlib compression level in PNG encoder via avctx->compression_level
      Add a read_timestamp function to ogg demuxer
      Remove ogg seek function, instead use generic seek via read_timestamp function
      Add RC4 encryption/decryption function
      Fix rc4 header, #ifndef -> #define
      Add support for DES en- and decryption.
      Add support for decrypting asf files
      Document ff_des_encdec
      wma_decode_superframe always returns s->block_align, so make
      Fix nodes[nb_codes*2-1].count being uninitialized and used to initialize
      Add missing const to last argument of decode_significance_8x8_x86.
      Use the correct "const float *" type for variable instead of casting const away.
      Make ff_mkv_codec_tags array const, it is never written.
      Some spelling fixes in comments
      Make ff_qt_default_palette_* arrays "static const"
      Add a missing break, before av_set_number would always return NULL for
      Add 'l' suffix where it is necessary because type can not always be
      Add FF_OPT_TYPE_BINARY and use it to add a cryptokey option
      Simplify av_open_input_file
      Simplify av_close_input_file similarly to av_open_input_file previously
      Add a av_close_input_stream function
      Increase minor version due to new av_close_input_stream
      Add a av_alloc_put_byte function.
      100l, av_read_packet must check for read_packet error, in case of
      begining -> beginning typo fix
      Check for avcodec_open codec parameter == NULL and return error in that case
      Use FFSWAP instead of "manual" swap
      Get rid of fft_rev table, use ff_reverse and a shift.
      Add mfspr-based AltiVec detection code.
      Fix typo in comment
      Extract the pure plaintext from mov text subtitles.
      Use DECLARE_ALIGNED
      Add CODEC_ID_MOV_TEXT and use it instead of modifying the stream in the mov demuxer.
      Use DECLARE_ALIGNED and remove unneeded attribute_used
      Use DECLARE_ALIGNED in yet another place
      Add and use DECLARE_ASM_CONST for constants used in assembler code.
      Use DECLARE_ASM_CONST where possible in libswscale code
      Add sbsf option for subtitle bitstream filter
      Add bitstream filter to convert between mov and plain text subtitles
      Make some assembler constants global instead of declaring them multiple times.
      Support MP4 subtitles as in
      Simplify: use DECLARE_ASM_CONST
      Missed a variable in postprocess.c that should use DECLARE_ASM_CONST as well
      Two tables that should be const
      Add support for ogg text subtitles.
      Add some const keywords in rtjpeg decoder
      LZO decoder input is const
      Some consts for cscd decoder helper functions
      Missing 'const' in cast.
      Put is_mms under ifdef CONFIG_MMSH_PROTOCOL, avoids warning:
      Fix warning:
      Add some const where appropriate
      Remove a pointless cast.
      mark qtrle input data as const.
      Use bytestream_get_be24 to simplify palette parsing.
      100l, since we already check for buf_size == 769 we should also
      Remove ff_mms_set_stream_selection again from asf demuxer,
      Fix a possible use of an uninitialized pointer.
      String pointers of ogg_codec_t should have const attribute.
      Check that we have enough input data in IMC decoder.
      Add support for speex in ogg
      Fix misplaced const, avoids a cast-discards-qualifiers warning
      Missing const
      Add casts to avoid warnings:
      Remove duplicate ;
      Fix warnings:
      Missing const, fix warnings:
      Add some const, fixes warnings:
      __asm __volatile -> asm volatile, improves code consistency and works
      __asm __volatile -> asm volatile part 2
      __asm __volatile -> asm volatile part 3
      Use cltd instead of cdq asm mnemonic, ICC and gcc support both, but
      Fix decoding of interlaced gif, e.g. http://samples.mplayerhq.hu/GIF/7up.gif
      Do not call update_initial_durations if pkt->duration is 0.
      Add checks to ff_split_xiph_headers to ensure that returned header_len and
      Simplify ff_split_xiph_headers
      Check for av_new_stream failure, fixes CID76 RUN2
      Check av_new_stream return value, fixes CID79 RUN2
      Check av_new_stream return value, fixes CID75 RUN2
      Almost-cosmetics: split out coeff_count from all_fragments struct into
      Remove the START_TIMER/STOP_TIMER from vp3.c, they clutter the output and
      Fix add_bytes_mmx and add_bytes_l2_mmx for w < 16
      Fix pkt->pos to really point at start of packet for mtv audio packets.
      Change mtv_read_packet so it does not break after seeking (displaying a shifted image).
      10l, fix nuv_header for the case when there is no video stream.
      Change MS ADPCM table so they fit into int8_t and change array type.
      10l, forgot to replace a / 256 for r14173.
      100l, fix MS ADPCM decoding for e.g. http://samples.mplayerhq.hu/mov/qtaudio/surge-2-16-L-ms02.mov
      Use new style static VLC tables for IMC decoder.
      Fix MS ADPCM de- and encoding: first and second sample are swapped.
      Use bytestream and AV_RL* functions in ADPCM code where possible
      clip MS ADPCM predictor range to correct [0..6] range, not [0..7]
      check that csize in ff_lzw_decode_init is < LZW_MAXBITS, <= is not enough and
      Make the id3v1_genre_str array const, not just the strings it points to.
      Remove useless empty close functions in mmf and mm demuxers.
      Change codec_tag type from const struct AVCodecTag ** to const struct AVCodecTag * const *
      Make ogg_codec_t descriptions const
      Mark "matroska" probe data string constant.
      Mark several libavformat arrays const
      Add ff_ prefix to ogg_codec_t structs
      Change doxygen comment: Clarify when av_write_trailer should be called.
      Fix nuv decoder to use reget_buffer for non-keyframes and correctly
      Use "static const" instead of "const static"
      Add (additional) const to many global tables.
      Remove useless cyuv_decode_end function
      Make 8svx codec context table pointer const to match the type of the
      Use M_PI from libavutil/mathematics.h instead of defining PI ourselves
      Make size variables in dyn_buf_write unsigned so gcc will not optimize the
      Use x86_reg type instead of long in float_to_int16 MMX/SSE functions.
      Replace long with x86_reg in postprocess_template.c like in all other
      Replace pushf/popf by explicit pushfl/popfl (32 bit) or pushfq/popfq (x86_64),
      Remove useless and confusing (since they are also used in gxfenc.c) /* gxf.c */ comment.
      Use "struct gxf_stream_info" instead of "st_info_t",
      Avoid _t in typedef type
      Some forgotten eax -> REG_a changes.
      Also include sys/types.h when testing for socklen_t, since unfortunately
      Remove OpenBSD SLIBNAME magic, it seems unnecessary and breaks compilation.
      Fix regression.sh to work on OpenBSD which has md5 instead of md5sum
      Define _BSD_SOURCE in bktr.c, dev/ic/bt8xx.h needs the u_short etc. typedefs,
      Allow setting a different nm binary from the configure line.
      Avoid _t in gxf enum type names
      Bump minor number for many small improvements, tiff fax decoding,
      av_free() -> av_freep(), patch by Reimar Doffinger, see discussion in
      My gpg key fingerprint.
      Replace nonsense -1234 return value in dyn_buf_write by proper AVERROR(ENOMEM)
      Use av_mallocz instead of explicitly zeroing in url_open_dyn_buf_internal.
      Use AV_WB32 instead of reimplementing it.
      Return AVERROR(ENOMEM) instead of -1 when malloc fails in url_open_dyn_buf_internal
      Make io_buffer_size unsigned to avoid a warning about comparing
      Merge declaration and initialization of io_buffer_size
      Use AV_WL32/AV_WB32 in bitstream.h
      Add "const" to AES function arguments where possible without generating
      Add av_ prefix to LZO stuff and thus make it officially part of the public API.
      Move doxygen documentation from lzo.c to lzo.h
      Documentation for LZO error return value flags
      Add and use a public API for RC4 and DES, analogous to the AES API.
      Add support for 3DES to DES module
      Update DES test code to use the new public API.
      Fix a bug in 3DEC CBC decryption and add more extensive tests based on
      100l, don't run tests twice, this was some forgotten debugging code.
      Fix compilation, there is no file vdpauvideo.c, only vdpau.c
      Place spaces consistently in pointer types in vdpau.c
      Instead of crashing, return from ff_vdpau_mpeg_picture_complete()
      Support playback of incorrectly cut DV files where the headers do not
      Check return value of frame_start(), this avoids a crash if AVCodecContext::get_buffer failed.
      100l, always declare variables at the top of a block, not inside a for ()
      Check return value of ff_fill_pointer in avcodec_default_get_buffer,
      Detect the case when the time base is exact but far finer than necessary to
      Hack: #undef printf so compilation works with DEBUG_IPMOVIE set to 1
      Simplify ipmovie.c pts calculation by using an appropriate time_base.
      Do not allow standard-frame-rate detection to increase the value of
      Remove unused variable in mpeg_mc_decode_init
      Fix braindead and broken way to calculate abase, fixes regression tests on
      Make AltiVec code write alpha as 255 instead of 0 when converting to RGBA
      Add warnings to yuv2rgb_vis.c because alpha is set wrong (0 instead of 255).

Reinhard Nissl (5):
      scenario: A properly coded frame is followed by an end of sequence NAL unit,
      Remove a NAL unit's trailing zero bytes even when dst_length is 1.
      When dst_length == 0 bit_length has to be 0, too, but the current code still
      Provide top_field_first logic for h264
      Fix problem with multithreaded decoding, introduced by r11214.

Reinhard Tartler (1):
      Fix the use of Requires.private and Libs.private in generated .pc files.

Rene Rebe (1):
      Fixes rgb32to16 conversion for I think all platforms since the int8

Reynaldo H. Verdejo Pinochet (26):
      anothe useless sizeof(char)
      new mtv demuxer
      Adds forgotten changelog's entry about the new mtv demuxer
      adds missing rgb/bgr 555 codec->fmt mappings, ok'ed by Michael
      adds missing rgb/bgr 565 codec->fmt mappings, ok'ed by Michael
      Fix obvious typo, BGR555 was missing
      now we set codec_tag, still have to figure out how to handle flipping
      Please keep original copyright heading and attribution lines...
      Change to file description suggested by Michael
      Cosmetics
      Credit Kenan Gillet for his contributions towards merging the SoC QCELP decoder.
      Trivial, Cosmetics
      Trivial, Cosmetics
      Trivial, Cosmetics
      Trivial, Cosmetics, mostly brace placement changes
      Trivial rephrasing, avoids too long line
      Removes misleading const qualifier, gets rid of two compiler warnings
      COSMETICS, add missing remarks crediting Ben and Kenan
      Add expected const qualifier on 'buf' to match AVCodec.decode's declaration.
      COSMETICS
      Cosmetics by Kenan Gillet. Part 1 of 3 of his 'qcelp: silence handling'
      Part 2 of Kenan Gillet's QCELP silence handling patch.
      COSMETICS Part 3 and final of Kenan Gillet's QCELP silence handling patch.
      Add my key fingerprint to the list.
      Add myself as qcelp* maintainer.
      COSMETICS. Write my full name instead.

Richard Hodges (1):
      fix pts type in avcodec.h for WIN32 patch by (Richard Hodges <rh at matriplex dot com>)

Richard van der Hoff (1):
      segfault in amr decoder fix patch by (Richard van der Hoff - richvdh0)

Robert Edele (5):
      First part of a series of speed-enchancing patches.
      Snow mmx + sse2 part 2
      Add the mmx and sse2 implementations of ff_snow_vertical_compose().
      snow mmx+sse2 optimizations, part 4
      add MMX and SSE versions of ff_snow_inner_add_yblock

Robert Lippmann (1):
      Fix link fail of ffserver on mingw with --enable-vorbis patch by ("Robert Lippmann" <rlippmann at imsventures dot com>)

Robert Marston (2):
      Maxis XA demuxer and decoder.
      documentation update for new Maxis EA XA decoder/demuxer

Robert Swain (73):
      update x264 wrapper.
      update x264 wrapper.
      alphabetic reordering of codec registration
      enable bidirectional me with x264
      Update specification and reference implementation links.
      Make the Kaiser-Bessel window generator a common function
      Add variable alpha and size of half window for Kaiser-Bessel Derived window
      Grammar fixes and improvements for the new ACELP code
      Remove leading whitespace in long codec name that was committed in r13017
      Add generic ff_sine_window_init function and implement in codecs appropriately
      Remove now unused variable
      Correct a couple of typos
      Typo and grammar fix
      AAC tables and data that have been OKed by Michael
      AAC definitions and structures that have been OKed from the SoC AAC decoder
      Move shared tables to aactab.c and add declarations to aactab.h. Also sync with
      OKed sections of code from the SoC AAC decoder
      Sync already committed code with that in SoC and commit more OKed hunks of code
      Bump minor version as av_log_missing_feature() was added to public API in
      More OKed hunks of the AAC decoder from SoC
      Synchronise code with AAC decoder in SoC
      Change wma.c to use the ff_sine_window_init() from mdct.c
      Add declarations for the sine tables used in wma.c (half window sizes: 128,
      The ff_sine_#[] should be aligned as they will commonly be used in dsputil
      More OKed sections of AAC decoder code
      Synchronise AAC decoder code with that from SoC
      More OKed AAC decoder hunks
      Don't try to free shared ff_sine_* tables pointed to by s->windows[]
      More OKed AAC decoder code hunks
      More OKed AAC decoder code hunks
      Last hunk of the AAC decoder code to be OKed and build system and documentation
      Add me as maintainer of the AAC decoder code
      Correct regex file name matching as per Måns' suggestion
      Reindent
      Remove unused variable
      Work around GCC variable may be used uninitialised warning
      Reindent after last commit
      Separate code in imdct_and_windowing() into distinct IMDCT, windowing and
      Whitespace cosmetics after the last commit
      Use ff_imdct_half() and vector_fmul_window() for IMDCT and windowing. Reduce
      Fix index to ff_sine_windows[]. Previously the index was usually in reverse
      Remove unnecessary portion of a condition. This fixes issue607 and avoids
      Fix typo in element type
      Correct pulse amplitude application - a negative or 0 coefficient implies the
      Reindent after last commit
      Operands of '+' are not necessarily evaluated in order so r15311 was pointless.
      Reindent after last commit
      libx264 ffpresets:
      Cosmetics: missing brackets
      Zero 'idx' for each iteration of the coupling gain loop and as it is only used
      Make myself maintainer of the preset files
      Synchronise presets with recent subme/brdo/bidir_refine -> subme changes in
      AAC: Cosmetics after last commit
      AAC: Reindent after last commit
      Change direct prediction for fast first pass to automatic as all other presets
      Be consistent and use b-frames throughout the presets to avoid confusion about
      Add a slow first pass preset
      Add some lossless h.264 encoding presets as suggested by Jason
      Support ADTS AAC files in the ffaac decoder (limited to streams containing one
      Add presets to use as constraints for H.264 Baseline and Main profiles
      Add presets for Apple iPod constraints
      Add maxrate and bufsize specifications for iPod presets. I omitted these
      The default subq value in the x264 CLI is now 6, not 5.
      Add (un)installation targets for ffpresets using the new DATADIR variable
      Make opt_preset() use FFMPEG_DATADIR from config.h rather than /usr and
      Doxygen defgroups throw a warning if they don't have a title as well as a name
      AAC: Cosmetic alterations after the last commit
      Relinquish maintainership of adtsenc.c from Måns grasp :)
      Make myself maintainer of the website code
      http://bytesex.org/xawtv/ permanently redirects to
      The Ogg muxer will handle Theora as well
      Add some release notes for this 0.5 release branch
      Correct grammar in one sentence and add a note about doc/APIchanges

Roberto Togni (76):
      Don't memcpy from unallocated memory
      100l: free previous frame, not current one!
      Fix some palette related defines, bump build number
      Make avi and asf demuxer export palette in palctrl
      Moved to new palette API
      100l: sometime palette was not copied to output frame
      Add cr codec capability and buffer hints
      Set buffer hints, use cr where available
      8BPS decoder
      4bpp support
      - Add reget_buffer() function to AVCodecContext
      MSZH and ZLIB decoder support
      Fix for older version of zlib
      Right fix for older zlib version
      Use reget buffer instead of copying from prev frame
      Use av_log
      rle_code can overflow when multiplied by 4
      Add buffer hints, enable DR1
      buffer hints, use reget_buffer, removed copy from previous frame
      Add support for qtrle4 (16 colors/gray levels)
      100l fix codebook usage for v1 vectors
      fix for width or height not multiple of 4
      Force alignement for rpza and smc
      Paletted cvid support
      Ignore palette for grayscale files, all my samples are inverted with
      h261
      AVC (H264 in mp4 files, fourcc avc1) support
      Zlib encoder: fix image orientation (was flipped), 100l in deflate error
      Fix mjpeg-b (padding bytes can exist between end of sos and image data)
      Fix compilation when both amr_nb and amr_wb are enabled
      Fix avc1 if there is nore than one nal per mov frame
      Check pointers before writing to memory, fix possible integer overflows
      Use avcodec_check_dimensions instead of custom hack
      1l, forgot to remove duplicated error message
      Decode MP3 in ADU format
      Check pointers before writing to memory
      Add 32bit RGB support
      Multichannel mp3 in mp4 support ISO/IEC 14496-3:2001/FPDAM 3 (MP3onMP4)
      Fix colors for YV12 case (u/v planes are swapped)
      Put correct fourcc in tag field for amr_nb (samr) and amr_wb (sawb)
      Theora fourcc
      Reverse patch 1.59->1.60, it break mp3 in ffplay and ffmpeg
      QDM2 compatible decoder
      QDM2 compatible decoder
      Increase version because of QDM2 decoder addition
      Increase version: 'wave' atom parsing added to mov demuxer
      Place qdm2 and truemotion2 in ifdef block so that they can be disabled
      Remove duplicated fft init
      Support for samples with fft_order = 7
      Add cook decoder
      Add credit for cook
      Fix for Real "old" files version 3 with no 4cc. Fixes thankyou144.ra
      Support for 28_8 in old ar. files
      wma2 fix
      Add missing external declaration for kmvc codec
      10000l don't reuse outer loop index in inner loops
      Support for AAC (fourcc raac and racp) in rm files
      rm_read_audio_stream_info return type is not void
      flv files from myspace with mp3 audio require need_parsing because the
      Change 4xm demuxer and video decoder to pass the video format version in
      Fix seek regression test for ac3 in rm after r10634
      Move dnet-ac3 byte-swapping code close to audio packet read code
      1000l: Fix check for max slice number
      Optimize memory management to create an av_packet from multiple slices:
      Check for memory allocation failure of temp buffer
      Cleanup: remove ssize
      No need to update pos, it's not used after that line
      Cosmetics
      Replace realloc with free+malloc, the previous content of the buffer is
      Revert r10892, it's wrong and no longer needed to prevent crashes
      Typo
      Move printing of the license status at the end of configure output, so
      Add support for S8 PCM (codec 2)
      Fix forgotten braces from r11039
      Remove my old email address.
      Remove my old email address

Roine Gustafsson (17):
      missaliged clear_blocks() and h264 not complied but referenced fix patch by (Roine Gustafsson <roine at users dot sourceforge dot net>) and me
      add -vtag and-atag support to mov patch by (Roine Gustafsson <roine at users sourceforge net>)
      Small configure fix for Darwin / Mac OS X.
      - Writes correct unknown aspect
      --extra-cflags and --extra-ldflags clobbers previous options. This
      handle fixed sample_size patch by Roine Gustafsson <roine AT users.sourceforge.net>
      handle fixed sample_size patch by Roine Gustafsson <roine AT users.sourceforge.net>
      add luma only support to yuv4mpeg patch by (Roine Gustafsson <roine users.sourceforge net>)
      console text to stderr : rev2 patch by (Roine Gustafsson <roine users.sourceforge net>)
      mov header field and compressor name patch by Roine Gustafsson <roine AT users DOT sourceforge DOT net>
      chunk support in mov patch by Roine Gustafsson <roine AT users DOT sourceforge DOT net>
      movenc illegal access patch by Roine Gustafsson <roine AT users DOT sourceforge DOT net>
      support colon-separated rates patch by Roine Gustafsson <roine AT users DOT sourceforge DOT net>
      Update yuv4mpeg to use PIX_FMT_NONE patch by (Roine Gustafsson <roine users sourceforge net)
      Adds read probe to y4m, and changes the extension to .y4m patch by (Roine Gustafsson <roine users sourceforge net)
      ff_reverse to utils.c patch by (Roine Gustafsson )roine users.sourceforge net)
      a few #ifdef CONFIG_X_ENCODER, patch by (Roine Gustafsson <roine users.sourceforge net]

Romain Degez (1):
      RTP/RTSP and MPEG4-AAC audio

Romain Dolbeau (16):
      altivec jumbo patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
      PPC perf, PPC clear_block, AltiVec put_pixels8_xy2 patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      More AltiVec MC functions patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      Altivec on non darwin systems patch by Romain Dolbeau
      Altivec on non darwin systems patch by Romain Dolbeau (NOTE: I would prefer using the existing CONFIG_DARWIN Makefile variable and suppress the redundant TARGET_OS variable, but it would break the current mplayer patch)
      PPC fixes & clean-up patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      1) remove TBL support in PPC performance. It's much more useful to use the
      AltiVec support in postproc/ + altivec optimizations for yuv2yuvX patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      AltiVec hScale, all size patch by (Romain Dolbeau <dolbeaur at club-internet dot fr>)
      forgotten copyright & dct_quantize_altivec disabled patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      better hadamard8_diff16 in AltiVec, and more patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      Newer version, using a vectorized version of the
      Another (final?) patch for libpostproc.
      remove AltiVec.linux warning patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      Altivec unscaled YV12 -> packed YUV patch by (Romain Dolbeau <dolbeau at irisa dot fr>)
      h264 qpel mc, size 16 patch by (Romain Dolbeau <dolbeau at caps-entreprise dot com>)

Roman Shaposhnik (146):
      zero sized malloc patch by Roman Shaposhnick
      support for PAL 4:1:1 SMPTE 314M DV streams patch by (Roman Shaposhnick <rvs at sun dot com>)
      correct AUDIO strf parsing patch by (Roman Shaposhnick <rvs at sun dot com>)
      making configure a shell script not a bash one patch by (Roman Shaposhnick <rvs at sun dot com>)
      OpenDML AVI > 2Gb support patch by (Roman Shaposhnick <rvs at sun dot com>)
      AVI type 1 support patch by (Roman Shaposhnick <rvs at sun dot com>)
      making configure sh script patch by (Roman Shaposhnick <rvs at sun dot com>)
      AVI > 2Gb (OpenDML) generation patch by (Roman Shaposhnick <rvs at sun dot com>)
      dv file format support patch by (Roman Shaposhnick <rvs at sun dot com>)
      bunch of small thing patch by (Roman Shaposhnick <rvs at sun dot com>)
      Changes for SPARC/Solaris compatibility. Now it should be possible to
      Complete support for OpenDML AVIs and AVIs > 2Gb.
      * Making AVI encoding predictable (all JUNK chunks are filled with 0)
      * fixing NTSC frame rate for DV streams
      * making it possible to override aspect ratio from the command line
      making it nicer to the client who doesn't expect errors messages in stdout
      * introducing new public interface in imgconvert.c
      * fixing a bug in avpicture_layout (PAL8 wasn't handled properly)
      misc. updates
      * First round of AV sync fixes.
      A/V sync patch from Gildas Bazin.
      patch for DV capturing by Dan Dennedy <dan at dennedy dot org>
      * fixing a regression in mpeg encoder (not setting pix_fmt),
      enabling Sun medialib support via --enable-sunmlib, the default
      all human-readable output should go into stderr for now. We really
      * ffmpeg was *silently* rejecting to deinterlace NTSC DV. The problem
      * Phase 1 for DV encoding support. Muxing/demuxing of DV streams is now
      * making resolution abbreviations right
      1000l in abbrv. table
      * patch for compile problem by "Steven M. Schultz" sms at 2BSD dot COM
      * fixing a typo ;-)
      * fix for integer overflow
      * DV handling was streamlined for both muxing/demuxing and
      * The functionality is now available in dv.[ch].
      * preAlpha DV encoding support -- there's still a truckload
      * fixing bug that prevented resampling the picture and
      * 10000l (cut'n'paste is evil!)
      * recommiting change from SF CVS
      * fix for a trivial 100l declaration typo
      * original fix still didn't align lines on s_align when CODEC_FLAG_EMU_EDGE
      * providing MPEG codecs with a generic fields in AVFrame to use.
      * introducing dct248 into the DSP context.
      * sh doesn't understand $(command) type of Command substitution.
      * adding integer/floating point AAN implementations for DCT 2-4-8
      * DV demuxer is now capable of decoding auxilary audio stream. So,
      * fixes for broken builds on Solaris, OS2 and all bingendian
      * 3x encoding speedup. Finally we seem to be on par with libdv
      * Initial implementation of the G.726 ADPCM audio codec.
      * docs update for G.726
      * gotta setup coded_frame for encoding. avcodec.h says that for decoding
      * turns out write_packet is supposed to return 0 on success, not
      * moving some of the commonly used bit reading/writing functions
      * DV decoder simplifications. Now it looks to be 6% faster. At least
      * DV decoding/encoding now supports MultiThreading for up to 324 CPUs ;-)
      * removing casualties of battle of the wits and English language
      * enabling seek in raw DV files
      * turns out raw DV files can have .dif extension
      * some significant clean-up of the dv_decode_ac (it looks real
      * enabling DV-in-QT
      * making .mov files generated by ffmpeg compatible with Apple software.
      * fixing 10l in movenc.c + updating regression patterns.
      * .mov files with uncompressed audio can't be correctly processed
      * making it possible to specify recording date and time in a stream
      * Enabling UltraSparc stuff wit _SPARC64
      * reimplementation using mutexes and condition variables.
      * IEC DV seems to accept SMPTE way of tagging 16:9 too.
      * cosmetics -- force "command not found" not to be printed
      * patch from James Morrison (getting rid of broken code)
      * fixing a small quirk with DV audio muxing
      * fixing a buffer overrun in gif.c
      * fixing a bug in DV timestamp generation. Now libavtest doesn't
      * simplifying OpenDML AVI handling.
      * replacing calls to not-always-available gmtime_r with our own code.
      * adding Michael's explanation to the FAQ
      * I was finally convinced. And besides, if err it's better to err with
      * misc. fixes and hacks to improve timing detection in raw DV
      * a couple of silly defaults to make everybody happy.
      * fix for a DV regression introduced with the last round of avi_read_packet
      * fixing a regression introduced by the last seek-support patch
      * let DV muxer generate audioless DV streams. This might not be 100%
      * experimental dynamic audio stream allocation for DV demuxer. This
      * static allocation for audio packets. This will make it a little bit
      * fixing a problem with ffplay being stuck (and not responding to
      * Introducing new UI for grabbing.
      * Introducing IIDC1394 grabbing interface.
      Moving dv_anchor back to the global scope. This creates a tiny memory
      Not that we're that far from being an *initial* implementation, but
      I still don't think this should be here. Note: without this code DV encoder
      * fixing -target *dv issues with not setting pix_fmt and adding dv50 target
      * fixing DV-in-AVI type1 packet production bug
      * Restructuring the division of labor between DV codec and DV format
      * Moving FifoBuffer out of libavformat/avformat.h and
      * Simplifying the setting of STYPE for AAUX AS
      * Fixing seeking with DV-AVI (by Jeff Downs <heydowns at borg dot com>)
      * Fixing a bug with incorrect bits set in AAUX source pack
      * Getting rid of the use of GCC language extensions
      * making profile selection a bit more tolerant to the badly broken streams
      replacing the use of deprecated unaligned*() routines by LD64 and AV_RB32
      * getting rid of code duplication
      * Making [START|STOP]_TIMER work on architectures that support gethrtime()
      * renaming (ST|LD)(16|32|64) -> AV_(R|W)N(16|32|64)
      * reinserting gethrtime in alphabetical order
      restoring the alphabetical order (this time with case ignored)
      * fixing a bug preventing default values to be set correctly
      Coalescing the un-weighting and de-quantization steps for faster
      Fixing the units in DV50 tables (both coordinates are
      Introducing DV_MAX_BPM macro instead of a hardcoded value for the
      Making the number of blocks per macroblock dependent on the DV stream
      Making block size in bits variable and dependent on the DV spec
      Simplification of how dct_mode affects decoding.
      Factoring out chan_slice / 27 into seq variable
      Cosmetic change aimed at making it easier to see how bits are consumed
      Two cosmetic changes for improved readability and ease of HDVCPRO HD integration
      Introducing video_stype for holding the value of VAUX source pack
      Intial implementation of the DV100 (AKA DVCPRO HD) decoder and demuxer as
      simplifying the code as per Michael's suggestion
      Doxygenizing the comments
      fixing a bug that prevented proper unweighting table to be selected
      unrolling the loops as per  Michael's suggestion
      vertically aligning as per Michael's suggestion
      cosmetic updates as per Michael's suggestion
      simplifying code as per Michael's suggestion
      replacing frame_rate and frame_rate_base with an AVRational time_base
      renaming macro in order to conform with the rest of libavcodec
      implementing more efficient (and direct) allocation of work for DV codec workers
      Making it easier to send arbitrary structures as work orders to MT workers
      Taking advantage of the new ->execute API
      Getting rid of unneeded extra wrapper functions
      Getting rid of huge static DV tables
      Change of indentation
      Making idct factor tables dynamically allocated
      additional recovery for the badly broken streams
      Replacing the constant with a macro
      Cosmetics: moving the function definition around
      Cosmetics: replacing 0 with NULL for pointer assignment
      factoring code into dv_init_enc_block
      Cosmetics: fixing the indentation
      simplifying DIF encoding process
      Cosmetics: simplifying static initialization
      unrolling encoding loops
      Cosmetics: fixing indentation
      No need to compute stype
      Making the arrays accomodate an extra intra 8x8 cmp function
      Adding 2 intra 8x8 cmp functions: vsad, vsse
      transitioning dv_guess_dct_mode to dsputil cmp function
      Fixing a value returning issue

Ronald S. Bultje (163):
      fix global name conflicts patch by ("Ronald S. Bultje" <R dot S dot Bultje at students dot uu dot nl>)
      non-avc h264 stream problems patch by ("Ronald S. Bultje" rbultje, ronald bitfreak net)
      add support for HTTP seeking
      makes the filename member of the URLContext a pointer, so that the
      proper error handling in file size retrieval, patch by Ronald S. Bultje rbultje at ronald bitfreak net
      protect the size=seek(SEEK_END,-1)+1 results with an if (.. < 0), else the
      Fix misbehaviour in url_fseek() when seeking fails.
      make HTTP MOVED messages also work in ffmpeg (302 is essentially the same as 303).
      Close connection on EOF
      Properly initialize filesize during input buffer initialization.
      Add libavcodec to compiler include flags in order to simplify header
      Do not specify directories in #include path, -I flags take care of this.
      emms --> emms_c, patch by Ronald S. Bultje, rbultje ronald.bitfreak net
      Fix compilation when MMX is disabled.
      CPU feature definitions should always be available, not just when the
      Add CPU feature definitions for SSE3 and SSSE3.
      Use FFmpeg-specific CPU feature definitions.
      emms --> emms_c, taken from a patch by Ronald Bultje
      Add some #includes to allow compilation without HAVE_AV_CONFIG_H.
      Adapt to API changes, fix compilation.
      Add a bounds check on AVProbeData input.
      remove decode_be32/64 and replace them by AV_RB32/64
      remove ENABLE_SWSCALER conditions
      to be able to compile without HAVE_AV_CONFIG_H
      Use AV_RB* macros where appropriate.
      Fix get_str/get_str8() to also work if the target string is not long enough to
      Merge get_str() and get_str8() functions.
      refactor url_split(), preparing for IPv6 support
      Add IPv6 support to url_split()
      Use emms_c() instead of ifdef
      cmdutils.c uses FFMPEG_CONFIGURATION from config.h without directly including
      Add checking for codecid to has_codec_parameters().
      Make url_split() strip url options (?opt=var) from the returned hostname or
      Take the target address family in account when determining the family of
      Split out the MDPR chunk reading into its own function.
      Specify the server address when opening an rtp:// URL in rtsp.c, so
      Change parse indexer.
      Split out the packet parsing from the main function body in rmdec.c
      Change a stream indexer.
      Isolate caching of audio frames in its own function.
      Split ac3 byte swapping into a separate function.
      Merge double-nested if into && condition.
      Drop RTSP default protocol.
      Make three rm demuxer functions non static.
      Make rtp_parse_packet() always call the vfunc of the dynamic payload handler
      Reindent after r11493 (always use parse_packet() vfunc in rtp_parse_packet()),
      Add a flags field to the RTPDynamicPayloadPacketHandlerProc (PKT_FLAG_*).
      Split the SETUP request into a separate function, as a prelude into allowing
      Allow cycling between different protocols (TCP, UDP or multicast) so that if
      Change protocol_mask into protocol, since we always just try a single one per
      Reindent after rtsp-alternate-protocol* patches.
      Allow NULL for read_header in AVInputFormat. See discussion in "[PATCH] allow
      Reindent after r14059, see "[PATCH] allow empty params for
      Set AVStream.discard, so that demuxers can choose which stream to select if that is a serverside option (as is the case for RTSP). See "Realmedia patch" thread on mailinglist.
      Implement url_open_protocol(), which is basiclly the former url_open()
      Remove useless comments. See "[PATCH] tcp.c/udp.c memleak?" for discussion.
      Fix memleak on some OSes in case network initialization fails. See
      Move malloc() down until after all initializations, so that the resource is
      On failure, return directly because the fail: case does nothing. This also
      Cosmetics after previous patches.
      Remove check for @ in tcp.c which removes the authorization data from the
      Use RTPDynamicProtocol parse_sdp_a_line() handlers in case of unknown SDP
      Export data_to_hex() as private API in lavf, rename to ff_data_to_hex() and
      Change implementation for ff_data_to_hex(), this is faster. See discussion on
      Implement RTSP/Realmedia-compatible OPTIONS command. See "Realmedia patch"
      Revert back to old version (r15103).
      Implement RTSPServerType enum as a way to identify the flavour of RTSP that
      Read RealChallenge1 field from the server.
      Send RTSP OPTIONS command to detect server type.
      Implement Realmedia-compatible DESCRIBE command.
      Some RTSP streams use SDP lines longer than 1024 bytes, so the SDP line
      Remove useless "else" case in if X { A; return }; else { B }. See discussion
      Use chunk-size in function calling mdpr_read_codecdata() rather than in the
      Move av_set_pts_info() inside the mdpr_read_codecdata() call, so that it is
      Implement Realmedia/RTSP-compatible SETUP command. This includes calculation
      Remove unused code that used to handle protocol concatenation, i.e. trying
      Give register_dynamic_payload_handler() in rtpdec.c a ff_ prefix and export
      Add dynamic payload handlers to rdt.c. These follow the same API as the ones
      Implement a RDT-specific SET_PARAMETER command that subscribes to the
      Reindent after previous patch.
      Implement RDT-specific data parsing routines. After these changes, simple
      Factorize out common code for opening of the RTP parsing context between
      Reindent after r15927, see discussion in "[PATCH] rtsp cleanup part 1:
      Rename RTSP_SERVER_RDT to RTSP_SERVER_REAL, because RDT (the transport
      Reindent after r15317.
      Rename RTSPProtocol to RTSPLowerTransport, so that its name properly tells us
      Send improper UDP SETUP request, which is what Realmedia servers expect.
      Remove access to rdt_data struct in functions called outside of the
      Implement a RTSPTransport field, which allows proper separation of server
      Reindent after previous patches.
      This patch refactors RDT packet header parsing so that it can be used in
      Use parse_packet vfunc in RTPDynamicProtocolHandler instead of hardcoding
      Rename RTP payload contexts to PayloadContext, suggested by Luca in
      Change function prototype of RTPDynamicPayloadHandler.parse_packet() to
      Remove access into RTPDemuxContext in rtsp.c, which allows making it opaque
      Make RTPDemuxContext opaque in rtsp.c, renaming it to tx_ctx (tx=transport)
      Implement RDTDemuxContext, which contains RDT-specific data (similar to
      Reindent after r15544.
      Add comment describing the RDT packet header. See discussion in "[PATCH]
      Rename variables in ff_rdt_parse_header() and callers to match the
      Call check_back_and_send_rr() function only in case of RTP as a transport.
      Add is_keyframe param to ff_rdt_parse_header(). See ML discussion in
      Reindent after r15833.
      Make function comment a bit more linguistically correct.
      Change status packet skipping to be more spec-compliant. See discussion in
      Add ByteIOContext argument to public ff_rm_* functions so that we can
      Fix memleak caused by the fact that url_open_buf() allocates a context
      Prefix all ff_rdt_parse_header() arguments with a p, preparing for local
      Change header reading implementation to using get_bits() instead of directly
      Read optional components of the RDT packet header, such as extended setID
      Reindent after r15853.
      Change type of prev_stream_id from uint32_t to int, since it has a max size
      Document 'ic' variable. See "[PATCH] RDT/Realmedia patches #2" thread in ML.
      Change function prototype from taking an AVStream to taking an index to the
      document ff_rdt_parse_open(). See ML discussion in "[PATCH] RDT/Realmedia
      Change RDTDemuxContext->st into an array of streams, which will make future
      Implement rule-number parsing, the initial step in stream (and bitrate)
      Do not access s->gb directly in functions that have a local copy of that
      Change function prototype of the sdp_parse_a_line in DynamicProtocolHandler.
      Remove static variable and a printf using it. See "[PATCH] asf.c: move
      Fix indenting.
      Separate the packet fetching from the data reading, so that the data reading
      Don't access RMContext directly in rdt.c. Rather, use the return value of
      Split RMContext into RMDemux/MuxContext and make them private in rmdec/enc.c.
      Increase SDP line buffer size because ASF headers in RTSP-MS are very big. See ML discussion
      Add comment to indicate why the SDP line buffer is as big as it is.
      Reindent something that looks weird.
      Use get_buffer() and url_fskip() for some loops of get_byte()s. See discussion in ML thread
      Add a ByteIOContext argument to ASF functions that will be shared with the
      Make asf_get_packet() and asf_parse_packet() public for use in RTSP-MS and
      Implement RMStream stream-specific private data object in the RM demuxer.
      Add RMStream object as function argument to public functions so that non-.rm
      Change audio_pkt_cnt from an array into a single field, since only a single
      Add audio_pkt_cnt return value to ff_rm_retrieve_cache(). See discussion in
      Reindent after r16368.
      Fix double free which happens for playback of .rm files after r16365. See
      Parse the ASMRuleBook SDP line to dynamically create one new AVStream for
      Parse the bitrate field in the ASMRuleBook ("AverageBandwidth") to fill in
      Parse the OpaqueData field for every AVStream represented by this "set of
      Allow subscription to any of the streams, not just the first, available in
      Don't use s->pb if a local ByteIOContext was already provided by the calling
      Export RTSPState and RTSPStream from rtsp.c into rtsp.h. This allows future
      Use the "server" RTSP field to detect whether the server that we're talking
      Fix doxy comments missing one '*'.
      Fix typo ("Standard-compliant" -> "Standards-compliant"), as noticed by
      Increase buffer size for RTP packet data because some ASF streams use a
      Apply rtpmap: SDP lines to the last m= line only, since they generally just
      Reindent after r16509.
      Reindent to properly fit a 80 chars terminal.
      Skip m= blocks in the SDP if the media type is unknown. This prevents
      Fix the Transport: line in the SETUP request so that it works with WMS
      Move enum RTSPTransport up a bit, so that all fields that are assigned a value
      Delete an enum and a function typedef that aren't used anywhere, and
      Rename "tx_ctx" and "cur_tx" variables to "transport_priv" and
      Add "AVFormatContext *ctx" (that being the RTSP demuxer's) as first argument
      Don't install rtsp.h. It is intended to be private, it depends on rtp code
      Rename RTSP_*_LAST to RTSP_*_NB in line with PIX_FMT_* in lavc. See "[PATCH]
      Change sizeof(struct_type) to sizeof(variable).
      Rename RTSPHeader to RTSPMessageHeader to reflect more clearly what the
      Document rtsp.h, see "[PATCH] document rtsp.h" thread.
      Don't (ab)use PKT_FLAG_* in RTP code, since the two have virtually nothing
      Implement marker bit, which is used for several RTP payloads currently
      Rename movenc.c MOVContext to MOVMuxContext, since MOVContext is already used

Roumen Petrov (2):
      don't use proxy and suppress verbose output patch by (Roumen Petrov <help at roumenpetrov dot info>)
      restore old tty in ffmpeg patch by (Roumen Petrov <ffmpeg at roumenpetrov dot info>)

Roy Sigurd Karlsbakk (1):
      largefile support patch by Roy Sigurd Karlsbakk <roy at karlsbakk.net>

Ryan Martell (12):
      Add support for H264 over RTP
      indentation fix, patch by Ryan Martell % rdm4 A martellventures P com %
      move up (in the file) AAC audio handling moved from below
      fix more dynamic protocol stuff, needed by the forthcoming h264
      switch to common base64 routines in lavf
      add valid statistics for the RTCP receiver report.
      Add support for getting duration of a RTP stream (for seeking in stream)
      add support for rtp/h264 streaming
      remove duplicate code, patch by Ryan Martell rdm4 A martellventures P com
      Minor modifications to handle mms streaming.
      Add the stream_bitrate GUID to the asf parsing code, and set stream bitrate
      Fix A/V sync error in ffplay with wmv3/wmav2 medias

Ryutaroh Matsumoto (1):
      mimetype fixes patch by (Ryutaroh Matsumoto <ryutaroh at it dot ss dot titech dot ac dot jp>)

Rémi Guyomarch (1):
      NAN doesnt exist on FreeBSD patch by (Rémi Guyomarch <rguyom at pobox dot com>)

Sam Hocevar (7):
      build error on Alpha patch by (Sam Hocevar <sam at zoy dot org>)
      minor compilation fix
      Add a leading zero to the sequence files generated by the testsuite
      Tell configure about hppa, s390, mipsel, m68k and ia64 architectures.
      Proper mangling information for symbols in OS X Mach-O shared objects.
      Do not return invalid pointer for non-audio or video streams.
      Add support for grayscale MJPEG streams sent by Axis cameras such as the

Sascha Sommer (11):
      10l another int i missing (without ^M)
      mangle on mingw32 patch by ("Sascha Sommer" <saschasommer at freenet dot de>)
      #undef DATADIR before inclusion of windows.h patch by ("Sascha Sommer" <saschasommer at freenet dot de>)
      libavcodec Cygwin compilation failure (II) patch by (Sascha Sommer)
      cygwin patch by ("Sascha Sommer" <saschasommer at freenet dot de>)
      nut files in cygwin patch by ("Sascha Sommer" <saschasommer at freenet dot de>)
      fix use of uninitialized variables
      rl2 decoder
      rl2 demuxer
      rl2 documentation update
      added my gnupg fingerprint

Sebastian Jedruszkiewicz (1):
      indeo3 for bigendian patch by (elf at frogger dot rules dot pl    Sebastian Jedruszkiewicz)

Sebastien Bechet (3):
      I420 patch by (Sebastien Bechet <s dot bechet at av7 dot net>)
      QT AVID 4CCs patch by ("Sebastien Bechet" <s dot bechet at av7 dot net>) (not tested)
      2000/00/00 to 2001/03/01 QT fileformat documentation patch by (Sebastien Bechet <s.bechet at av7 dot net>)

Sebastien Lucas (1):
      add MMX version of vp6_filter_diag

Sergey Vlasov (3):
      DTS decoding was broken for every speaker configuration except 5.1.
      Fix dca.c compilation with #define TRACE
      fix dca_bitalloc_index decoder init

Shane (1):
      Fix intended order of operations for 4 assert() checks.

Siarhei Siamashka (9):
      armv5 optimizations
      Proper armv5te instructions support detection.
      add FASTDIV macro for ARM. Reported speed-up on MPEG-4 decoding: 1.8%
      Optimization of 'dct_unquantize_h263_intra' for ARM (armv5te)
      fix lowres decoding support on ARM CPUs
      Add some initial optimizations for ARM VFP (floating
      OGG: Add padding at end of Vorbis headers
      Use DECLARE_ALIGNED_16() on vorbis windows as they are used in dsputils
      Interleaved forward/backward channels processing in order to increase chances

Sidik Isani (1):
      dont add padding in the middle of the data patch by (Sidik Isani <isani at cfht dot hawaii dot edu>)

Sigbjorn Skjaeret (2):
      restore GCC2/3 support, patch by Sigbjorn Skjaeret %sskjer-1 A broadpark P no%
      kill one vector constant value load by the right combination of vec_splatX/vec_sl

Sigbjørn Skjæret (3):
      fixes transcoding to vorbis with ffmpeg on big endian machines patch by (Sigbjørn Skjæret {sskjer-1 broadpark no})
      Disable AltiVec IDCT for lowres decoding in lavc patch by (Sigbjørn Skjæret: cisc, broadpark no)
      Add a couple of missing consts.

Sisir Koppaka (4):
      BFI demuxer
      BFI video decoder.
      Add .longname to bfi_decoder declaration.
      Add complete multimedia Wiki URL, patch by Sisir Koppaka.

Slavik Gnatenko (2):
      os2 support patch by ("Slavik Gnatenko" <miracle9 at newmail dot ru>)
      restoring OS/2 compatibility patch by ("Slavik Gnatenko" <miracle9 at newmail dot ru>)

Stanislav Brabec (2):
      Fix warning:
      Replace hackish support for amr-nb and amr-wb. Instead of including the source

Stefan Gehrer (68):
      staric frame_rate_tab -> non static ff_frame_rate_tab part of the cavs patch by (Stefan Gehrer stefan.gehrer gmx de)
      CODEC_ID_CAVS (part of the cavs patch by Stefan Gehrer stefan.gehrer gmx de)
      cosmetic patch:
      cleanup macroblock layer: merged decode of skipped MBs
      missed removing this line in my last commit
      even more cbp safety
      make use of 16x16 motion comp functions, 16x8 and 8x16 do not seem to be worth it
      hopefully better for platforms not supporting misaligned read
      avoid overflows of qp and pic_type
      set mvP2 also in PSKIP macroblock
      change partition flags SYM -> FWD|SYM
      replace run by run+1 in VLC tables
      avoid double copying of left border in luma prediction
      some MMX optimizations for the CAVS decoder
      add new tag 'CAVS' for Chinese AVS codec
      make mpeg4video_split public as ff_mpeg4video_split
      move all cavs-parsing to cavs.c
      rename ff_cavs_flush to cavs_flush and make it static
      fix comment, get/show_bits(0) does not work
      make vector constants global
      move context structure definition from cavs.c to cavs.h
      move existing inline functions from cavs.c to cavs.h
      move preparation of chroma intra prediction into its
      move modification of intra prediction modes into
      setting intra prediction modes to default values
      setting special motion vectors in intra macroblocks
      cosmetic: rename VLC tables to make clear they
      move dequantization into its own inline function
      prepare splitting decoder-only parts into own file:
      prepare splitting decoder-only parts into own file:
      split decoder-specific parts into their own file
      have less functions as inline
      mark read-only data as const
      revert accidental const slipping in
      remove duplicate tables
      make ModeAlphabet read-only and use a custom mode alphabet
      move ff_log2_run to bitstream.c and reuse in ffv1.c
      fix size of header data
      reduce code duplication by moving common header parsing
      add 1bpp decoding function and extend 4bpp function to
      rename vector_t to cavs_vector
      rename XXX_t enums to cavs_XXX to fix POSIX namespace issue
      use assignment of structure instead of memcpy
      remove duplicate tables
      added demuxer for FunCom ISS audio files,
      add myself as maintainer of ISS (with Jai's blessing)
      fix header parsing, the NULL terminated File ID is followed by a space
      return an error if we can not get a full packet,
      avoid duplication of dquant_code
      avoid duplication of size_table and offset_table
      remove zigzag_scan8x8, it is the same as ff_zigzag_direct
      remove alloc_table_0 and alloc_table_2, the same
      fix doxygen file reference
      avoid indeterminate order of accessing src
      remove duplicate tables
      ADPCM found in Xbox wav files can be decoded with
      check validity of pointer srcC
      do not check for minimum slice start code, 0x00000100 is valid
      time_code has 24 bits, not 16
      avoid duplicating dsputil's clear_block
      fix parsing of pic_structure
      fix multiple slices per frame
      introduce a macroblock index to avoid a few x*width+y calculations
      add raw demuxer for Chinese AVS elementary streams
      consistent naming of Chinese AVS raw demuxer
      mention raw Chinese AVS video demuxer
      change macroblock loop in inter frames so that after a
      check for a stuffing byte at the end of a slice

Stefan Huehner (8):
      Patch by Stefan Huehner / stefan % huehner ! org \
      Add (mostly) const to variable and parameter declaration, where a char* was
      Add const to (mostly) char* and make some functions static, which aren't used
      Remove unused variable, patch by Stefan Huehner stefan__at__huehner__dot__org.
      Mark read-only data as const. Patch by stefan huehner org.
      Mark some read-only datastructures as const.
      changes some function declarations from () to (void) as per ansi c.
      Make some functions which aren't used outside their declaring source file

Stefan Lucke (1):
      Prevent infinite loop when buffer holds SEQ_END_CODE only. Fix for r9870

Stefan de Konink (2):
      Implement the reading of the video bitrate of flv movies out of the meta data,
      Splits the totalbitrate in a videodatarate and audiodatarate part.

Stefano Sabatini (331):
      add a vstats_file command line option
      add a vstats_file command line option
      Add support for listing the supported pixel formats using the option
      Move the video size and rate abbreviations system from libavformat to libavcodec
      replace calls to deprecated parsing functions to calls to new ones
      split frame rate and frame size abbreviation into two structures
      me option implemented as an AVOption
      Make ffmpeg and ffplay exit 0 after displaying help
      Prefix with "opt_" the functions ffmpeg.c:show_{version,license,formats}.
      Make ffmpeg -formats exit 0.
      Make show-license exit 0.
      Make show-version exit 0.
      Make ffserver -h exit 0.
      Make cmdutils.c:parse_options accept as argument a function to parse
      Rename parse_arg_file function to opt_input_file for consistency.
      Implement the cmdutils.c show_license function.
      Make ffmpeg use cmdutils.c show_license function
      Make ffserver use cmdutils.c show_license function
      Cosmetic, change "fvstats" to "vstats_file".
      Document libavutil/mem.h:av_strdup.
      Make distclean should remove .vhookdep.
      Define mandir as $(PREFIX)/share/man rather than as $(PREFIX)/man
      Add explanatory comments to the parse_date function.
      Make parse_date return INT64_MIN in case of unparsable input.
      Implement common show version and banner.
      Make ffplay show a banner at startup.
      Remove unnecessary copyright notice from ffplay.c:show_help.
      Make ffserver show a banner at startup.
      Remove redundant banner output from ffserver.
      doxy/cosmetics fixes. Patch by Stefano Sabatini %stefano P sabatini-lala A posteP it %
      Doxyfication, patch by Stefano Sabatini %stefano P sabatini-lala A poste P it%
      Improve unsupported (pixel) format error message.
      Fix nellymoser decode_tag return value, patch by Stefano Sabatini
      Move up opt_show_license to get rid of ugly forward declaration.
      Fix wrong error message.
      Clarify documentation for avcodec_encode_audio.
      More explicit unsupported pixel format error messages.
      Remove a superfluous call to atoi in ffmpeg.c:opt_verbose
      Export and document in cmdutils.h the function parse_number_or_die().
      Make parse_options() use parse_number_or_die().
      Clarify the documentation for 'ffmpeg -v'.
      Provide a context to parse_time_or_die().
      Make option time parsing functions take argument from options
      Send context to the opt_* functions in ffplay.c
      Export parse_time_or_die from ffmpeg.c to cmdutils.c
      Make ffplay.c:opt_seek use parse_time_or_die
      Send context to ffplay.c:opt_seek
      Make the install-man target depend on $(MANPAGES).
      Add a missing comma in tiff_encoder.pix_fmts
      Complete argument information in ffmpeg.c:options for non-number arguments
      Mark option arguments as variables using @var{}.
      remove useless extern keyword.
      Add descriptive long_name to AVCodec
      Typo
      Make two comments doxygen compatible.
      Make some ffmpeg opts use parse-number-or-die.
      Add supported_samplerates field to AVCodec
      Send context to some ffmpeg.c opt* function
      Enhance documentation for AVClass
      Add long names to many AVCodec declarations.
      Make 8SVX codec names just a single word for consistency with other codec names.
      Add long names to AVCodec declarations.
      Add long names to some AVCodec declarations.
      Transform codec names into single words.
      Add some long names to AVCodec declarations.
      Add long names to some AVCodec declarations.
      Transform codec short name into a single word.
      Document AVCodecContext channels field.
      Fix a typo.
      Make 'ffmpeg -formats' print codec long names.
      Update offset in the option string later, to have a better error message.
      Add codec long name, patch by Stefano Sabatini, stefano.sabatini-lala poste it.
      Make av_set_string() fail when number could not be set.
      DPCM codec long names, patch by Stefano Sabatini, stefano.sabatini-lala poste it
      cosmetics: s/Id/id/ in libavcodec where Id refers to id Software.
      cosmetics: Make 'ffserver -h' output consistent with other ff* tools.
      Let ffplay support "-version".
      Improve parse_primary() error message.
      Make ffserver use parse_options().
      Implement ffserver -version.
      Implement ffplay -L.
      Move show_formats() to cmdutils.c.
      Add -formats option to ffserver.
      Add -formats option to ffplay.
      small consistency fixes for the ffserver manual
      small consistency fixes for the ffplay manual
      Clarify avcodec_encode_audio() vs. PCM audio.
      Update documentation of show_version() to account for latest changes.
      Slightly clarify bitstream filter help output.
      Simplify show_banner() so that it does not require arguments, similar
      Add ASUS V1/V2 decoder codec long names.
      Add a few newlines to 'ffmpeg -h' output to make it more readable.
      Macro suggested by Michael which will be used to disable the
      Make long_names in lavf/lavdev optional depending on CONFIG_SMALL.
      cosmetics: s/Id/id/ in libavformat where Id refers to id Software.
      use program_name to be more consistent with the banner, patch by Stefano Sabatini, stefano dot sabatini-lala at poste dot it
      use unsetenv, fix warning, patch by Stefano Sabatini, stefano dot sabatini-lala at poste dot it
      Factorize options context setting.
      Change opt_sync function behavior to print an error message.
      Make ffmpeg.c:opt_thread use parse_number_or_die().
      Make AVCodec long_names definition conditional depending on CONFIG_SMALL.
      Improve ffmpeg error reporting in case of input/output stream mismatch,
      Make ffmpeg.c use print_error() when it can't read an ffserver stream.
      Cosmetics: s/Id/id/ in libavcodec/roqvideodec.c when Id refers to id Software
      Fix an example command in ffserver-doc.texi.
      Don't show subtitle and grab options within the ffmpeg -h generic options.
      Don't show advanced subtitle options within the ffmpeg -h generic
      Cosmetics: display ffmpeg -h generic advanced options just after
      Add missing SMALL_IF_CONFIG_SMALL macros for beosaudio long_names.
      Add an "How does it work" section in ffserver-doc.texi.
      Cosmetics: s/stats/status/ in ffserver.c.
      Improve an lclenc.c error message in case of unsupported input pixel
      Add a "Status stream" section in ffserver-doc.texi.
      Extend documentation for AVOption.
      Cosmetics: move ffmpeg.c dframes option definition from within the
      Document ffserver no-launch mode.
      Document ffserver daemon mode.
      Document ffserver debug mode.
      Make ffserver daemon chdir()s just after log init. This way a
      Fix typo in libavcodec/opt.h.
      Incorporate ffserver documentation Quick Start chapter into the
      Make ffserver show each launched commandline just before to launch
      Use http_log() rather than printing on stdout to show the launched
      Make ffserver show launched commandlines not only in debug-mode.
      Reindent.
      Remove unnecessary function opt_qdiff(), max_qdiff is set now through
      Update ffserver docs to match r13978, ffserver in debug mode now
      Remove credit to Philip Gladstone contribution from the ffserver
      Clarify/simplify documentation for the default_val field in AVOption.
      Document av_find_opt().
      Print libavformat version in outgoing SDPs.
      Print the program_name rather than always "FFmpeg" when showing the
      Fix a warning in ffserver.c caused by the use of the function
      Fix the logic to access the location of a string to free when setting
      strdup() the default value for rc_eq in AVCodec when setting it in
      Make v4l_read_header() don't free a stream in case of failure when
      Make v4l2_read_header() don't free a stream in case of failure when
      Free in avcodec_close() avctx->rc_eq. Fix a memory leak.
      Make ffmpeg -t X with X<=0 transcode nothing.
      Make configure --help document the --enable-debug option.
      Complete documentation for av_set_string().
      Improve long names for FFM muxer and demuxer:
      Make ffmpeg dump output format in case of codec type mismatch between
      Fix a typo in sdp_write_header(): change "IPV4", which is not a valid
      Increase the maximum number of supported stream mappings in ffmpeg.c.
      Improve docs for the ffmpeg -map option.
      Distinguish the error reporting for the cases of wrong size and wrong
      Distinguish the error reporting for the cases of wrong size and wrong
      Cosmetics: make a v4l2 error message consistent with r14647.
      Implement avutil_version().
      Doxygenate avcodec_version().
      Implement avformat_version().
      Implement avdevice_version().
      Implement avfilter_version().
      Cosmetics: fix weird indent.
      Remove reference to the unsupported master format from ffserver.conf.
      Cosmetics: move at the beginning of the file the av_frac_* functions, avoid
      Add to the ffmpeg docs an usage example showing how to extract images
      Make show_banner() and show_version() print both the compile-time and the
      Implement av_fifo_realloc2().
      Replace invocations of av_fifo_realloc(), which is going to be
      Deprecate av_fifo_realloc(). av_fifo_realloc2() should be used instead.
      Cosmetics: remove a redundant and misplaced doxy.
      Provide a context to opt_frame_rate(), and make it print the wrong
      Simplify the indexing logic in the av_encode() streams mapping code.
      Reword a sentence in a doxy making its interpretation clearer.
      Fix a typo.
      Make X264_init() no longer assign avctx->rc_eq to libx264's context.
      Add to the ffmpeg manual an example showing how to add streams to the
      Implement swscale_version().
      Make show_version()/show_banner() print the libswscale versions if
      Cosmetics: vertical align in print_all_lib_versions().
      Fix typo.
      Globally rename the header inclusion guard names.
      Change headers' guardnames consistently with r15120.
      Implement postproc_version().
      Make show_banner()/show_version() print the libpostproc version
      Make the maximum number of simultaneous HTTP connections handled by
      Make the nb_max_connections and nb_connections int variables unsigned.
      Remove wrong and unnecessary condition. VHOOK filters are not special
      Log an error message in case of invalid number of channels.
      Implement av_nearer_q() and av_find_nearest_q_idx() functions.
      Implement the force_fps option, which disables the automatic selection
      Document the preset files system.
      Remove unused and invalid code from ffmpeg.c.
      Make less verbose a sentence in the preset system documentation.
      Document libavfilter filters options in configure -h.
      Make configure --help return 0.
      Implement CPP condition check support in configure.
      Check requirement for libx264 version in configure using
      Fix inconsistent indent.
      Add preset options documentation strings.
      Provide more context for the error message issued when no preset file
      Make ffmpeg abort if a preset file contains an invalid option or
      Provide more context to the opt_preset() error messages, printing the
      Make more clear the error message for the libx264 version check.
      Remove the ugly and unnecessary require_cpp_condition() function.
      Add checks on the _mingw.h and w32api.h versions.
      Fix argument constness for the av_codec_get_id() and
      Make the library names for the require checks more explicit.
      Remove unnecessary calls to show_help(), make ffmpeg/ffplay print a
      Improve messages printed when there is no input/output file, making
      Improve error message in case of output file with no streams, dumping
      Make ffmpeg print in the error message the exact line of a preset file
      Make ffmpeg print in the error message the exact line of a preset file
      Make the input realtime framerate emulation code use ist->pts for
      Remove AVInputStream frame variable definition and usage, unnecessary
      Make rate_emu a global property rather than to be settable for a
      s/FFserver/ffserver/ in the help message invocation synopsis.
      Rename functions in sdp.c for consistency's sake.
      Replaces not the anymore valid -flags +trell option with -trellis 2.
      Improve documentation for the -pass and -passlog options. Add mention
      Make more informative the error messages regarding the access to the
      Remove unnecessary ret variable.
      Make av_log_missing_feature an internal function, and change its name
      s/#AVClass/AVClass/, consistent with the rest of the doxy.
      Cosmetics: remove empty comments.
      Use a more explicit "codec" rather than "format" as the parameter of
      Make register_avcodec() call avcodec_init().
      Document register_avcodec().
      Remove set_all_opt() and its use in av_set_string2(). Make
      Move documentation of avcodec_register_all() from allcodecs.c to
      Remove ending period in a log message, which was inconsistent with
      Implement the av_set_number2() internal function, which makes possible
      Implement av_set_string3().
      Fix opt_default(), making it exit immediately in case of an invalid
      Replace frame_rate and frame_rate_base fields in VideoData with an
      Replace the width and height fields in VideoData with a struct
      Implement size autodetection if one of the width and height
      Fix weird indent.
      Remove calls to deprecated av_set_string2() with calls to
      Cosmetics, change variable name "o2" to "o" in ffserver_opt_default().
      Put under #if LIBAVCODEC_VERSION_MAJOR < 53 the deprecated functions
      Cosmetics: put a space after each comma, improve consistency and
      Add missing check on the result of a VIDIOCCAPTURE ioctl.
      Remove unnecessary temporary variable "frame_size" from
      Check that the device is a capture device immediately after opening
      Fix documentation for avfilter_parse_graph().
      Choose better names for the variables in create_filter().
      Use more descriptive long_names for Video4Linux input devices.
      Fix documentation for -passlogfile, the argument is not a file name
      Rename variables:
      Put pass_logfilename_prefix check on a single line, improve readability.
      Fix typo: ommitted -> omitted.
      Fix and extend documentation for avfilter_graph_get_filter().
      Change thisIsAVar variable names to this_is_a_var style, more
      Add missing check on the return value of the VIDIOCSWIN ioctl.
      Make the VIDIOCMCAPTURE ioctl error message consistent with the other ones.
      Use avcodec_check_dimensions() for checking the frame size validity.
      Mention in the documentation of avfilter_poll_frame() that it returns
      Fix typos.
      Cosmetics: break line to make it not longer than 80 chars.
      Bump year number.
      Drop the deprecated parse_image_size() and parse_frame_rate() functions
      Move the AVCodecContext options definition to a dedicated file, reduce
      Move the AVFormatContext options definition to a dedicated file,
      Fix capitalization and missing periods issues.
      Implement the avfilter_default_draw_slice() handler and use it in
      Add missing inclusion of libavutil/intreadwrite.h, fix compilation when
      Make avfilter_open() set to NULL the pads and the filters when the
      Reindent after the last commit, while at it also add some spaces for
      Include libavformat/avformat.h before all the other libav* headers.
      Print a warning and fail if the graph description cannot be
      Implement a var for containing the current year number rather than
      Deprecate av_init_random() in favour of av_random_init(), with a more natural
      Replace calls to the deprecated function av_init_random() with
      Drop the deprecated av_init_random() at the next libavutil major bump.
      Move documentation of av_register_all() from allformats.c to
      Make more descriptive the long names for the various variants of H.263.
      Change "H.263 Intel" long name to "Intel H.263".
      Implement avfilter_graph_check_validity().
      Update preset files documentation to match with r16795.
      Remove broken test program.
      Add a new test program for base64, based on that removed in r17024.
      Rename register_avcodec() as avcodec_register() and deprecate the old
      Replace the calls to the deprecated function register_avcodec() with
      Drop the deprecated function register_avcodec() at the next major
      Deprecate av_malloc_format_context() in favor of
      Replace the calls to the deprecated av_alloc_format_context() with
      Make av_base64_encode() do not require the user to provide an
      Cosmetics: consistently prefer "size" over "len"/"length" for the
      Cosmetics: prefer out/in over buf/src for the parameter names of
      Cosmetics: rename the "size" parameter of av_base64_encode() to "in_size".
      Cosmetics: "* out" -> "*out" for consistency with the other
      Document the av_base64_encode/decode functions.
      Fix a typo: lumaSarpen -> lumaSharpen.
      Replace int type with enum PixelFormat for the dstFormat/srcFormat
      Drop the deprecated sws_scale_ordered() at the next major version
      Document coeff and length fields in SwsVector.
      Add a @deprecated notice to swscale_get_ordered().
      Remove pointless comment regarding sws_scale_ordered().
      Move documentation of sws_getCachedContext() from swscale.c to
      Document sws_getConstVec().
      Document sws_scaleVec().
      Fix typo: lenght -> length.
      Document sws_cloneVec().
      Move sws_getGaussianVec() documentation from swscale.c to swscale.h.
      Remove @deprecated notice from AVFrac, it is not deprecated anymore.
      Cosmetics: vertical align "*" for consistency with the other doxyies.
      Document sws_getIdentityVec().
      Document sws_getContext().
      Bump micro version, related to r28491.
      Apply misc fixes for sws_getCachedContext() documentation.
      Fix avfilter_parse_graph() invalid graph description detection.
      Document sws_normalizeVec().
      Implement sws_printVec2() and deprecate sws_printVec().
      Use sws_printVec2() instead of the deprecated sws_printVec().
      Document av_iformat_next() and av_oformat_next() functions.
      Rename register_protocol() to av_register_protocol() and deprecate
      Replace calls/references of the deprecated register_protocol()
      Document av_protocol_next().
      Do not export to the public the first_protocol symbol at the next
      Document av_codec_next().
      Improve grammar: "X next after Y" -> "next X after Y".
      Add const qualifier to the error var, fix warnings.
      Assign const qualifier to the font variable, fix warning.
      Use parentheses around assignment used as truth value, fix warning.
      Use const qualifier for the variable "tbp", fix warning.
      Document sws_scale().
      Rename the "src" parameter in the sws_scale() declaration to
      Fix typo: explixitly -> explicitly.
      Split avutil.h, move all the pixel format definitions to the new
      Implement in AVFilterGraph the scale_sws_opts field, and pass its
      Rename avfilter_destroy_graph() to avfilter_graph_destroy(), for better
      Rename avfilter_parse_graph() to avfilter_graph_parse(), for better
      Add a @file notice to avio.h.
      Mark as "internal but installed" the avio.h file.
      Fix subtitle selection logic, make ffplay select the subtitle stream N
      Simplify the logic for the subtitle stream selection.
      Apply the same logic used for subtitle stream selection to audio and video
      Document ffplay -ast, -vst, and -sst options.

Stephen Davies (1):
      qscale estimate fix, diff extended to 64bit wide. patch by Stephen Davies <steve at daviesfam.org>

Stephen Warren (1):
      Fix MBAFF/PAFF samples.

Steve L'Homme (25):
      init_get_bits() requires size in bits, not bytes
      MSVC-compatible __align8/__align16 declaration
      (f)printf --> av_log conversion
      duration and subitle fix from the patch by Steve Lhomme
      ST16 is undefined for non-GNU compilers.
      Rename time variable to the more descriptive block_time.
      #define fseeko/ftello to fseeko64/ftello64, fixes MinGW compilation.
      warn user about crappy OS patch by (Steve Lhomme <steve.lhomme at free.fr>)
      Add YV12 support, patch by Steve Lhomme % steve P lhomme A free P fr %
      value_num should also be 64bit (1 hunk from the asf seeking patch by DrDivx/Steve Lhomme)
      AVISynth support, patch by Steve Lhomme % slhomme A divxcorp P com %
      add support for some MS DIB formats for avisynth
      Remove empty structures,
      Don't include inttypes.h/stdint.h directly, there is an abstraction in
      reindentation, patch by From: Steve Lhomme, slhomme divxcorp com
      Define fallback for PRIX64, taken from a patch by Steve L'Homme.
      Replace most of the %lld and %llx by their (cleaner) PRI*64 counterparts.
      Properly quote paths on grep invocations for codec lists.
      remove floats in 10 bytes/byte support (no version of matroska used it)
      guess matroska files frame rate
      Ensure codec_id = CODEC_ID_NONE for unknown codecs (don't reuse previous one).
      Add support for block duration.
      add "seek by bytes" support to ffplay
      Add A_AAC codec identifier support in matroska.
      Fix wrong comment.

Steven Johnson (1):
      Implement DELTA_FLI opcodes correctly. Patch by Steven Johnson

Steven M. Schultz (25):
      -c patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      'ffserver=no' default for Darwin patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      yuv4mpeg.c extra space patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      recommit (of patch, as cvslog msg didnt apply cleanly)
      lrintf patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      lrintf patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      av_log patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      av_log() patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      av_log() patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      yuv4mpeg output 'C' tag patch by ("Steven M. Schultz" <sms at 2bsd dot com>)
      gcc 2.95.3 patch by ("Steven M. Schultz" <sms at 2bsd dot com>)
      libavformat/mov.c fails with gcc 2.95.3 patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      libavformat/os_support.c HAVE_LOCALTIME_R patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      FreeBSD pthread patch by ("Steven M. Schultz" <sms at 2BSD dot COM>)
      changes 'make' in a couple places to $(MAKE) patch by ("Steven M. Schultz" <sms 2bsd com>)
      NetBSD 2.0 patch by ("Steven M. Schultz" <sms 2bsd com>)
      liba52/crc.c fails to compile patch by ("Steven M. Schultz": sms, 2bsd com)
      remove -d to make libtool on OSX happy patch by ("Steven M. Schultz": sms, 2BSD COM )
      Need -maltivec in addition to -faltivec for OSX patch by ("Steven M. Schultz": sms ; 2BSD COM)
      '#elseif' needs to be '#elif' patch by ("Steven M. Schultz": sms, 2BSD COM)
      reversing:
      vhoook vs vhook typo, patch by Steven M. Schultz < sms - at - 2BSD - at - COM >
      Darwin install strips all global symbols, which gives undefined symbols
      s/log2/av_log2/
      Add the necessary 'strip' commands to the BSD/OS section of configure because

Sylvain Corré (1):
      MacOS X stack overflow? patch by (Sylvain Corré <sylvain dot corre at wanadoo.fr>)

Tao Huang (1):
      Fix put_*_xy2_arm bug and enable put_pixels16_arm and put_pixels8_y2_arm.

Thierry Foucu (2):
      compute b_per_i_or_p value correctly, patch by Thierry Foucu, tfoucu at gmail dot com
      use correct field number for video according to specs, patch by Thierry Foucu, tfoucu at gmail dot com

Thijs (2):
      make ffmpeg able to send back a RTCP receiver report.
      Make it possible to reuse UDP socket (optional, disabled by default)

Thijs Vermeir (3):
      mpeg4probe patch by (Thijs Vermeir ; thijs vermeir barco com)
      Add support for options
      Add transport method options

Thomas Jarosch (2):
      libav -> libavformat in Makefile patch by (Thomas Jarosch <tomj at simonv dot com>)
      pts encoding fix patch by (Thomas Jarosch <tomj at simonv dot com>)

Thomas Raivio (3):
      mov/mp4/3gp muxer improvements patch by (Thomas Raivio <tjraivio at cc dot hut dot fi>)
      correctly identify AAC audio in mov/mp4/etc. files. patch by ("Thomas Raivio" <tjraivio at cc dot hut dot fi>)
      aac decoding patch by ("Thomas Raivio" <tjraivio at cc dot hut dot fi>)

Thorsten Jordan (7):
      Undefine DEBUG_SEEK.
      set start_time to 0 to enable seeking in mp3-files
      Check context before accessing its field.
      Remove useless buffering of input data, so that avcodec_decode_audio never
      Cosmetics: indentation
      Make Xvid wrapper use threads (with support for newest CVS Xvid only)
      Set pixel aspect ratio for libxvid wrapper.

Tim Allen (1):
      Shared library for libavformat support patch by (Tim Allen <tim at proximity dot com dot au>)

Tim Chick (2):
      enable iwmmxt support, removing mm_support() stub
      iwmmxt ffmpeg code does not implement dct_unquantize_h263_inter_iwmmxt, but

Timo Lindfors (1):
      Add support for files that begin with a dash.

Timofei V. Bondarenko (3):
      me_range patch by ("Timofei V. Bondarenko" <timm at ipi dot ac dot ru>)
      Remove unnecessary guarding &=0x0f around adpcm_ima_compress_sample().
      fix predictor initialization for adpcm-ima encoder not to lose first sample

Tinic Uro (1):
      - made --extra-cflags option work on darwin

Todd Kirby (7):
      padding support in ffmpeg patch by (Todd Kirby <doubleshot at pacbell dot net>)
      Fix description of '-padcolor' option patch by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      UYVY support patch by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      Add UYVY support to libavcodec/raw.c patch by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      enrtries for PIX_FMT_XVMC_MPEG2_MC and PIX_FMT_XVMC_MPEG2_IDCT patch by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      removing the .so file when make clean is run patch by ("Todd.Kirby" <doubleshot at pacbell dot net>)
      Animated GIF looping patch by (Todd Kirby // ffmpeg.php gmail com)

Tom Harper (4):
      outputs bit spent on various encoding functions (motion vectors, overhead, etc)
      Fix linking with H.261 support disabled.
      Fix linking on Intel Mac due to local relocation entries in
      Fix compilation of os_support.c for MinGW with --disable-network

Tom Joseph (1):
      H.263: more descriptive error message

Tomer Barletz (3):
      Extract into its own function the code to compute frame delay.
      Do not force last delay value to its own value.
      Move debug trace to the new compute_frame_delay() function.

Torsten Martinsen (1):
      Support for AMR Interface Format 2 patch by ("Torsten Martinsen" <tma gatehouse dk>)

Toshimitsu Kimura (1):
      Gopher protocol, patch by Toshimitsu Kimura, lovesyao gmail com

Tuukka Toivonen (1):
      disable luma in postprocess filters patch by (Tuukka Toivonen |tuukkat ee.oulu fi)

Uoti Urpala (6):
      dtsdec.c copies one input packet at a time to a (static) buffer of size
      Vorbis specs requires blocksize_1 >= blocksize_0, error if it's false.
      Use COMPILE_MMX2 instead of HAVE_MMX2 to determine whether to compile
      Fix compilation with -no-PIC and without -fomit-frame-pointer (used by
      mpegvideo.h has two function declarations with the 'inline' specifier
      Mark add_png_paeth_prediction_* functions which are only used within this file

Victor Pollex (1):
      Make LOAD4/STORE4 macros more generic.

Vidar Madsen (5):
      New options for intra/inter matrices patch by (Vidar Madsen <vidar at prosalg dot no>)
      -target option patch by (Vidar Madsen <vidar at prosalg dot no>)
      Add Yamaha ADPCM encoding/decoding patch by (Vidar Madsen: vidarino, gmail com)
      Yamaha ADPCM in wav patch by (Vidar Madsen: vidarino, gmail com)
      Yamaha SMAF file format support patch by (Vidar Madsen: vidarino, gmail com)

Ville Skyttä (3):
      Support modular X11 path to rgb.txt patch by (Ville Skyttä - scop from SF-bugtracker)
      Remove obsolete and non-functional support for external LZO.
      Specify in documentation that avcodec_find_(en|de)coder* only searches for

Ville Syrjälä (1):
      Improved NV12/NV21 support.

Vincent Fourmond (2):
      Indent for next commit
      Allow pipe: URL to take fd number as input

Vitor Sessak (711):
      RoQ audio encoder
      RoQ muxer, patch by Vitor, vitor1001 gmail com
      cosmetics and function rename
      Split RoQ decoder to accommodate future encoder patch
      Codebook generator using the ELBG algorithm
      Fix typo
      RoQ video encoder
      add width and height in context and use them
      right copyrights
      indentation
      remove empty write trailer functions
      use the right stride
      Mention unneeded casts and parentheses
      use the right stride
      use properly AVFrame.linesize
      use properly AVFrame.linesize
      Remove variables made unused by the last two patches and test svn access.
      Remove useless parentheses.
      Remove more useless parentheses.
      Make error message more helpful and forbid random()
      Add my last name to copyright headers
      Port some useful comments from newest version of David Hammerton's decoder (from which alac.c is based)
      Remove some duplicated code
      Prepare to remove more duplicated code
      Use proper bytestream functions
      Make "channels" variable mean the number of channels, not the number of
      Remove code duplication. Based on a patch by Matthieu Castet.
      Move var. declaration to allow further clean up
      Remove more code duplication. Based on a patch by Matthieu Castet.
      Cosmetics: alignment
      Cosmetics: indentation
      Test for supported number of channels
      Remove reimplementation of av_log2
      More cosmetics
      Yet more cosmetics
      Minor simplification
      Make deinterlace_16 receive an array as a parameter and not two separated vars
      Another minor simplification
      Minor cosmetics
      Rename function
      Cosmetics: indentation
      More indentation
      Replace two #define's by inline functions
      Rename function
      Comment about unsupported sample size
      Cosmetical: Empty line removal
      Remove reimplementation of get_unary.
      AMV audio decoder
      Documentation update for AMV audio decoder
      spelling
      Fix alignment broke by my last patch
      First libavfilter commit.
      avfilter_formats_unref() should remove the given reference, not the last
      10l. parameter was unused and redundant.
      Need to be careful when dealing with references.
      Make regression tests to pass with libavfilter.
      Oops, my last commit broke compilation...
      Do not hardcode the type of AVFilterFormats.formats in allocation
      Do not hardcode the list of pixel formats and use PIX_FMT_NB instead.
      Factor duplicated loop
      More code factorization
      Coding style changes
      Factor bytewidth determination in its own function
      Split avpicture_fill() in two functions. This will be
      First version of libavfilter Makefile
      Merge of several revisions adding filters
      Add rule for make clean.
      work into the ffmpeg build system
      Merge of several revisions that add more targets to the
      Remove test program and test video output module since ffplay can be used
      Targets addition and removal
      More targets addition
      Link to libraries in dependency order.
      Cosmetics: split long line
      Add movie file video source filter.
      True conditional compilation
      Fix compilation
      Improve conditional compilation of vsrc_movie.c
      Remove everything not yet commited to svn
      My commit in r11942 broke compilation.
      This should not be part of the public API
      Add header for declaration of ff_xxxx functions in imgconvert.c
      Simplify avcodec_default_get_buffer()
      100l
      Initial stab at an API
      Some simple filters for testing
      Use memcpy() to copy an array.
      Make linesize a per-reference property
      Fix a pointer type
      typo
      Even better way to pass size to memcpy.
      Cosmetic: align arguments to memcpys
      Support passing parameters to filters.
      Rework filter initialization sequence.  Now supports passing user parameters,
      cosmetic: more vertical alignment
      Provide some default functions to simplify the implementation of common one
      Add a filter to cut video frames into smaller slices.
      Add a simple video source filter which loads a PPM file and outputs it as
      Add an RGB24 <-> BGR24 conversion filter
      Make avfilter_ref_pic also handle removal of permissions for the caller.
      Missing semicolon typo
      Adjust the default start_frame() and end_frame() implementations to simplify
      Provide a default for request_frame() which does the right thing for simple
      Actually fail when colorspaces don't match
      Add pts to frame references
      cosmetics: vertical alignment
      Allow giving filter instances names, in anticipation of the upcoming
      Add a simple filter graph structure and functions
      Use av_strdup/av_free instead of strdup/free
      Split filter graphs out into their own source & header files.
      Allow code to pass data to filters it creates.
      Fix some potential segfaults.
      Cosmetics: fix indentation
      Fix warnings
      Make input/output pads modifiable on a per-instance basis.
      Make the filter graph just another normal filter.
      Move default entry point implementations to their own file.
      Separate the process of creating links between filters from that of configuring
      Helper functions for adding new pads to filters at runtime
      Add a vertical flip filter
      Let request_frame() indicate success or failure
      Add a filter to modify the framerate.
      Add a filter to overlay one video on top of another, picture-in-picture style.
      Allow creation of filter graphs from a graph description structure which
      Remove test program and test video output module since ffplay can be used
      Remove the data parameter from draw_slice().  It's unneeded and often more
      Use macros for better readability
      Allow filters to set the requirements on permissions for incoming buffers.
      Track the permissions that have been given out to each picture.
      Revert braindead linked list of permissions
      replace memcpy with assignment
      Be more consistent in style of code for falling back to defaults.
      Change registered filter list to a linked list
      Rename avfilter_create to avfilter_open and remove the wrapper function
      Use av_freep instead of av_free
      Add a filter to act as a buffer between filters which may output multiple
      Add a filter to output the same video to two outputs
      Rename "buffer" filter to "fifo"
      Prefix names of filter structs
      Add scale/colorspace conversion filter
      Auto insert scale filter to convert colrospaces when needed
      Image negating filter
      Remove dummy and ppm filters which were only for testing.
      Check for error so that
      Remove no-longer accurate comment
      Make filter auto-insertion code its own function.
      Rewrite colorspace negotiation.
      When inserting a filter, don't lose any information we may already have
      Need to be careful when dealing with references.
      Add horizontal flip filter.
      Rework link property configuration system.
      Indentation
      Change the passthrough filter to a noformat filter.  It's now similar to
      Add an equivalent to the mplayer format filter
      90 degrees counter-clockwise rotation filter
      Rename filter
      Presentation timestamp (PTS) modification filter
      Add movie file video source filter.
      Implement poll_frame() method. Fix ffmpeg.c bug with
      The filter framework as is hangs with filters that
      Add const to silence warnings
      Move setting the pts of the copied frame to a more adequated place.
      Nits: more punctuation and capitalization
      Remove unneded headers
      Coding style changes
      Merge loops
      Add context to av_log()
      Avoid link_dpad() overuse in avfilter_start_frame()
      Remove unused var (dead code leftover?).
      Do not always fail for circular filter chains
      Remove useless ()
      Nit: Alignment
      Cosmetical: indentation
      Remove unneded else
      Indent after last commit
      Simplify
      Simplify
      Missed this file in my conditional compilation patch
      Remove unneeded function.
      Remove header now made useless
      Do not copy more data than needed
      Move default entry point implementations to their own file.
      Separate the process of creating links between filters from that of configuring
      Allow filters to set the requirements on permissions for incoming buffers.
      Make default end_frame() implementation more useful.
      indentation
      Track the permissions that have been given out to each picture.
      Revert braindead linked list of permissions
      Ensure that buffers returned by the default allocator are readable, or else
      Make an invalid use of a default callback implementation return an error.
      Rewrite colorspace negotiation.
      Rework link property configuration system.
      Force linesize to be a multiple of 16
      Force alignment of pic->linesize
      Add last file of main filter framework
      Initial stab at an API
      Make linesize a per-reference property
      Support passing parameters to filters.
      Rework filter initialization sequence.  Now supports passing user parameters,
      Make avfilter_ref_pic also handle removal of permissions for the caller.
      Adjust the default start_frame() and end_frame() implementations to simplify
      Add pts to frame references
      Allow giving filter instances names, in anticipation of the upcoming
      Add a simple filter graph structure and functions
      Split filter graphs out into their own source & header files.
      Allow code to pass data to filters it creates.
      Fix some potential segfaults.
      Make the opaque init parameter non-const.
      Make input/output pads modifiable on a per-instance basis.
      Move default entry point implementations to their own file.
      Separate the process of creating links between filters from that of configuring
      Helper functions for adding new pads to filters at runtime
      Let request_frame() indicate success or failure
      Add many new and improve existing doxy comments.
      Remove the data parameter from draw_slice().  It's unneeded and often more
      spelling fix
      Allow filters to set the requirements on permissions for incoming buffers.
      Track the permissions that have been given out to each picture.
      Revert braindead linked list of permissions
      Rename avfilter_create to avfilter_open and remove the wrapper function
      use units of 1/AV_TIME_BASE for pts rather than milliseconds
      Add another "reuse" permission, to distinguish between the case where a
      Add pixel aspect ratio to AVFilterPicRef structure
      Rewrite colorspace negotiation.
      Fix misc. warnings
      Rework link property configuration system.
      Indentation
      Implement poll_frame() method. Fix ffmpeg.c bug with
      Add const to silence warnings
      Nits: punctuation and capitalization
      Update comment to match code
      Group avfilter_poll_frame() with related funcions and add doxygen comment
      Nit: remove odd whitespace
      Use CodecType for pad type
      Remove author field
      Missed this file in my conditional compilation patch
      Oops
      Remove unneeded function.
      Fix bug I introduced in r11962.
      Make libavfilter define its version like the other libav* libs
      Build system changes for libavfilter addition. Add --enable-avfilter configure option.
      10l, alphabetic ordering
      Print libavfilter version if it is enabled in configure
      Fix libavfilter compilation and make it more in line
      Fix avfilter compilation with --enable-shared and no libswscaler
      Add correct version number to shared libavfilter
      Split filter graphs out into their own source & header files.
      Move destruction of the graph's filters to its own helper function.
      Move simple filter chain loading code over to the filter graph.
      Fix possible segfault
      Ensure that the filter_count member is reset to zero when emptying the graph
      Don't mangle the input strings.  The caller may want to reuse it later.
      Allow code to pass data to filters it creates.
      Make the filter graph just another normal filter.
      Fix memory leak.
      Separate the process of creating links between filters from that of configuring
      Let the filter graph export input and output pads from the filters it contains.
      Fix behavior when exporting output pad with default query_formats()
      Need to copy over colorspace property too
      Let request_frame() indicate success or failure
      Allow creation of filter graphs from a graph description structure which
      Remove the data parameter from draw_slice().  It's unneeded and often more
      Doxygenize some comments
      Rename avfilter_create to avfilter_open and remove the wrapper function
      Prefix names of filter structs
      Minor simplification
      Silence warning caused by my last commit
      Remove useless empty functions
      Handle failure properly
      initialize filter graphs completely even if there is no list of filters
      Fix a small memory leak
      Rewrite colorspace negotiation.
      10l
      Use av_log(ctx, ...) instead of av_log(NULL, ...)
      Make filter graph description parsing code independent of the code to
      Use a filter graph description for creating simple chain graphs so we
      Use separate fake filters for exporting inputs and outputs from filter graphs.
      Rework link property configuration system.
      Implement poll_frame() method. Fix ffmpeg.c bug with
      Nits: more punctuation and capitalization
      Header inclusion redux
      Use CodecType for pad type
      Remove author field
      Remove header now made useless
      Make multiple inclusion guard adhere to convention
      Make comment match code
      Spelling and puctuation
      Fix doxy comment
      Remove ugly forward declaration
      Almost from scratch rewrite of filter parser.
      Remove declaration of non-existing functions
      Split graph_load_from_dest() to have a version independent of avfilter_vf_graph
      Remove code made unused by the two last patches
      Remove more leftovers from the old parser
      Remove avfilter_vf_graph
      Merge avfiltergraphdesc.c in avfiltergraph.c
      Rename variable GraphContext -> AVFilterGraph
      Simplify graph_load_from_desc3()
      Remove usage of AVFilterGraphDesc outside avfiltergraph.c
      Move mess (to be removed) to where it is actually used.
      Move funtion to avoid forward declaration
      Rename functions now static
      Rename uninit() to avfilter_destroy_graph() and make it non-static
      Remove AVFilterGraphDesc struct.
      Add backslash '\' support to the parser
      Replace consume_char() function by *(*buf)++
      I should not have merged the graph parser with the graph
      Remove declaration of nonexisting function
      Remove some unwanted todos
      Cosmetical: alignment
      Remove unused check
      Merge two ifs
      Give a more meaningful instance name to auto-inserted scaler
      Use sizeof(var) instead of sizeof(type)
      Handle av_realloc() failure
      Remove wrapper around av_log2()
      Remove unneeded variables from context
      10l: my last commit broke compilation and introduced warnings
      Remove another unneeded var from context
      cosmetics: Reindent file, patch by Vitor Sessak, vitor1001 gmail com,
      Rewrite unpack_input() completely, patch by Vitor Sessak, vitor1001 gmail com.
      Remove useless memset, patch by Vitor Sessak, vitor1001 gmail com.
      Simplify a list of array allocations with a loop.
      Remove unused defines from ra144.
      Replace some hardcoded swapping with FFSWAP.
      Simplify do_voice().
      The first table of decodetable[] is unrelated to the other. Rename it.
      Cosmetics: remove useless parenthesis
      Use correct type in FFSWAP (thanks to Benoit Fouet for pointing it out)
      Read the bitstream where the data is actually needed instead of doing it in unpack_input()
      Remove unneeded fields from the decoder context
      Do not copy tables to a context var, use them directly
      Declare vars as const when possible
      Do not pass the context to functions that do not use it
      Simplify rotate_block()
      Remove useless buffers
      Convert ra144.h tables from hex to decimal
      Remove a useless temporary buffer
      Simplify ra144_decode_frame()
      Cosmetics: rename variable
      Cosmetics: indentation
      Minor simplification of rotate_block()
      Cosmetics: move function to remove forward declarations
      Use ff_sqrt() function instead of using a table
      Simplify ra144_decode_frame() by unrolling a senseless
      Make etable1 and etable2 tables instead of arrays.
      Only one out of two values of gbuf1 is actually used. Make it half the size.
      Simplify add_wav()
      Simplify irms()
      Cosmetics: indentation
      Remove unneeded var from context
      Make gbuf2 a table instead of an array
      Make wavtable a table instead of an array
      More simplifications of ra144_decode_frame()
      Simplify rms()
      Simplify final()
      Simplify eq()
      Handle the case where we do not have enough input
      Use an intermediate variable for overflow testing
      I should not have merged the graph parser with the graph
      Last commit was missing header file
      Handle av_realloc() failure
      Better error messages
      Avoid searching by instance name when it is not necessary
      Use link_filter() instead avfilter_link() when clearer
      Better instance name for parsed filters
      Simplify
      Simplify consume_string() as Michael suggested
      Cosmetics
      Rename AVFilterInOut.instance to AVFilterInOut.filter
      Rename avfilter_graph_parse_chain() to avfilter_parse_graph()
      Use '[' and ']' for label naming
      10l: Forgot to consider the null byte at the end of the string when alloc'ing
      Update comments to new syntax
      Cosmetics
      Merge declaration and initialization
      Simplify
      More simplifying
      Better error messages
      Make two variables const
      Cosmetics: alignment
      Remove unneeded var
      Proper handling of trailing whitespaces
      More improvements on the handling of whitespaces
      Even more improvements on whitespaces handling
      Cosmetics: if(x != NULL) -> if(x)
      Link filters in the same pass as the parser
      Simplify
      Better error handling
      Cosmetics
      Improve error handling
      Use a user-provided context for av_log()
      Simplify
      Move code from handle_link() to the only place the function is called
      10l: Missed that in one of the last commits
      A semi-colon is also a string end
      The name field of AVFilterInOut can be declared const
      Change the parser logic following Michael's review
      Simplify extract_inout() as suggested by Michael
      Simplify while(pad){pad--; etc} to while(pad--){etc}
      Remove useless assignment
      Replace if(!a){B}else{C} by if(a){C}else{B}
      Move var declaration to where it is needed
      Rename var
      Rename another var
      Factor common code out of if
      Cosmetics
      Cosmetics
      Move code out of if
      More cosmetics
      More cosmetics: make indentation more consistent
      Cosmetics: more indentation consistency
      Pass the inputs and outputs of avfilter_parse_graph() with a AVFilterInOut linked list
      Simplify
      Cosmetics: change function order
      Make consume_whitespace() return a number of bytes to be skiped
      Cosmetics: more function reordering
      Make parse_link_name() return a pointer to the name read
      Fix doxy comment
      Cosmetics: split setting a var and checking for error
      Remove senseless comments
      Simplify
      10l: fix previous commit
      These error messages should print the filter name, not the instance name
      Remove comment redundant with those in graphparser.h
      Factor linked list insertion in its own function
      Use av_mallocz instead of av_malloc to simplify some code
      Simplify
      Remove unnused var
      Cosmetics
      Merge variable initialization and declaration
      Plug some memory leaks
      Cosmetical: move statement
      One more memory leak
      Split openLinks linked list into openInputs and openOutputs
      Remove another useless variable from context
      Indent
      Parameter n of dec functions is always 3. Hardcode it.
      Remove yet another val (gbuf1) from context
      Make tables that fit in one byte (u)int8_t
      Use (u)int16_t instead of short for tables
      Remove gbuf2 from context
      Make add_wav() receive a vector instead of three integers
      Move evaluation of variable a to where it is trivial
      Cosmetics: braces removal
      Cosmetics: merge declaration and initialization
      Remove useless variable
      Minor simplification of dec2()
      Minor simplification of dec1()
      Rename var
      Cosmetics
      More cosmetics
      Yet more cosmetics
      Use (u)int16_t instead of (unsigned) short
      Merge two if's
      Rename context struct to be more consistent with the rest of ffmpeg
      Context vars are not global vars. glob is a bad name for it.
      Rename var: buffer_2 is an adaptive codebook
      Rename var: swapbuf* are LPC related
      Rename var: val -> energy
      Update comment
      Rename var: add read coefficients a decent name
      Fix doxy comments
      Add comment
      Simplify implementation and use of dec2()
      Make lpc coefficients 16 bit wide
      Use ff_acelp_weighted_vector_sum() instead of reimplementing it
      Remove dec1() function
      Remove useless intermediate var
      Remove unused var
      More intermediate vars removal
      Rename variable
      Revert r13499, log:
      Remove ra144.o dependency added in revision r13500, now reverted
      Remove useless intermediate variable
      Rename variable
      Minor simplification
      Add doxy comments to eq() and do_voice()
      Rename variables
      Rename functions
      More variable renaming
      Yet more variable renaming
      The function dec1() do two completly unrelated things. Split the rms calculation out of it.
      Rename dec1() to a more descriptive name
      Rename var
      Rename dec2() function
      Cosmetics
      Replace comments by error messages.
      Remove warning that could never be triggered.
      Split function rms() in two (it does two unrelated things)
      Remove *lpc_refl from the context. Only the value
      Rename var
      Rename function
      Rename vars: s/decsp/out/
      Rename table
      More table renaming
      10l: fix typo
      Yet more table renaming
      Last table renaming
      Remove useless comment
      Misc cosmetics
      Do not write junk in the end of truncated files. FATE
      Avoid a memcpy. Code is also simpler now in my opinion.
      Cosmetics
      Simplify rotate_buffer()
      Move code unrelated to lpc filtering out of lpc_filter()
      Rename var
      Add doxy comment
      Simplify t_sqrt()
      Rename rotate_buffer() to copy_and_dup() and doxy it
      Remove useless define
      Document defines better
      More comments improvement
      Remove useless comment
      More comments removal/fixes
      This should be signed
      Move factor multiplication out of irms()
      rescale_rms() should use unsigned ints
      Remove unneeded check
      Cosmetics: indent
      It makes more sense to return 1 here in case of overflow.
      Simplify
      Simplify
      Use sizeof() intead of hardcoding the values
      Consistency: make the output buffer the first argument of functions
      Merge the >> 4 calculation of cb1_base in the table
      Now these tables fit in 16 bits.
      Remove unused values from gain_{val,exp}_tab
      Remove unused entries from energy_tab
      Remove unused entries from lpc_refl_cb*
      Use lpc_coef[2] instead of lpc_coef and lpc_coef_old
      Cosmetics
      More cosmetics
      Simplify interp()
      Doxy comment
      Add a rounding parameter to ff_acelp_lp_synthesis_filter()
      Use ff_acelp_lp_synthesis_filter() instead of duplicating it
      Update copyright
      Make ff_acelp_lp_synthesis_filter() receives a pointer to the actual filter coefficients and not the pointer minus one
      Reindent the whole file
      Remove useless initialization
      Move function to avoid forward declaration
      Rewrite unpack() using the bitstream reader
      Clean up error message
      Remove unpack() function, read the bitstream as needed
      Remove useless wrapper around ra288_decode_frame()
      Reindent table
      Use a matrix instead of a vector
      Remove unneeded var
      Simplify
      Move vars declaration to inner loop when possible
      Write for loops in a more standard way. In my opinion, they are much more readable now.
      Create a function for float dot product instead of duplicating all over the file
      Cosmetics: remove braces
      Do not reimplement memmove()
      Another use of scalar_product_float()
      Cosmetics: remove braces
      Typo in a comment: s/two avoid/to avoid/
      Slighly faster operation
      Remove phasep context var, it is just phase*5
      Cosmetics: reindent tables
      Simplify: glob->phase is always == 3
      Add av_clipf() function to common.h and use it in ra288.c
      Simplify: use FFMAX
      Do not declare as double a var that only stores a float
      Another av_clipf() usage
      Simplify
      Revert r14218. FFMAX is a macro, so
      Declare function parameters as const
      Simplify
      Simplify
      Another simplification
      Declare parameters of pred() that could be const as such
      Simplify: do not overuse pointer aritmetic
      Set avctx->coded_frame in RoQ encoder. At some point in
      Simplify co(): remove variables that are only used once
      Simplify co(): do not abuse pointer aritmetics
      Cosmetics: alignment
      The codebook generator algorithm involves picking three
      When picking a "high utility centroid" do not pick one
      Do declaration and initialization separately in elbg.c
      My commit at r14340 was not the right solution. For a monochromatic
      Simplify co(), use memcpy/memmove and colmult() when useful.
      Simplify co(): write constant in a more readable way
      Rename co() function to the much more descriptive name of
      Give parameters of do_hybrid_window() more meaningful names
      Add comments to do_hybrid_window()
      Give windowing tables a more descriptive name
      Rename pred() to eval_lpc_coeffs() and add a doxy comment
      Follow FFmpeg convention of returning negative values on error in eval_lpc_coeffs()
      Rename and comment bandwidth broadening tables
      Rename update() function to backward_filter() and add a doxy comment
      Cosmetics: grammar nits
      Give the context structure of ra288 a better name: s/Real288_internal/RA288Context/
      Give the context variables of ra288 a better name: s/glob/ractx/
      Remove an useless comment describing prodsum()
      Remove an useless comment describing colmult()
      Move colmult() function to the beginning of file to group DSP-related functions.
      Cosmetics: remove useless parenthesis
      Rename st{1,2}{a,b} context variables to more meaningful names: {sp,gain}_{rec,hist}
      Remove st1 and st2 variables from context, since they could be just local vars
      Change slightly the meaning of RA288Context.phase. This
      Remove st1 and st2 temporary buffers in backward_filter() and use instead
      Rename two context buffers: pr{1,2} -> {sp,gain}_lpc. Also add a few doxy
      Remove the history buffer from the context. It can easily be evaluated
      Add a few comments pointing to the G.728 specification
      Remove RA288Context.output buffer. This buffer is just RA288Context.sb
      Change the way the input is passed to do_hybrid_filter(). Before, in[0] was
      Cosmetics: remove braces and useless newline
      Cosmetics: do not align with now removed next instruction
      Remove redundant comments
      Declare temporary buffers to be only of the necessary size
      Rename two context vars: s/sb/sp_block/, s/lhist/gain_block/
      Cosmetics: align comments
      Cosmetics: make comment match var name
      Remove the now useless phase variable from context
      Cosmetics: new line
      Cosmetics: rename loop counter vars to i,j
      Cosmetics: rename loop counter vars to i,j
      Do not declare a counter as unsigned when it is not needed
      Rename prodsum() function to convolve()
      Flip (by making buf[i] -> buf[size-i-1]) two buffers: {sp,gain}_block. This
      Cosmetics: alignment
      Check *data_size in decode_frame()
      Check available size before writing in decode_frame()
      Cosmetical simplification in t_sqrt()
      Simplify rms(): merge a few operations in the same statement
      Simplify eval_refl(): return directly when needed instead of storing the return value in a variable
      Simplify do_output_subblock(): Remove a unneeded memcpy
      Cosmetics: reindent
      Simplify range checking in eval_refl(). Before the
      Simplify eval_refl(): s/b == 0/!b/
      Simplify eval_refl(): do not store a buffer value in a var when the buffer could be used directly
      Cosmetics: rename loop counters in eval_refl()
      Apply parts of MXF muxer ok'ed by Michael
      Import more ok'ed chunks of the mxf muxer from the soc tree
      Import more MXF muxer code from the SoC tree
      Remove useless zeroing of a buffer
      Merge declaration and initialization of a few vars in compute_lpc_coefs()
      Remove useless var in compute_lpc_coefs()
      Misc cosmetics in compute_lpc_coefs(): braces removal and empty lines
      Avoid duplicating compute_lpc_coefs() function in both the RA288 and AAC decoders.
      Multiply table by -1. This avoid doing this calculation (that was introduced
      Cosmetics: s/LPC_type/LPC_TYPE/
      Reindent mace.c, its indentation was completly inconsistent with the coding rules.
      Remove useless comments
      Simplify mace_decode_frame()
      Simplify: use two distinct functions to decode MACE3 and MACE6, since the
      Functions mace{3,6}_decode_frame() are just wrappers to Exp1to{3,6}(). This commit
      Simplify: use a for instead of unrolling by hand
      Cosmetics: indent after last commit and remove useless braces
      Remove useless parentheses.
      Test a var for overflow just after it is set
      Remove output pointer from context
      Do not share context variables between channels and do not zero them at
      Use the same 8 bit -> 16 bit conversion as QuickTime.
      Cosmetics: whitespace/linebreaks
      Factorize broken clipping in its own function and document it.
      Use FFMIN() instead of rewriting it.
      Remove commented out code.
      Remove unused table entries.
      Convert tables to signed and decimal.
      Remove cast, now tables are signed.
      Cosmetics: s/short/int16_t/
      Exploit symmetry to reduce size of tables by half.
      Cosmetics: rename vars s/ctx/chd/
      Check output buffer size before decoding.
      Cosmetics: Make the two additional units in size more clear
      Add av_cold attribute to decode_init()
      Simplify if() in copy_and_dup()
      Pass directly variable 'a' to interp instead of evaluating it from num_blocks
      Normalize triplets in gain_val_tab[][] so gain_exp_tab[] can be just an
      Remove useless comment
      Remove ChannelData.lev var and use ChannelData.level instead
      Cosmetics: remove useless parentheses
      Convert two tables to signed decimal
      Convert table to integer.
      Rename function: s/colmult/apply_window/
      Cosmetics: align spec references
      Simplify: use a single history buffer for gain and a single one for
      Cosmetics: line breaks
      Use SAMPLE_FMT_FLT instead of doing the float->int conversion in the
      Add support for RVTR fourCC
      Simplify: move division by constant off the loop
      Cosmetics: move convolve() together with the other DSP functions
      Move constant multiplication out of the loop
      Do instead of divising a float by a constant, multiply by its inverse
      Simplify: avoid duplication backward_filter()
      Factorize duplicate table reading code.
      Use a for() instead of triplicating code.
      Avoid code duplication between mace3_decode_frame() and mace6_decode_frame()
      Cosmetics: rename mace3_decode_frame() to mace_decode_frame()
      Add a new unscaled PAL8 -> RGB converter.
      Do not set src[1] to the palette, it is now in the context
      Fix broken palette8to*.
      Cosmetics: alignment
      Silence GCC warning
      Silence GCC warnings:
      Use ff_dot_productf() in ra288.c
      Cosmetic: Fix typo
      Remove hack added to make "make checkheaders" pass, it is unneeded now
      Silence some ICC warnings. Patch by Vitor Sessak.
      Rename ROQDPCMContext_t to ROQDPCMContext to avoid _t reserved prefix.
      Avoid POSIX reserved _t suffix.
      Use shared sine window instead of defining another one.
      Do not allocate RoqTempData on the stack
      10l: fix compiler warning:
      Document old scaler as deprecated
      Do not reimplement ff_celp_lp_synthesis_filterf().
      Do not read uninitialized buffer, no matter if it will be multiplied by
      Reindent after last commit

Vladimir Dergachev (1):
      This fixes segfaults because of uninitialized s->mbintra_table variable - patch by Vladimir Dergachev <volodya at mindspring.com>

Vladimir Voroshilov (34):
      Add variables to prepare to the AMV decoder patch.
      AMV video decoder.
      Add support for AMV variants of AVI files.
      Fix audio clicks in ADPCM IMA AMV by casting predictor as "signed short"
      Replace "signed short" typecast with "int16_t" as suggested by
      Fix MJPEG decoder for AMV files.
      Decrease returning probe score for WAV demuxer to avoid
      Implement FFMAX3(a,b,c) - maximum over three arguments.
      Fixed-point implementation of ff_cos, ff_log2, ff_exp2.
      Fixed-point LSP and LPC decoding routines for ACELP-based codecs
      Replace hardcoded LP filter order with parameter.
      various filters for ACELP-based codecs
      Fix compilation error.
      Update comment to version, negotiated with Diego, and
      Implement bidirectional (positive offset - left, negative - right)
      adaptive and fixed codebook vector operations for ACELP-based codecs
      Move pitch vector interpolation code to acelp_filters
      gain code, gain pitch and pitch delay decoding for ACELP based codecs
      G.729 decoder main code
      first and second stage LSP codebooks for G.729
      Move from g729.h all definitions which are used only in g729dec.c
      Remove unnecessary header inclusion from g729.h
      Another set of definitions for G.729
      (cosmetics) Align comments consistently.
      (cosmetics) Remove wrong comments.
      (cosmetics) Describe for which tests clipping is required.
      (cosmetics) Remove incorrect comment.
      Add explicit (int) cast to i386 optimized MUL* macros.
      Parity bit calculation routine for G.729
      Another set of approved G.729 chunks (from decoder core)
      gain codebook (first stage), 8k mode for G.729
      Remove not approved parts.
      Another set of approved changes for G.729: includes, formats structure,
      Second-stage gain and Moving Average Predictor codebooks.

Volker Moell (1):
      patch by (Volker Moell <volker-ml at die-moells dot de>)

Víctor Paesa (57):
      ogg has to be at the right of vorbis patch by ("Víctor Paesa" wzrlpy hotmail com)
      add loop_input to AVFormatContext, getting rid of old hack
      Cygwin shared libraries support, patch by Víctor Paesa wzrlpy at arsystel com.
      Add missing EXESUF to executable rules.
      Explain usage of image formats other than GIF.
      Add $(EXESUF) where necessary, fixes regression tests under MinGW.
      convert null vhook to libswscale
      Re-Add support for PNG or PNM as watermark images, which was broken
      port ppm vhook to swscaler
      Add sws_getCachedContext(), which checks if context is valid or reallocs a new one instead.
      Fix vhook build on Cygwin, shared builds only.
      Document how to read DirectShow files,
      Explain lmin, lmax, mblmin, mblmax options.
      Quote names individually, since otherwise the whole list was understood
      port imlib2 vhook to swscaler
      port watermark vhook to swscaler
      simplifies fish vhook by using sws_getCachedContext()
      simplifies null vhook by using sws_getCachedContext()
      Document -newaudio, patch by Victor Paesa, taken from an explanation written by Baptiste Coudurier in -users
      add moving images feature to imlib2 vhook
      Some documentation on the vhook imlib2 moving images feature.
      Some notes about the order of options on the command line.
      Current FFmpeg expects bits/s instead of kilobits/s in the -b option.
      Fill in some missing AVOption descriptions.
      Remove hardcoded dependency of imlib2 on Xlib.
      libswscale is needed in Cygwin vhooks
      --enable-mingw32 --> --target-os=mingw32
      Replace valid use of exit by _exit, patch by Víctor Paesa, wzrlpy arsystel com.
      Add some help for the analyzeduration option.
      Print a list of valid AMR bitrates if a wrong one is used.
      A bit more clear FAQ 1.2
      Add FAQ entry for video joining.
      Wording improvements for the previous commit.
      Remove space before question marks.
      Add a few missing Advanced Options.
      Add myself as Cygwin maintainer.
      On Cygwin, libswscale is needed in VHOOK(SHFLAGS|LIBS) only when
      The watermark vhook cannot handle truncated bitstreams.
      Announce that vhook is deprecated, and its substitute.
      Revert r9952, as per discussion at
      The 'Video Filter API' is better known as 'libavfilter'.
      Leverage MinGW's llrint() in Cygwin compilations.
      Link to libraries in dependency order.
      Alignment
      Fix Cygwin compilation.
      Add libavfilter_version
      Fix issue 68:
      Fix spelling
      The diffutils package is also needed (for the 'cmp' command).
      Bring up to date the Cygwin docs.
      Update Cygwin Ports URL.
      Advice on compiler versions
      Complete the list of external libraries.
      Update again Cygwin Ports URL after my sloppy previous commit.
      Document the same functions, in the same order, that eval.c has.
      Document the ld, st, while statements.
      Explain how to add IPv6 support to Cygwin.

William Juwono (1):
      av_resample_close in avcodec.h patch by ("William Juwono" <william.juwono at sensestream dot com>)

Wolfgang Hesseler (8):
      more #ifdef CONFIG_ENCODERS patch by (Wolfgang Hesseler <qv at multimediaware dot com>) with modifications by me (s/WOLFGANG/CONFIG_ENCODERS/ and some other fixes)
      CONFIG_ENCODERS patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      debug parameter patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      move motion_val & mb_type to AVFrame patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      cygwin fix and dont average interlaced MVs patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      frame stepping patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      motion vector vissualization improvements patch by (Wolfgang Hesseler <qv at multimediaware dot com>)
      export DCT coefficients patch by (Wolfgang Hesseler <wolfgang.hesseler at imk dot fraunhofer dot de>)

Wolfgang Scherer (1):
      Fix SIGSEGV in rv10-null-pointer-500k.rm.

Wolfram Gloger (26):
      Support raw mpegts streams from dvr patch by (wmglo at dent dot med dot uni-muenchen dot de)
      Off-by-two in mpegts.c patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      Fix synthetic next_pts calculation / audio pts patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      user specified start time offset
      Correct audio parameters when grabbing and using "-acodec copy" patch by (Wolfram Gloger <wmglo at dent ot med dot uni-muenchen dot de>)
      Avoid segfault on ffmpeg "buffer flush" in mpeg12.c patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      print start_time patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      update sync_opts for video frame copy patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      fix -pixfmt mono[bw] patch by (Wolfram Gloger <wmglo at dent dot med dot uni-muenchen dot de>)
      10l (a level of indirection too much) patch by (Wolfram Gloger <wmglo dent.med.uni-muenchen de>)
      Writing to freed memory in write_section_data() patch by (Wolfram Gloger <wmglo dent dot med dot uni-muenchen de>)
      Correct number of bits for PTS.
      sync output stream against any input stream (extension of map) patch by (Wolfram Gloger: wmglo, dent med uni-muenchen de)
      Fix url_fsize for large files patch by (Wolfram Gloger: wmglo, dent med uni-muenchen de)
      minor fixes for invalid audio data patch by (Wolfram Gloger: wmglo, dent med uni-muenchen de)
      dvbsub encoder, patch by Wolfram Gloger < wmglo AH dent POIS med POIS uni-muenchen POIS de >
      gcc-2.95 compile fix, patch by Wolfram Gloger <wmglo A dent PIS med PIS uni-muenchen PIS de>
      Fix audio corruption when cutting.
      av_estimate_timings_from_pts() flushes the packet queue but doesn't
      more detailed error message for buffer underflow
      Fix loop condition so it can be run through more than once.
      Give seektest .libs dependency so that it gets updated on libavformat changes.
      fix probe buffer issue. patch by Wolfram Gloger {wmglo at dent.med.uni-muenchen.de}.
      Correct initial timestamp
      set video_enc->sample_aspect_ratio to fix stream copy
      Offer an option (disabled by default) to enable the copy of the frames before

Xiaohui Sun (2):
      Changed the rle encoder a little and made it more universal.
      SGI image decoder ported to the new image API.

Yartrebo (4):
      Snow Slicing patch by (Yartrebo) yartrebo earthlink net
      Snow segfault bug in revision 1.43 patch by (Yartrebo /yartrebo earthlink net/)
      fix overflow in mc_block() patch by (Yartrebo )yartrebo earthlink net
      tighter snow slicing patch by (Yartrebo || yartrebo earthlink net)

Yoav Steinberg (1):
      If we find a VBR tag at the beginning of the file don't attempt to

Yong (1):
      Add fourccs for Wavpak and TTA

Zdenek Kabelac (189):
      * prevent crash in decoder
      * simple fix to avoid memcpy with len < 0
      * made by Juergen - should be useful for Solaris
      * fixes to avoid crash when broken mp3 header is found
      * and finaly avoid deadlock at the end of audio stream
      * dependency for static linking
      * fixed decoding busy loop
      * temporal solution for shared lib compilation
      * continue after error in msmpeg4_decode_mb - helps for some movie samples
      * using static instead of extern
      * eliminating duplicated code from MPV_common_end
      * simplified indexing
      * simplified indexing
      * some modifications to allow gcc to compile same code for -fPIC
      *  Also include time.h for nanosleep()
      * encoding of AC3 with more than 2 channels
      *** empty log message ***
      * support for .au .gif .mov .mp4  by François Revol <revol at free.fr>
      * fixing some warnings
      * more warning cleanups
      * some minor modification by Philip Gladston
      * using pixtype as enum - by Philip Gladstone
      * support for simple_idct config
      * no error message without config.h
      * minor - slightly modified debug message
      * solaris does not support -q
      * add support for liba52
      * using some small char buffer - needed for sprintf
      * using liba52
      * path to ffmpegs config.h
      * fixes problem with -funroll-loops and buggy gcc compiler
      * cleanup
      * by Thomas Sailer:
      * replaced with liba52
      * first shot for the new avcodec API
      * fixed --enable-shared handling
      * avoid conflicts with non-ffmpeg config.h and also it's consistent
      * using ../ for bswap.h
      * restored previous non '..' version
      * restored previous behaviour for this moment
      * oops - these are the midnight commits...
      * minor cleanup
      * fix for -fPIC compilation - compiles with 2.95.2 as well - any
      * added simple test main - see comments about how to
      * fixed contrains and avoid usage of scale index access
      * align the source initialy address
      * somewhat more advanced test suit for dsp code - for now i386 only
      * a little bit extended dsptest also suports dsptestpic target
      * minor update
      * cleanup for put_pixels_mmx
      * always baling even though it seems to produce sometime worse
      * cosmetic
      * baling 8  seems to have the same speed
      * avoids double inclusion of this header
      * compatibilized declaration with its original definition
      * added
      * undefine local defines when they are no longer needed
      * yet a bit more extended
      * for this moment  bswap.h included ahead of HAVE_AV_CONFIG_H check
      * removed TESTCPU_MAIN - not needed for testing
      * more PIC friendly and faster code
      * started to cleanup name clashes for onetime compilation
      * optimized avg_* functions (except xy2)
      * optimized remaing avg_pixels_xy2
      * cosmetic minor change
      * new mmx code - based upon http://aggregate.org/MAGIC
      * code with new PAVGB for MMX only CPU splited into separate file
      * reenabled original xy2 put routine - rounding error is really bad with
      * using unrolled loops seems to help to MMX Celerons
      * added dependency on dsputil_mmx_rnd.h for dsputil test
      * removed MANGLE from macros for setting constants
      * implemented remaing avg_ pixel functions (these are not used offen)
      * reimplemented remaing avg_ pixel functions
      * turned into debug message - it's annoying when watching mjpeg files
      * redundant declaration removed
      * declare av_free
      * clean compiler warning
      * using time.h for Cygwin systems as well
      * ok - let's start with avifile->ffmpeg morphing
      * added support for multiple payloads
      * resync without endless loop
      * using statics
      * skiping unwanted packets when it's possible
      * removed extra_data from AVStream (it's part of Codec)
      * always set extradata_size
      * avoid negative numbers as well
      * DR1 support
      * don't show header errors when eof is reached
      * fix leak with output file closing
      * free extradata
      * allocate slightly more data - so decoders won't be touching memory
      * use larger buffer - the size should be calculcated from the bitstream
      * calculating with the full packet size
      * put there more reasnable 15fps speed (though it's probably meaningless)
      * cleanup statics & const
      * cleanup statics & const
      * cleanup statics & const
      * cleanup statics & const
      * added dsp_mask
      * consts
      * avcodec_init  could be removed I guess...
      * useless commit - ignore
      * using DSPContext - so each codec could use its local (sub)set of CPU extension
      * using DSPContext - so each codec could use its local (sub)set of CPU extension
      * using DSPContext - so each codec could use its local (sub)set of CPU extension
      * using DSPContext - so each codec could use its local (sub)set of CPU extension
      * update
      * quick fix for bit exact setting
      * using macros to shorten init part
      * init table once
      * init for inv_zigzag_direct16 moved to init block
      * fixing some minor const warnings
      * fix double free
      * not checking for get_buffer_callback - it would be user's error
      * minor warning fix
      * const warning fix
      * compilation fix (ARM users please check)
      * fixed prototype
      * oops fixed bad initialization of ff vals.
      * cut&paste fix
      * remaining part of John Ryland's patch
      * two functions to handle allocation of static data more simple
      * fill codec_id in codec_open
      * fix for empty image queue
      * first shot for  generaly usable option parser for codecs
      * avoid usage of next/last buffer after flush
      * code for parsing options
      * headers valid for C++ compilers
      * static,const,compiler warning cleanup
      * still unfinished code for Options
      * still unfinished code for Options
      * opts compilation
      * static,const,compiler warning cleanup
      * check for potentialy problematic field len
      * UINTX -> uintx_t   INTX -> intx_t
      * prototypes at one place
      * int64_t is now the right name
      * update cleanup rules (Steven M. Schultz)
      * sync with main liba52 sources
      * using $(MAKE)
      * fixed IMA WAV decoding (though still might not be fully correct)
      * const arrays
      * using const buffers for reading
      * compile PNG only when ZLIB is available
      * minor optimalization
      * more generic avoption_parse
      * support FF_MM_FORCE
      * oooooops - sorry for this one - wrong logic
      * 'externaly' visible option list begins avoptions_ prefix
      * quick fix to support make -j x
      * moved os_support.h into libavcodec
      * update
      * cosmetic doc fix
      * added more options
      * don't rebuild ffmpeg when libs were not modified
      * fixing cut&paste bug
      * filling bitdepth
      * fixed initialization and some memleak
      * more strict types
      * simplified parameters for parse routines
      * keeping compiler happy and quiet
      * support for AAC audio streams via libfaad
      * support for AAC audio (esds decoding - using extradata)
      * uninitialized/unused vars cleanup
      * simple options to enable faad support (no runtime compile checking is made)
      * quiet missing EOF \n  warning
      * link libfaad when needed
      * passing ImageDescriptions for SVQ3 (Vertical300K.sorenson works)
      * added s263 (mapped to h263 - 3gp files)
      * sync
      * oops - reversed last commit
      * compile fixes by Mitch at Bits.COM
      * 3gp patch by Joca
      * amr format - by Johannes Carlsson
      * removed redundant includes (they were even wrongly used)
      * *FILE needs <stdio.h>
      * sys/types.h for size_t
      * const
      * const warn elimintation
      * const warn elimination
      * more consts
      * missing opts for make dep
      * removed config.h
      * these vars could be either static or extern (compilation fix gcc-3.5)
      * fixing a few of gcc 'clean-code' warnings
      * missing targets cleanup
      * bugfixing call reference
      fix mem leak
      Correctly clean up IntraX8Context upon codec close.

Zhentan Feng (5):
      Factorize common code out of the mxf demuxer.
      Ok-ed parts from patch by zhentan feng.
      Ok-ed parts from patch by zhentan feng.
      Remove useless #ifdef DEBUG (patch by Zhentan Feng).
      Ok-ed parts from patch by zhentan feng.

Zoltán Hidvégi (1):
      MMX code for (put|avg)_h264_chroma_mc8

Zuxy Meng (27):
      use xorps instead of mulps to toggle the sign of a float, as suggested by Software Optimization Guide for AMD64 Processors.
      3DNow! & Extended 3DNow! versions of FFT
      Remove unused and unsupported Cyrix's "Extended MMX",
      Fix ASF format parser's broken UTF-16 string handling
      Use common macro PUT_UTF8 in asf.c
      Set correct aspect ratio (it's sample aspect, not display aspect)
      set bit rate for asf format, patch by Zuxy Meng, zuxy meng gmail com
      use asf stream information to set bitrate
      cosmetics: Fix indentation after last commit, patch by Zuxy Meng.
      Remove trailing / from -I flags, MinGW appears to dislike them.
      Faster SSE FFT/MDCT, patch by Zuxy Meng %zuxy P meng A gmail P com%
      Building both shared and static libraries is now possible on MinGW.
      3DNow! and SSSE3 optimization to QNS DSP functions; use pmulhrw/pmulhrsw instead of pmulhw
      Add --enable-auto-image-base for building mingw32 & cygwin DLLs
      Remove incorrect comment; MMX2 is preferred over 3DNow! on Athlon
      Fix a typo that causes an assertion to always fail.
      Pure, const and malloc attributes to libavutil.
      Reapply r12489: Add pure, const and malloc attributes to proper functions
      Apply "alloc_size" attribute to av_alloc, av_realloc and av_mallocz
      'malloc' attribute isn't supported in old gcc.
      Cosmetic change: remove the whitespace after 'defined'
      Apply 'cold' attribute to init/uninit functions in libavcodec
      Apply 'alloc_size' and 'malloc' attributes to av_mallocz_static()
      Apply 'alloc_size' attribute to ff_realloc_static()
      Add missed call to ff_cavsdsp_init_3dnow() in dsputil_init_mmx()
      add SSE2 version of vp6_filter_diag
      Remove CPUID availability check on AMD64 as it's architectural.

anonymous (8):
      Fixes:
      use dc tables from msmpeg4 instead of duplicating them patch by anonymous
      In that patch:
      - add another way to decode norm6 VLC; modify VLC_NORM6_METH0D define in
      Documentation patch by anonymous
      - a lot less overflows when decoding frames
      - samples from mplayer ftp in the "adv" profile seem to have profile=2,
      - renaming the AC coding set variable names to more meaningful ones

avcoder (7):
      Typo errors patch by (QuickTime | ffmpeg gmail com>
      Change names of codecs provided by external libraries in examples to account
      Avoid void*-arithmetic.
      Remove useless casts.
      Remove useless casts.
      The VC-1/WMV3 decoders depend on ENABLE_MSMPEG4_DECODER.
      av_alloc_format_context has been renamed, update AVFormatContext av_class

borat at subsense.com (1):
      support for sub-only streams; patch by borat subsense com

bubu (1):
      warnings patch by (bubu <bubu at bubu dot net>)

diane_cartman at gmx.de (1):
      13% faster decode_residual (cavlc).

j at v2v.cc (4):
      add libavutil.pc + fix version in pkg-config files patch by (j, v2v cc)
      Add all possible shared libraries to the pkg-config files.
      Move all definitions of external dependencies into configure instead of
      support for libswscale.pc without --enable-swscaler

joca at rixmail.se (1):
      AMR-NB audio support patch by (<joca at rixmail dot se>)

jwestfall at surrealistic.net (1):
      Off by one fix to prevent possible segfault. Patch by jwestfall at surrealistic dot net.

kweiner at spamcop.net (1):
      Add "genre" command line option

lars.taeuber at web.de (1):
      A few lines of text about the range coder, part of a patch by lars punkt taeuber ät web punkt de.

tjcannell at blueyonder.co.uk (1):
      Sony PSP variation of MP4 patch by <tjcannell AT blueyonder DOT co DOT uk>

tjraivio at cc.hut.fi (1):
      quick patch for adding 3g2 support to ffmpeg (muxer and demuxer). No movie fragment support yet patch by (tjraivio cc.hut fi)

tripp (5):
      Clarify -minrate/-maxrate description in the documentation.
      Correct b_quant offset comment in avcodec.h, it is the description for
      Add complete listing of bitstream filters and 2 examples to the documentation.
      Enhance qblur and qcomp documentation.
      Fix and clarify invalid examples for framerate override in the documentation.

wg (1):
      Fix crash in PCM decoder when number of channels is not set.

İsmail Dönmez (4):
      Replace non-existing @fixme doxygen tags with @todo.
      Do not add deprecated -no-cpp-precomp to CFLAGS on Darwin.
      Fix MANGLE macro on Mac OS X.
      Replace hackish CFLAGS 64 bit test by a proper compilation check.

-----------------------------------------------------------------------

-- 
Libav/FFmpeg packaging



More information about the pkg-multimedia-commits mailing list