[Pkg-e-commits] [SCM] Core abstraction layer for enlightenment DR 0.17 annotated tag, debian/1.0.0-1, created. debian/1.0.0-1
Albin Tonnerre
albin.tonnerre at gmail.com
Fri Nov 11 14:53:30 UTC 2011
The annotated tag, debian/1.0.0-1 has been created
at 5291aa082da074411cf3e8ffbd39d81eeae828eb (tag)
tagging dbccddaeb22a826443689dbd4f5e078683f141af (commit)
replaces debian/1.0.0_beta-1
tagged by Albin Tonnerre
on Sun Mar 13 18:12:04 2011 +0100
- Shortlog ------------------------------------------------------------
Debian release 1.0.0-1
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)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=fohM
-----END PGP SIGNATURE-----
Albin Tonnerre (8):
Import new upstream snapshot
Merge commit 'upstream-vcs/1.0.0' into upstream
Merge commit 'upstream/1.0.0' into new-svn-snapshot
Upstream release 1.0.0
debian/control: use architecture wildcards for libecore-fb (Closes: #604654)
Enable c-ares support for dns lookups in ecore-con
Add patches/02_fix_missing_flags_in_ecore_suite.diff:
Debian release 1.0.0-1
barbieri (2):
log domains in lower-case only please.
fix screensaver events, /me spanks raster (r53942)
bdilly (4):
In practice, it was already being treated as an Eina_Bool, so let us
unused_warning--
Document ecore_con_url_verbose_set's behaviour.
Fix typo in ecore AUTHORS
caro (37):
fix doxygen warnings
Ecore_Fb doc finished
typo
improve doc layout
[ecore] fix compilation when pthread is not available
[ecore-file] fix doc
[ecore-file] use function pointers for callbacks
[ecore] remove messages about ecore_time on Windows
fix doc
Killer feature: Windows app can have icons, now.
fix mouse event on Win XP, should be fixed on Win CE
declare alloca()
on Windows, "C:" is not a directory, but a drive.
add native win32 thread support in ecore_thread. Not tested a lot at all.
typo
more fixes, but needs to remove Windows 7 code, now...
remove win7 code and fix other minor problems
[ecore_win32] Better move and resize of windows. Needs a lot of love though
change the size of the buffer
declare variables at the beginning of the block
use '#pragma message' instead of '#warning' when vc++ is used
fix potential problem with Windows 64 bits API.
Add Visual Studio 2008 files
remove useless private libraries in ecore-win32.pc.in
autom4te.cache is deleted at distbuild time, so at
* check if dlopen() is in libc first, then in libdl
order of .la (cosmetic)
add guards in config.h to avoid multiple inclusions
AC_SYS_LARGEFILE defines _FILE_OFFSET_BITS
Windows CE clean up
whitespace--
use configure-time headers checks instead of compiler ones
put again ecore tests in the source tree
add space key support on Windows
use typedef type for ecore_main_loop_select_func_get() API
typo
Use memcpy instead of strndup() as we know the length of the string.
cedric (33):
* ecore: fix build without X.
* ecore: remove thread from the correct list.
* ecore: refcount send and received notify to prevent leak on
* ecore: fix typo.
ecore: return more information on pipe_read error.
* ecore: fix ecore_fb build.
* ecore: turn on and off eina threads support when needed.
* ecore: fix possible race. It also make ecore_thread_run
* AUTHORS: don't forget to blame T_UNIX for his work on xrandr.
* ecore: don't leak pipe.
* ecore: improve and fix ecore_thread destruction.
* ecore: remove ecore_con_url_http_post_send as it expose CURL
* ecore: proper protection of newer constant.
* ecore: remove now useless structure.
* ecore: time to cleanup ecore_con_url time.
* ecore: always check for func_cancel being not NULL before using it.
* ecore: make Ecore_File.h match others EFL headers.
* ecore: don't expose Ecore_Evas internal magic outside.
* ecore: reduce strdup call from ecore_file, but this file need
* ecore: fix mouse wheel for frame buffer.
* ecore: break Ecore_Thread API once and for all to make it consistent.
* efl: beta2 !
* ecore: raise gl_sdl priority over software_sdl.
* ecore: fix event_window for Ecore_SDL.
* ecore: revert comment.
* ecore: small cleanup.
* ecore: improve speed by reusing pipe and allocated structure.
* ecore: fix build on Mac OS X.
* ecore: forgotten file to pass distcheck.
* eina,eet,embryo,evas,ecore,edje,efreet,e_dbus,eeze: beta3 !
* ecore: fix obvious bug. But Ecore_Con with UDP session is
* ecore: SPANK ! SPANK ! SPANK !
* ecore: SO_REUSEADDR in all case.
davemds (4):
ecore_file_download() is BROKEN!!
typo in doxy
fix the return value of the progress callback
Try another file to download, the ubuntu one had some strange lag
devilhorns (1):
Attempting to track a bug in ecore_x_randr_12 where we get a segfault
discomfitor (113):
<discomfitor> ecore_con_url.c: In function 'ecore_con_url_ftp_upload':
2 obscure bugs in one commit:
stop trying to use ssl for local connections
make this more explicit in case compiler is dumber than expected?
add info when generating dh params so people will know wtf is going on when their app hangs for an hour on startup
typedef function pointer
require gnutls 2.10.2 since earlier versions don't work
fix some warnings with --disable-curl
fix potential crash when someone specifies a weird/nonexistent ssl type
hmmm actually this is probably the better way to fix it with maximum laziness
forgot a return statement
fix some random error cases where events would fire too fast and cause a segv
only set openssl verify when needed to avoid some handshake errors
add more gnutls info for dhp generation so you can tell if it ever stops
wnen I said gnutls on my last commit, I really meant openssl. but when I say gnutls on this commit, I mean gnutls. really.
remove unnecessary ifdef
remove dead code
a little bit of reformatting, be explicit with dhp allocation, and add server verification for client certs (apparently I was off my cut and paste game that day)
comment out all psk code since I haven't implemented it yet, fix gnutls servers (how the hell did it break?????)
add a sweet DBG logging function for gnutls because logging info rules
fix incorrect realloc usage
eh I guess I'll clean up a little bit on failure too
only free when exists
because SUCCESS is not an error.
add some safety checks so people can see wtf went wrong when sends fail
reformat using ecrustify because uncrustify dicked it up previously
use safety macros to check mallocs, remove some dead code, optimize read loops by removing them, refactor a lot of code to not be indented halfway across the screen
more refactoring, fix bug that was exposed where a completed ssl read (which returns 0 bytes) causes disconnect
for some reason we weren't setting write on fds when all data had not been sent. this should make ecore_con MUCH snappier
malloc check
++ecore_event_handler_data_{set,get}
fix disconnect conditions for non-ssl reads
fix some cases where refcount was <0 and so things were not being freed
remove erroneous return, add some debug info
add ecore_con_client_connected_get to avoid people scratching their heads and wondering how the hell they're supposed to know if a client has disconnected
delete servers on 0
typo fix
add magic strings for ecore_con
add ecore_con strings
I got overexcited
robustify ssl handshaking by catching extremely rare cases when the session has been deleted between handshake calls usually caused by thousands of concurrent sessions. yes, I'm benchmarking.
fix bug where server disconnect event was not emitted, only set magic to none on clients/servers after final event has been emitted, and only null the data pointer at this point as well
only free clients that aren't waiting on events to free them to prevent a segv
avoid looping infinitely while freeing servers, and don't log while freeing servers in case ecore_con's log domain has already been unregistered
should be bool since we already include eina
unsigned char -> Eina_Bool cleanups; these files are in great need of documentation! if you know the code, please write help!
revert bitfielding of a bool to avoid possible abi break
optimization pass #1: 5% server speed increase
add doxy for ecore_main_fd_handler_prepare_callback_set, add massive optimization for fd handler scaling (only noticeable with 10000+ fd handlers) to only attempt running prep functions for handlers which have them
clean up unused warnings
scaling optimizations round 2: add list for fd handler buffer functions
fix list removal for prep/buf functions to avoid annoying eina warnings
benchmark/test apps for ecore_con server/client as well as fd handlers
fix error handling of plaintext connect, print ERR on connect failure
spruce up error message a bit
remove the last (I think?) O(n) operation from ecore fd handling
fix annoying uninitialized value warning
to fix the bug that morlenxus found, I have crashed my own computer.
fix bug where getopt_parse returns 1 when there are no args
fix bug where getopt_parse returns argc as the first nonopt argument when there are no nonopt arguments
more correctly fix no-option parsing to still parse when no non-arg options are found
I'm tired of you, ecore_getopt. now you get the hammer.
revert this for crankypants
fix annoying printfs that should be ERRs
fix weird bug where fdh in a list could become null
null pointer to avoid partial free in rare cases
explicitly set eina error level, add a printf
use EINA_(TRUE/FALSE) instead of 1/0
switch to EINA_BOOL values for delete_me, fix huge O(n) behavior in _ecore_main_fd_handlers_call; all fd operations now should see an exponential speed increase based on the number of fds that ecore is watching
type this as EB for correctness because I'm feeling pedantic
remove redundant check
only enable extensive ssl logging with compile define set to avoid massive slowdown
copypastecopypaste
formatting cleanups
avoid segv
avoid segv
aha! found the leak and eliminated it!
avoid magic fail on already-freed handler
handle ARES_ENODATA more gracefully
avoid loooooooooooooots of ecore spankies
reorder a little, fix bug where client_del event was never sent if it had send data in its buffer
this error fix is dedicated to Tommy[D] for continuing to prove that the impossible can still be possible on his system
avoid writing <= 0 bytes to clients
handle ARES_ECONNREFUSED
use more explicit error checks
fix usage of ecore_con functions to avoid magic failures: SOMEONE WHO KNOWS THIS CODE PLEASE DOUBLE CHECK FAILURE CASE RETURN VALUES
forgot a null check
validate name better, fix super frustrating bug where connections would inexplicably die if prefixed with http[s]://
finally fix stupid ssl read bug for long transfers! also added a couple more error checks and reordered things to read a little better
add notes to prevent future breaking of ecore_con_ssl
stop trying to rm a dir and rm -rf the dir
From: Mike McCormack <mj.mccormack at samsung.com>
this adds a small helper function for use in finding cases where the epoll fd gets deleted and calls it after each fdh function
some updates to ecore_fd_valid: add dummy closes to track with strace (sup k-s) and switch sleep() to pause()
remove dummy close() calls
complete rewrite of c-ares fd handling: fixes an instance of ticket #637 and also integrates with main loop much more nicely
avoid closing stdout (fd0)
handle ares timeout more gracefully
null variables at start
avoid passing NULL to exec functions
if alloca fails you're screwed anyway, so not much point checking for it
avoid passing NULL to memcpy
avoid null deref
remove redundant line
avoid undefined array access
I think someone meant to set it like this originally but was confused by the double negative.
more descriptive error messages when epoll fails
From: Mike McCormack <mj.mccormack at samsung.com>
From: Mike McCormack <mj.mccormack at samsung.com>
add @file for doxy
unsigned int : 1 -> Eina_Bool : 1
fix segv
fix doc typo
englebass (18):
Check for gnutls >= 2.6
We always require gnutls >= 2.10
Check whether we actually want to alloc data
Right fix for alloc
Declaration at beginning of block
formatting
Check events on right handler
Only remove from existing list
right func to check for dir
print all inotify events
print buf, not file
listen for IN_ATTRIB and IN_CLOSE_WRITE
TODO++
Fix me as author
only emit ECORE_CON_EVENT_SERVER_DATA when data
don't accept 0 return
formatting
only check errno if num < 0
etrunko (3):
Ecore: No need to #define _GNU_SOURCE
Ecore: Decide whether library will be built or not in upper level
Add basic cookie support to Ecore_Con_Url.
glima (1):
Yet another int -> Eina_Bool fix, similar to the one done to
kwo (1):
Fix warning.
lucas (8):
Fix signal accounting
Remove erroneus __UNUSED__ mark.
Add a missing leading underscore to static variables.
Fix some indentation problems in ecore_con_url.
Simplify if clause.
Always use ecore_con_url_ftp_use_epsv_set's parameters.
Simplify the checks in ecore_con_url_http_post_send.
Check for libcurl's return codes when appropriate.
maxerba (2):
updating portuguese and italian translations
updating german translations
mej (2):
Same effect, more portable method, and fixes an obsolete syntax.
Oops. Forgot to remove that.
morlenxus (1):
Fix _ecore_file_path_from_env(), wrong strcpy source.
raster (60):
fix double/triple click
aaagh ecore_fb needs double/triple fix too.
ecore-dfb - double+triple too.
empty dir - remove.
lets set num to 0 shall we? :)
From: Mike McCormack <mj.mccormack at samsung.com>
delay comp setup until show. (and unset on hide)
if pipe handler deletes the pipe... be safe!
and.. be safe! for funcs so they dont go closing things when they
remove changes in case developer dumber than expected?
note that its experiemntal
make sure we have undefed stuff
note -= we may leak display struct here...
patch from mike - fix epoll on fork thing.
from mike:
patch from mike to fix kernel issue with closed fd's+ epoll to re-init
From: Raphael Kubo da Costa <kubo at profusion.mobi>
fix ecore-evas to use stride return for buffers.
ecore_x -> add mapping change event for when key/pointer/modifier
alloca - not strdupa.
make sure it has dynamic hint set.
fix - don't call new event handler added for the SAMe event. defer
no more debian dir in our svn - debian now handles this tehmselves.
see comment.
so... why.. do.. we .. not init.. all? :) how do we expect https and
u know.. eina_list_free.. returns... NULL :)
compacter code.
formatting!
and break api in ecore_input_evas - not used anyway except within
discomfitor... SPANK! REVERT REVERT REVERT! i've been sitting in a
From: Mike McCormack <mj.mccormack at samsung.com>
From: Mike McCormack <mj.mccormack at samsung.com>
From: Mike McCormack <mj.mccormack at samsung.com>
no - dont do http:// - then englebass is right need to do ftp://,
if data is NULL for curl... make postfieldsize 0.
formatting
remove test suite from ecore. gentoo users/ebuild maiontainers are
and we were missing vsync set support in ecore-evas, but had indirect
wtf? why a ! ... ??? that was wrong-ish.
just in case - check if malloc fails.
ok rafael/rui - i've deprecated send and made a get and post call that
hve to revert these commits. they hve created a whole chunk of 64bit
geez! i used the wrong atom for pixmap prop! fix fix fix! it worked
fix possible ptr deref bug waiting to happen if crtc_info is null.
fix oh so many warnings and potential errors - scan report.
oh and i didnt realise - gl enigne can avoid sync as its swap based :)
BAD DISCOMFITOR! BROKE ecore_exe!
warning--
missing IN_MODIFY!
dont call pipe handler if its deleted
dns async lookup fix - if u del svr before dns lookup done
exe none enum++
ecore_con_url_get needs no args.
re-enable progress callbacks.
fix documentation so vtorri can write a correct example :)
,++ in docs
provide enums for progress return.
branch and tag 1.0
aaaah farg. forgot the changelogs. fix.
geeze. get year right! spank
rui (2):
http auth was implemented way back in r48715 (one less item in FIXMEs).
Assorted fixes for building RPMS, probably a good thing that the release source archives can generate the rpms as expected.
sachiel (4):
Dead code bye bye
Use event_win when it exists to match source of events
Revert the changes on ecore_x, seems not needed and break composite
ffs, bring these back in
tasn (1):
Ecore imf: Fixed documentation issues, patch by Jihoo Kim.
-----------------------------------------------------------------------
--
Core abstraction layer for enlightenment DR 0.17
More information about the Pkg-e-commits
mailing list