[libpoe-perl] annotated tag upstream/1.3650 created (now 0b7ed34)
gregor herrmann
gregoa at debian.org
Tue Oct 14 20:42:06 UTC 2014
This is an automated email from the git hooks/post-receive script.
gregoa pushed a change to annotated tag upstream/1.3650
in repository libpoe-perl.
at 0b7ed34 (tag)
tagging c44d4a3558a0c71a5abceb59d52162ce0ac542ed (commit)
replaces upstream/1.3640
tagged by gregor herrmann
on Tue Oct 14 22:22:06 2014 +0200
- Log -----------------------------------------------------------------
Upstream version 1.3650
Adam Kennedy (3):
Make the POE::Filter::Map synopsis a little more readable
Moving a variable inside a scope to save an instruction per event
Whenever you work with relative time, you have to measure "now" absolutely as soon as possible. Moved the time() call from when delay_set sets the actual alarm to the very top, which is much closer to where the caller meant "now" to mean.
Apocalypse (44):
Add doc blurb about pty issues on linux
Tweak the PWR exitval test for MSWin32, turning it into a TODO so we can study what the heck happens
Enable USE_SIGCHLD on perls >= 5.8.1, as they support safe signal handling.
add tracing regression test
add tests and smart has_forked() behavior to simplify fork management
add POD blurb about smart has_forked management
add socketfactor-timeout.t to MANIFEST
squash the AF_INET6 warning for good
finalize the AF_INET6 woes
stylistic tweaks to the recent AF_INET6 patch
shut up warning about constant in void context on old perls
rip out use of base.pm to be more compatible with 5.6.2
be more careful when doing ->blocking(0) on perls older than 5.8.1 as apparently they dont like tied FHs
add doc blurb about tied fh woes on 5.6.2
fix incorrect Time::HiRes pod link
remove t/20_resources/00_base/statistics.pm as we ripped out stat support in 2e5258fd1c98c4ed676befc30a6f0f67855d0e97
add note about NEVER doing ->autoflush, thanks dngor
apparently something in 5.6.2 exports NULL, arg
remove Storable from optional prereqs as its already in required prereq list
add missing POE_DANTIC env check
remove obsolete code and comments pertaining perl-5.005
remove comments pertaining perl-5.004
make code follow what docs say regarding usage of ClientFilter/ClientInputFilter/ClientOutputFilter
add todo to poco-client-tcp about missing (input|output)filter param
Merge branch 'master' of git at github.com:rcaputo/poe
make code follow what docs say regarding usage of ClientFilter/ClientInputFilter/ClientOutputFilter
add todo to poco-client-tcp about missing (input|output)filter param
fix incorrect Time::HiRes pod link
remove t/20_resources/00_base/statistics.pm as we ripped out stat support in 2e5258fd1c98c4ed676befc30a6f0f67855d0e97
add note about NEVER doing ->autoflush, thanks dngor
apparently something in 5.6.2 exports NULL, arg
remove Storable from optional prereqs as its already in required prereq list
add missing POE_DANTIC env check
remove obsolete code and comments pertaining perl-5.005
remove comments pertaining perl-5.004
make code follow what docs say regarding usage of ClientFilter/ClientInputFilter/ClientOutputFilter
add todo to poco-client-tcp about missing (input|output)filter param
make code follow what docs say regarding usage of ClientFilter/ClientInputFilter/ClientOutputFilter
add todo to poco-client-tcp about missing (input|output)filter param
make the ipv6 stuff confess if it isnt available
re-tweak the ipv6 stuff *AGAIN* so it checks Socket.pm for getaddrinfo to be future-compatible with 5.14
add proper checks for ipv6 stuff in Makefile.PL
add TODO about possible improvement to ->kill on MSWin32
Bump Term::Cap prereq to 1.10, thanks Hinrik and Chris Cobucci for tracking it down
Benjamin Smith (35):
Changes to mylib/coverage.perl:
Added some commandline flags to coverage.perl to allow it to run prove(1)
Modified gen-tests to use CORE::exit instead of POSIX::_exit, and to use
Removed the old test upload programs, and their test harness, and references
Removed use lib qw(lib) from tests because we should be use libraries from blib.
Relay bot for commits (with colour removal).
Small smoker script from BinGOs that posts results to the pastebot.
New tests for POE::API::Ctl.
Added 02_api_ctl.t from the previous change to MANIFEST.
Change do back to require as require correctly rethrows any errors produced
4 small fixes to POE::Filter::RecordBlock:
Tests for POE::Resources.
Add subtests to 10_units/05_filters, including:
Stub tests for POE::Wheel::Curses and POE::Wheel::ReadLine.
Adjusted POD to say POE::Loop::Gtk instead of POE::Loop::Event.
Close ticket 20127 by adding an idle watcher that runs _loop_resume_timer if no other
Refactored and extended tests for POE::Resource::FileHandles.
Change defaulting of filters in POE::Wheel::ReadWrite constructor:
The resources test which looks for XS modules being preferentially loaded
Set $SIG{ALRM} back to IGNORE after the kernel exits so that on slow systems with Devel::Cover the test doesn't fail when the alarm fires later.
Change how cover, prove and make are search for to work better on windows (especially where make is called dmake).
Fixed a typo in an error message and gave the $foo variable the name $using_watermarks.
Extended the tests of POE::Kernel's public and private interfaces.
A set of tests that just examine POE::Wheel::ReadWrite (it was already well
Test the POE::Component::Client::TCP constructor, and specifying the filter
Further POE::NFA tests (options, postbacks, callbacks, etc)
Extend POE::Session tests to cover option().
Rewrite of the POE::Wheel::Run tests.
Updating the MANIFEST for the newly added test files.
Skip some POE::Wheel::ReadWrite tests on windows.
My extensions to the POE::Session tests accidentally reverted the signal
Reworked the test for POE::Wheel::Run->shutdown_stdin to work portably.
Skip tests that fail on win32.
(#19971) Added docs for POE::Resources. Adjusted package statements, and docs
* Renamed tests/ to t/
Brian Fraser (1):
Workaround for systems without getprotobyn* (e.g. android)
Casey West (6):
POE::Component::Server::TCP->new() now accepts an 'Args' parameter.
Added create() parameters SessionType and SessionParams. SessionType
Added new() parameters SessionType and SessionParams. SessionType
Resolving Ticket #1597 on rt.cpan.org
The code to validate SessionParams was broken. Even when
From the mailing list:
Chris 'BinGOs' Williams (2):
Fix two typos in PoCo-(Client|Server)-TCP documentation, spotted by Maddingue
Fix regression in SocketFactory with getnameinfo()
Chris Fedde (4):
RT#33718: pravus' patch for mst's error report
#33901: Patch to re-work exit in t/90_regression/agaran-filter-httpd.t
Make the sense of the network test use positive logic
backing out -r2277:2278 sense of network tests restored
Chris Williams (34):
Amended documentation for Args and ClientConnected to match implementation
Made wheel_readline tests skip all on MSWin32 when standard input is not a terminal.
Added pod tests, fixed pod errors and improved pod coverage. Set POE_TEST_POD env var to enable tests.
More pod coverage fixes. Wheel::ReadLine
Changed as per Adam Kennedy's suggestion.
Moved the MSWin32 optional requirements to core requirements. libwin32 installs correctly from CPAN as a dependency on Vanilla/Strawberry now.
MSWin32 hacks to POE::Wheel::Run to use Win32::Process to spawn a separate process. Skipped tk_wheel_run test on MSWin32, which is causing hangs on AS perl
Documented new POE::Wheel::Run MSWin32 behaviour
Rolled back changes to POE::Wheel::Run
Minor documentation fix to amend typo of filter when wheel was meant
Fixed a mistake in the documentation
Borland C++ doesn't specify ENOSYS in Errno.h. Check for existence of ENOSYS and substitute EIO if not found.
Fixed a pod error that was making pod test fail.
Added a regression test for the FollowTail problem on *BSD.
Applied a patch to Filter::HTTPD from Maciej Pijanka (agaran). Created applicable regression test.
Patch from agaran to fix a problem with his original patch to Filter-HTTPD. Adjusted so the tests pass.
Patch from Martin Ferrari to fix tests not honouring the "don't use the network" flag. RT #33276
Added setpgrp() to POE::Wheel::Run to change process group of child process, if supported by the OS.
Added runstate parameter to POE::NFA, similar to POE::Sessions heap. Documented and updated nfa test
Fixed some of the links in ReadWrite POD
Corrected typo in =head1 NAME
Fixed typo in =head1 EVENTS;
Minor pod fixes in POE::NFA
Amended the META.yml generation to include repository meta
Typo in object_states POD, spotted by Jay Hannah
Fixed typo in FollowTail example code for ErrorEvent
RT #47852 'POE::Component::Server::TCP - ClientArgs is passed to the listener, too' applied patch from Michael Fowler, signed off by bingos
RT #47853: document Inline/Package/Object states as applying to client sessions not the listening session, by Michael Folwer
RT #47854: Fix and document the default setting for Port option, by Michael Fowler
RT #47892: added regression test for PoCo-Server-TCP from Michael Fowler.
Make the somni-poco-server-tcp test explicitly bind to the loopback address
Added 'homepage' meta to the META.yml generation code.
Updated agaran-filter-httpd.t to bind server poco specifically to the loopback as per RT #48341
Fixed a mistake in the alias_list() example code in POE::Kernel documentation, spotted by Paul G Webster
David Davis (11):
CALLER_STATE and get_active_event added
CALLER_STATE added
CALLER_STATE modifications
add CALLER_STATE test
added support in HTTPD filter for 'options' requests, and added test for it
remove 100's of PID:0 warnings
data dumper and http resposne aren't needed for the filter test
fixed test skip count
added get_one interface to HTTPD filter
switch to a more portable way of emitting STDERR in a subprocess
add blurb about myself to the contributers list
David Huebner (3):
Added 'immediate' option for changing states immediately. [call rather than post]
Added a bit of documentation about the new immediate option as well as how POE::NFA operates by default.
Moved documentation change to a less illogical spot.
David Steinbrunner (19):
typo fixes
typo fix
typo fixes
typo fix
typo fix
typo fix
typo fix
typo fix
typo fix
typo fix
typo fix
typo fix
typo fixes
typo fix
typo fix
typo fix
typo fix
typo fix
typo fixes
David Webb (2):
Resolve a race condition that caused the last line of rapid input to be delayed. Resolves rt.cpan.org ticket #63242.
Fix read bug in select based polling loop
Hinrik Örn Sigurðsson (24):
Fix 'CleintInput' typo in poco-server-tcp's Pod
Mention that "use POE;" warrants has_forked() after forking
Add test for Win32 socket bugfix in commit 8265022
Allow changing the ReadLine prompt on the fly
Make get() preserve ReadLine's current prompt
Return from sub via "return", not "next"
Reinstate transposition regex which failed on 5.004_05
Add TODO about missing features in transpose-chars
Move the TODO items to the transpose-chars where they belong
Use $stdout instead of *STDOUT, fixes transpose-words
Make transpose-chars work when cursor is at end of line
Return from sub via "return", not "next"
Reinstate transposition regex which failed on 5.004_05
Add TODO about missing features in transpose-chars
Move the TODO items to the transpose-chars where they belong
Use $stdout instead of *STDOUT, fixes transpose-words
Make transpose-chars work when cursor is at end of line
Use Z<> to remove all TODOs from formatted Pod
Small grammar fix in Pod
Fix mangled stdio checks. This one passes all tests.
Document using getsockname to verify a listening socket
Add note about Perl 5.6 issues with '>&' open mode
Document using getsockname to verify a listening socket
Add note about Perl 5.6 issues with '>&' open mode
Jonathan Steinert (20):
Documentation cleanup in PoCo::Client::TCP and PoCo::Server::TCP
Bug squashing attempt, sessions that used postbacks may in the past get garbage collected twice, there was an internal consistency check to this effect which made a confess() call.
In an endeavor to use SocketFactory without the POE::Kernel, discovered that IO::Handle needs to be loaded to work properly. This glitch never showed up before because Kernel.pm has use IO::Handle in it for us. Added for completeness.
Fix make distclean so test directories (not just the tests themselves) are cleaned up. This helps to stop CVS from screaming at you if you've run tests and then want to commit some changes (still have to run make distclean, but at least it works.)
Duplicated test, same test is a couple lines up verbatim.
Minor documentation gap, and some bad grammar.
Arguments to an event should be cleaned up after the event is called, but before we return control back to the Kernel (queued events) or the calling Session (called events).
SIGCHLD bug reported by Merijn Broeren on October 13 2005 on the mailing list. This test evokes the proper results for me (failure on new POE, success on old POE) on Linux 2.6, should be portable but I'm not positive.
Start making this test more portable, someone please suggest something I can use on Windows
We need to watch for the child process ending in this test so we are a good example for the rest of people, and to avoid a mandatory warning that may be added in a near-future revision of POE.
Disable use of $SIG{CH?LD} = "IGNORE". This cannot be done in perl without breaking system() and friends. This fixes a failing tests/90_regression/merijn-sigchld-system.t
Change signal watchers so they keep sessions alive.
Probably should test after removing the signal watcher again.
Mandatory warning for reaped processes to pair with sigchld semantics changing. This loop will block the Kernel shutdown until all processes are reaped. All tests pass.
Welcome to recursive signals, this test makes sure that the sig_handled() flag does not affect outer signals during recursive dispatch.
Forgot to add new test to MANIFEST
This patch fixes the recursive signal testing added earlier. It uses package globals instead of lexicals for the bookkeeping vars in POE/Resource/Signals.pm, this should be examined for a faster way if possible.
Change exceptions test to better determine if POE has shut down properly when the DIE signal is not handled
Better testage for signal recursion causing corruption of the bookkeeping values.
Change exceptions to propagate out and kill the whole process if they are unhandled.
Joseph Huckaby (1):
[rt.cpan.org 87922] Don't crash when ClientPreConnect returns undef.
Larwan Berke (54):
This patch is the beginning of RT#1692, hopefully I got them all, help me test, test, test!
Actually document the new() deprecation as per RT#8469
Patched it to cache the default freezer for performance reasons - brought up by bline,
Andrew A. Chen reported on the mailinglist that inet_aton wasn't imported, thanks!
Updated to check for number of sessions in run() and skip the event loop if sessions == 0
#18665: PoCo::Client::TCP sometimes doesn't send Disconnected callback
#18552: POE::Filter::Stackable does not check it's params
Oops, bad typo!
fixed RT#18646 - finally!
Removed the tests that checked for the old ASSERT death behavior
Patch to clean up POD for Test::Pod and Test::Pod::Coverage
POD tests are now required! This fixes the Loop::Tk* issue too...
added documentation about the ID() method
POD error fixes, thanks Rocco
added missing accessor for TRACE_PROFILE's %profile, stat_getprofile()
fix TRACE_STATISTICS 30s hang because of wrong idle_queue_size count
add TRACE_PROFILE support for transient session profiling
tweak default ports and add die() for MSWin32 on console example
fix Tk warning ( argument idle isn't numeric in multiplication ) exposed by the $^W mess
fix test failure with 90_regression/bingos-followtail.t on MSWin32
shut up prototype mismatch errors
removed yet another warning: Use of uninitialized value in multiplication (*)
removed yet another warning :(
added missing nfa test to MANIFEST
added missing localized nowarnings so we skip 'used only once' on TRACE_DESTROY
skip test because of broken alarm() on MSWin32 with perl5.6
added placeholder test in root t/ directory to get around perl-5.6.2+Test::Harness failures
tweaked the F_GETFL detection routine to be more portable, thanks dngor/p5p for the help!
Backport the Fcntl tweak to the rest of the POE distribution - thanks dngor for pointing this out!
Typo in POD, pointed out by Hinrik - thanks!
POD tweak, thanks crab from irc!
Add sanity check to Server::TCP's InlineStates/PackageStates/ObjectStates constructor, thanks crab at irc!
Use POE::Kernel::ASSERT_USAGE as recommended by rcaputo + POD tweaks
tiny comment tweak
Added RELEASE_TESTING env var checks
Fix a documentation bug, thanks zephg! ( reported by arcanez at irc )
Fix loading of XS loops via POE.pm, thanks TonyC!
Remove any XS modules from @modules, as usual TonyC spotted it :)
Tweak the @modules logic, thanks to TonyC ( it might not be perfect but good enough for now )
Added a lot of loop-loading tests, so we can start work on the "naive" loop load hacks
After a lot of hacking/digging Cygwin with a TERM=dumb does not work with POE::Wheel::ReadLine.
Update the Wheel::ReadLine docs for a better explanation
Fix a doc typo in my previous commit :(
Fix a documentation typo in PoCo-Client-TCP
Fix a doc typo in ReadWrite in get_output_handle()
Automatically clean META.yml and CHANGES files
Back out of my previous commit to auto-clean
Patch POE::Wheel::Run for RT#56417 and add a test+more docs
Add deprecation warning for 'Winsize' to POE::Wheel::Run->new
Add proper exitval handling to POE::Wheel::Run, the previous patch munged it :(
Remove the obnoxious warning about the win32 leak in Wheel::Run
Fix invalid POD links
Tweak the test so it uses RELEASE_TESTING
Add TODO test for double POE::Loop loading
Mark Nunberg (5):
Allow ::Run to discard stdio file descriptors
Merge remote branch 'upstream/master'
STDIN stays open by default
Allow ::Run to discard stdio file descriptors
STDIN stays open by default
Markus Müller (1):
[rt.cpan.org 87635] Fix PoCo::Client::TCP "Connected" event w/watermarks.
Martijn van Beers (8):
Remove the POE::Loop modules from the modules that need to be loaded
Create a instantiate method, so you can override that when you want
remove superfluous regenerating of dynamic states when changing filter.
Edited the docs for new() to reflect the deprecation status
document instantiate and try_alloc, and thus remove them from the
use POE::Test::Loops
update version
add package/object states to POE::NFA
Matt Cashner (83):
added "isa" syntax to the import. allows modules to import the macros and constants of another class
rocco's nits picked
add caller information to the state arguments
drop the spurious warn from the makefile
default the gtk/tk/event tests to not installing all their deps
add a script to generate xml test reports. will eventually upload them automatically but for now just outputs a file
add perl version to the data
move testreport.pl to lib. add failing test reporting to testreport.pl
correct the test directory.
strip file names better. add username, hostname, and gmtime to the report
get the actual user id not the root one.
add new version of test::harness for convenience use in testreport.pl
update manifest for test::harness goodness so it'll ship with the tarball
correct the test::harness::iterator entry
fix use lib to make testreport runnable from the root of the checkout. stop deleting detailed results. fixes failing test reports. add skipped test detail to the report.
adding closing quotes on the skip reason helps with parsing :)
strip the t/ off the file name in case its run from the root of the checkout
add make testreport target to, guess what, generate the test report :)
make the testreport target smarter. checks for existence of poe_report.xml and doesnt rerun the report generator if it exists.
script to upload a test report
add make targets for reportupload
document the test server and the makefile targets
remove Macro/UseBytes from the manifest
revert last patch. my mistake
convert the last remaining macros to subroutines. samples/readline.perl runs flawlessly and all tests pass
add a warning about ithreads bug to Makefile.PL.
remove -l check on files which fixes spurious warning when encountering a filehandle
first stab at a benchmark of various event loops. run with --help option for usage
comment out the linux select timeout workaround and comment why
oops. still need to set a minumum select timeout
add the ability to disable Time::HiRes usage via a constant. document this interface
remove poe_report.xml in make clean/realclean
add POE::Resources as a place holder until some Kernel internals get cleaned up ala resource loading. add POE::API::ResLoader which allows POE::Resources a standard initialization method and the ability to add custom resources to the kernel. clarify the select() call in FileHandles.pm to prevent ambiguitity with the select() in Kerne.pl. Fix sessions so they load their resources into the kernel in the original and documented spot
add lots more needed things to manifest.skip
avoid the docs/ tree
update the manifest with the new files
add documentation and $VERSION to POE::API::ResLoader. also remove unnececessary use of Kernel in same module
add some comments about where resources are and maybe should be loaded from
load signals hash back into the kernel
load the aliases hash into the kernel in the documented spot
put session_ids hash back into the kernel in the documented spot
update the location of things in some comments.
KR_QUEUE was a ref to ref. now is just a ref :)
add a ref to sid_seq into the kernel in the documented spot
add the kr_filenos hash back into the kernel in the documented stop. violate the naming convention a little bit to avoid trampling the preexisting initialization routine
move two initializeii to preload. we need to formalize names here. did this to avoid redefine warnings
add META.yml to make activestate automated ppd building happy again. add Build.PL i used to make the meta file
add gtk initialization error checking. fix gtk test suite to skip the whole suite if gtk initialization fails.
remove spurious warn i stupidly left in
change init_check to simply not initialize gtk a second+ time. not sure why init() is passing the second time but init_check() doesnt but whatever.
style nit pick. replace some random hard \t indents with the majority 2 space indents
add error messages to every build_error call
(!!) Add core prerequisite of Test::More. Add the beginnings of tests of Filter::HTTPD
keep NEEDS parser from looking into the bundled AUtoInstall.pm for strange non-needed needs
add the new test to the manifest
add HEAD tests. add PUT and POST tests which are currently skipped due to known bugs
turn on the simple post tests after working around the HTTP::Request::Common bug that was causing them to fail
remove skip block around no longer skipped tests. remove a warning
rip out the hack workaround. specify the protocol using the http::request method
add a better error condition in the case of a POST request in a 0.9 transaction.
remove spurious warn
add tests for large multipart form data posts
add docs to warn about possible brokenness with libwww
add kr_run_warning into the kernel so it is usable outside of Kernel.pm. all tests pass.
up poe version for development
expose kr_active_event
add in the new POE::Resource::Controls and POE::API::Ctl modules. yay ra
update my email address
remove 'use warnings' everywhere to keep stuff from breaking on older perls
fix some pod issues that pod2man was complaining about
unfortunately there are really two patches in here. first, remove the "no warnings" calls that were breaking poe on perls older than 5.6. replace them with $^W invocations as suggested by perllexwarn. second, fix the TRACE_FILE magic that was causing older perls to lose their mind. the comments here explain this in detail so i wont bother explaining it here. its kind of magic but works well. i have tested all this on 5.004 and 5.8.3 and all is well there (at least in relation to t [...]
yet more "older perls" support fixes. only run the tests that require qr// on perls that support qr//. tested on 5.004 and 5.8.3
merge in lotr's try_alloc patch. closes rt #6045
mod the test suite so tests pass despite evil nasty garbage collection bug in perls older than 5.6. the workarounds are commented with nice blocks explaining the problem. the only real solution is to get a perl without the bug. sorry folks. there's only so much i can do.
(!!) complete XyzState to XyzEvent transition. This removes all errors and warnings caused by use of the old State parameters in Wheels. all tests pass in 5.004_005, 5.005_004, 5.6.2, and 5.8.3
find the tests in a different (less shell dependent) way so that tools like testreport work as expected
add ability to opt out of network tests
comment out the testreport functionality
change a raft full of ok($string eq "string") tests to is($string, "string") for easier test debugging
roll exception handling (ala POE::Exceptions) into the core. this patch creates a new terminal signal called DIE. if an exception occurs, this signal is sent to the session that caused the exception. by the nature of signals, if the child fails to handle the excpetion, the signal propogates up the session tree until either someone handles the signal or the poe environment collapses. two small tweaks were necessary outside of _dispatch_event. session_alloc needed to cache the value o [...]
add documentation for the new exception handling code
tests for new exception handling code
update manifest for new tests. apparently the manifest got rearranged internally so the diff looks way larger than it really is. add run_network_tests file to MANIFEST.SKIP.
Matt Sickler (19):
Patched Filter::Map to close rt18551
Added test to check PoCo-Server-TCP's Filter arg mechanism
Code audit to PoCo-Server-TCP
Edited mistake in POE::Filter::Map introduced in r1948
Added paramter checking to Filter::Map and Filter::Grep
Fixed array slice syntax error in POE::Filter::clone()
Updated the filter mechanism again, hopefully I have it right this time. This also addresses rt19106.
Updated the filter mechanism. This addresses rt19106.
Fixed Stackable.pm to make Filter optional again.
Added callbacks exposing High/Low-Mark in the ReadWrite wheel of PoCo-Client-TCP and PoCo-Server-TCP
Removed depreciated code from POE::Session::new()
Added patches to PoCo-Server-TCP and PoCo-Client-TCP such that new() returns the spawned session's ID. Included documentation.
Updated PoCo::Server::TCP and PoCo::Client::TCP docs to match new Filter argument style
Fixed POE::Kernel::TRACE_STATISTICS documentation bug reported by joel.
Finally removed all traces of new() from the POE::Session docs.
Added handle accessors to Wheel::ReadWrite
Documentation edits.
Removed my old, godawful slow test.
Removed the test from MANIFEST.
Mike (1):
Add a test case for the MSWin32 stat() bug fixed a few commits ago.
Nicko Dehaine (1):
[rt.cpan.org 87721] Clear $! in Driver::SysRW::flush().
Nigel Kukard (1):
Added support for Stopped callback in TCP server
Perl Whore (1):
Resolve a test failure on CentOS, which appears to be a race condition in the test itself. Resolves rt.cpan.org #61727.
Philip Gwyn (47):
Mid-sized documenation update
Middle-sized documentation review
All L<module> to L<module|module> to prevent pod2html outputing 'the module manapge'
Small POE::Session doco patch from pravus.
Mid-sized review of POE::Session documentation
Win32 fixes :
Added rt47966-sigchld.t which tortures the signal-pipe with CHLD, for testing
Wait for longer for all children to exit
Small comment tweak re: top/bottom handlers
Roll back previous patch. We need to call ->close_slave() on the _read
Added test to make sure we get the 3 "completion" events from
Hiding Event's exceptions is no longer necessary
Fixes to POE::Filter::HTTPD:
POE::Wheel::Run : Mask signals before fork(). Unmask when safe again.
Document the fact that the signal pipe isn't used on Windows.
This patch sucks less. _warn not defined at that point. Teach me to commit
Prevent the signal pipe from being used during signal finalization
Reset the idle queue size in Kernel->stop()
Document the ordering of ErrorEvent, StdoutEvent CloseEvent and CHLD
Unit test for r2687
Add more to what I've done.
Added a guard to the signal pipe
delete $ENV{POE_ASSERT_USAGE} so that 11_assert_usage.t and 13_assert_data.t
Delete $ENV{POE_ASSERT_USAGE} so that 12_assert_retval.t always works.
[rt.cpan.org 79886] Fix SIGDIE handler save/restore for Perl 5.8.8.
Added POE::Resource::Clock which exports monotime() and walltime()
Added POE::Resource::Clock
Calculate window pixel size from columns, lines.
Added leolo-alarm-adjust.t to test alarm_adjust() and delay_adjust()
Fixed my regression test
Fixed POE::Kernel->delay_adjust()
Fixed previous patch for delay_adjust( $id, 0 ), which means the delay happens now.
Added Streaming and MaxContent to POE::Filter::HTTPD
Make sure Content-Length is a number
Added POE::Filter::HTTPD->get_pending
POE::Filter::HTTPD Streaming mode no longer requires switching filters
Merge branch 'master' of ssh://git.code.sf.net/p/poe/poe
Added MaxBuffer and MaxLength parameters to POE::Filter::Block
Added MaxLength and MaxBuffer to POE::Filter::Line
Added DEBUG constant
Added MaxBuffer to POE::Filter::HTTPD
Added FIRST_UNUSED
Added MaxBuffer support
Added POE::Filter::BlockRecord::FIRST_UNUSED
POE::Filter::Block->new complains about unknown params
POE::Filter::HTTPD will use Email::MIME::RFC2047::Encoder to convert
Fixed the doco
Rocco Caputo (2327):
Initial revision
Initial revision
First version. Tasking works; tested with forkbomb and about 1500 sessions.
First version. Tasking checks out, and nothing leaks. Next, to do the
First code.
Initial revision
Fixed a typo. Whee!
Finished selects support, and added the beginnings of signal support.
Initial revision
Finished signals implementation.
Shut down sessions properly when receiving a terminal signal.
Debugged signals.
Tweaked to better test signals.
Minor tweaks & FreeBSD/Linux/OS2 portable.
Initial revision
Added Wheel/Driver/Filter support (and state() for self-modifying sessions)
Removed some debugging code; moved functionality to Session.pm
Added POD documentation.
Overhauling the interface.
Testing cperl-mode vs. RCS things.
Initial revision
Initial revision
Initial revision
Initial revision
Documentation is hard!
Initial revision
Tweaking to get Makefile.PL working correctly.
Initial revision
Added FollowTail and tests
Update docs (added alarm() docs)
updated for POE.pm auto-import
Initial revision
explained more about what the hell this is
updated docs
updated to reflect new version and features (d'oh)
added Curator.pm and curator.perl
updated to reflect Curator.pm and test
make _parent/_child/_stop events (and table cleanup) more consistent with
Clean up signals, and make sure they pass all existing tests.
made signals work better, and updated documentation
add signal handler semantics (return value)
update signal handlers to match new usage
explained UberSys a little
update for 0.02, state new directions, etc.
update docs for new signal processing
progress!
Documented special states' parameters.
allow subsecond alarms if Time::HiRes is present
added and documented delay()
adapted to use Kernel::delay() in a meaningful way
documentation patch for session_free
Initial revision
updated docs for tests/proxy.perl
fixed a "blessing" bleed... event handlers could return blessed things
removed some things that won't make it into 0.02 :(
updated... :)
reinvented it all again
fix a line of spaces, so =head1 DESCRIPTION would be recognized
pointless...
added Session::new($kernel, $object, \@methods) syntax
Initial revision
added docs for tests/objsessions.perl
updated the object syntax docs to be even more confounding
updated to include Object syntax
fixed _default handling
inlined _check_session_resources
updated
added objsessions.perl to the tests
d'oh! d'oh d'oh d'oh!
incremented the version
checks objects to see if they can handle methods
add a SIGZOMBIE handler
added aliases
Initial revision
added names.perl test
documented 0.04 changes
updated the version to 0.04
update for new signals
signal() and call() added
tests call()
documented signal() and call()
dunno why
updated signal handling
added credits; updated availability
updated the error handlers to not report error 0
moved post() in 'increment' to test that call() doesn't GC
updated garbage collection, documented call()
tweaked to check for delayed-GC on new sessions
delayed _start garbage collection
tweak error handling
tweak signal handlers for possible hp-ux bugfix
enhanced logging.
optimize shutdown
dispatch SIGPIPE directly to the offending session
skip fake and nonexistent signals... was causing trouble
Initial revision
Initial revision
Initial revision
changed the comments. big deal.
commented a bit, so i could understand it
fixed to display references better
rewrote to use POE to exercise Filter::Reference::Put
remove dependence on length of frozen version
put() no longer concatenates @_ before sending it to the driver.
whee
updated documentation for new contributions
updated version; documented contributions
updated put() documentation
added Reference.pm and associated tests
accidentally broke the flush event... now it's fixed
Initial revision
preliminary check-in before replacing IO::Select
localize $SIG{'__DIE__'}
localize $SIG{'__DIE__'} in evals
remove local $SIG{'__DIE__'} damnit
added the infinite loop test
Initial revision
added inline markup; whee!
unbroke block-quoted text's leading whitespace
Initial revision
fixed the copyright
fixed to match documentation; removed docs
removed documentation
removed documentation
bumped up the version, and removed the documentation
Initial revision
changed to not use IO::*
made debugging prints switchable on/off instead of commented
added changes for 0.05_03
rewrote signal handlers.
update to flush as much as possible all at once.
i forget :(
bump the subversion
updated to be more like a-mused's tests
allowed command line parameters to override what's sent.
added POE.pod, Filter::Stream, and some new tests
Storable isn't portable across different endianisms; made FreezeThaw
Initial revision
updated for 0.05_03
propagate error messages from modules to user
made Storable preferred again :)
forward module exports to the package using POE directly
updated for new parameters
revise for new Kernel semantics
revised for new _parent/_child semantics
update _parent and _child parameters
Initial revision
many updates for 0.06
updated for 0.06
bump the version to 0.05_04
update for 0.06
Initial revision
commented the master socketfactory, and made fork failure friendly
updated to use POE::Session instead of POE
update to use POE::Session instead of POE
depreciate HEAP as a parameter to kernel methods
Initial revision
updated for 0.05_52 and win32 support
bumped the version to 0.05_52
added tutorial-chat.perl to the tests
updated to manage the output queue as a queue instead of a stream
updated to work with blocks instead of streams, internally... it's
updated to work with blocks instead of streams... for compatibility
updated for future datagram support
updated signal handler error messages; removed SEGV signal handler;
update for future datagram support
added Win32 non-blocking support
updated
removed spurious HTTP response
fixed a bad package at the very end, causing the program to fail to
added a new message... whee!
fix indenting; unlink unix sockets when done
update comments a little; add more friendliness
fixed win32 support
added section numbering for .pod output
added the Crimson Scope Hack(tm) :)
add a bunch more comments
updated for 0.06
document the one-way protocol better
added preliminary code for Win32 support... needs work
modified log writers to test write-only ReadWrite wheels
changed event names around, slightly... eliminated bareword warnings
fixed indenting
fixed FailureState to generate an actual handled event
remove AUTOLOAD and post methods; _debug becomes option()
added the Crimson Scope Hack(tm) :)
update Changes for 0.06
updated version for 0.06
make InputState optional (write-only ReadWrite)
remove some problematic barewords
ignore SIGWINCH until we can get safe signals
updated for 0.06
update to test that => does not "stringify" objects
sorted; added some new tests
dorked with indenting. whee!
Added session_create for Randal.
Changed internal 'debug_flags' field to 'options'.
Add the event() method.
Added event() method.
added event() method
Allow packed addresses as parameters.
bump the version to 0.06_03
Added some new redirections to test things.
Initial revision
added changes for 0.06_03
bump the version number
add changes for 0.06_04
croak on old constructor syntax
removed import for Time::HiRes::sleep()
bump the version to 0.06_05
fixed a condition where error events sometimes would not fire
document prerequisite autoinstallation
add prerequisites
document changes
add defaults
document SocketFactory defaults
update to use new SocketFactory semantics
update to use new SocketFactory semantics
use new SocketFactory semantics
typo fix
look for RemoteAddress instead of ListenQueue to determine connecting
added STATE
bumped the version to 0.06_06
documented STATE
added AUTHOR and ABSTRACT
made all the version numbers 0
Initial revision
included STATE constant (whoops).
added DEB_STRICT debugging constant.
bump the version
added create.perl
0.06_07
start revision history for 0.06_08
s/tests/samples/
Initial revision
moved out-out to docs, added extension
Backed out Leolo's patches. A new set of patches plus a test are
Added shebang and made executable.
Added things to research.
Changes: Documented these changes.
Updated manifest to include new files
POE::Runtime - Made this a "package" object that needs to be
Added POE-HINTS target to the docs makefile
Added Philip Gwyn's serialization patch to Filter::Reference.
Applied Philip Gwyn's get_pending patch to allow filters to be
Fixed t/test.t to point to samples/
Updated samples/*sessions.perl to test $kernel->state(...) and
Bump the version to 0.07
Revised all the manpages. Split the manpages into separate PODs, and
Dropped version back to 0.06_09 from 0.07
Tweaked documentation for fakelogin.perl
Tweaked version number to be a number number
I always forget the date in the Changes file
POE-TODO partial update
Major revision to POE-TODO, for HTML inclusion on the web site.
Document exported constants in Kernel and Session
Proofed and revised the manpages
Bumped the version to 0.07
Fixed SocketFactory connect error misfire
Updated Filter::HTTPD.pm to advertise protocol version 1.0
Fixed the 32KB framing error in Reference.pm again
Bumped the version number to 0.07_01.
fixed typo in Changes
Made BindAddress optional for Unix client sockets in SocketFactory
Updated documentation
Fixed a minor typo in SocketFactory
Made _enqueue_state use binary insert, if available.
Added date to Changes file
Shut down sockets in thrash.perl
Added pong.perl
Renamed pong.perl to poing.perl.
Tweak poing.perl some more
Remove stupid closure tricks
Add samples/poing.perl to manifest
Add missing classes to CVS tree and MANIFEST
Bump revision; update Changes
Apply patch from John Labovitz: SysRW read block size
Added time limit for samples/poing.perl
Reduced small/large queue limit. Removed hard requirements for optional modules.
Changed IO::Socket::INET uses to IO::Socket in samples. Cleaned up poing.perl a little.
Added &POE::Kernel::fork() and changed preforkedserver.perl to use it
Defer SIGZOMBIE until after a kinder, gentler SIGIDLE has failed to wake things up.
Reordered the lookups in alias_resolve
Fixed Changes to reflect alias_resolve changes
Fixed POE::Kernel::fork
Added Session IDs, finally
Bumped version to 0.0802
Added Kevin Lenzo's support for ReadWrite using two handles
Bumped version to 0.08_03 for later
Made SocketFactory states accept coderefs
Fix SocketFactory states coderef acceptance
Fix SocketFactory states coderef acceptance
Fix up documentation and comments in preparation for 0.0803 release
Fix Changes to have today's date
Fixed ReadWrite and FollowTail wheels
Added getsockname method to SocketFactory
0.08_05
SocketFactory UDP (and maybe other protocol) support added
Minor comment changes.
Added timestamping to out-out.html; featuritis in poing.perl
More poing.perl fixes.
Several changes, including 0.08_06 version; fixing the samples
Recommit from previous broken one.
Document Changes better
Fix update breakage in the sample programs
Preliminary work on compressed Filter::Reference, checked in to transfer to FreeBSD box for testing
Refserver corrections to test compressed references.
Document compression mode in Changes
POE::Session::ID was commented out and never brought back in.
Moved Reference's inclusion of Zlib to runtime; replaced length() on Reference packets (was dropped in previous patch)
Bump the version and fix the Changes.
Documentation and version bump for the 0.09 release
Split Filter into InputFilter and OutputFilter
Fix another Abigail bug! Yay!
Add Win32 code back
Artur's high/low water mark patch; needs documentation and testcase
Drivers now return pending-write octets instead of objects
Changed SysRW to be a list reference internally, and documented Artur's watermarks
Documented changes.
Documented changes.
Bumped the version to 0.09_04 and documented the signals.perl change.
Updated Changes to reflect the possible 0.0904 release date
Fix queue_peek_alarms
Added driver statistics to Wheel::ReadWrite
Documented changes.
Prepare for 0.0905
Added watermarks.perl to manifest
Preparing for 0.0906
Fixed a reference leak in Wheel::SocketFactory; similar to the ones in Wheel::ReadWrite
New features and fixes before 0.0908
Fixed tests' lib paths
Added P::C::Server::TCP
Add strict and POE::Wheel::SocketFactory to the TCP component
Re-unbreak closure breakage in wheels' states
keep track of event generators' source file/line
resolve session id to reference in alias_resolve
oops! fixed Changes
bumped the version number and fixed a typo
Bump version and tweak MANIFEST for 0.10
Fix the README WRT Win32 support
macros, optimizations, comments. a lot of changes.
Add Preprocessor
optimized regexp optimization in the preprocessor
optimized regexp optimization in the preprocessor
optimized regexp optimization in the preprocessor
Work around a filter bug
removed poing.perl from the samples; it's all growed up now *sniff*
added two new test suites
fix test 4 to use pipe instead of socketpair
tone down the selects testing a bit
fixed wrong variables near alias_resolve macro invocations; tweaked changes
Major source overhaul.
optimize reference compares
documentation revisions for 0.1001
fixed 0.1001 release date
committing the new test
split queue and bugfix
fixed macros test
fixed Changes date
Many changes, which have been described in Changes
internally tested Tk support; removed debugging info
Oops! Bump the version to 0.1003!
fixed Changes merge
fix cross merge
Big documentation rewrite
Initial Event support; new Tk test; more manpage revisions
test initial tk test's test for tk initialization
better tk testing... moving to freebsd machine to run it
tweak tk test
better tk test; need to move to os/2 for plain testing now
back to freebsd
got tk shutting down ok; time to make an event test on the os/2 side
initial Event support is in
bump the version number to 0.1005
tweaked events tests
move it over to freebsd for make dist
final tweaks and 0.1005 released
lots of testing changes and additions
tweaked postbacks to work on older perl versions
many signal changes
minor tweaks
have Event feed into the child reap loop, too; add a test for it
0.1007 is in the can
move to freebsd for X testing
many postback tweaks
garbage collection fixes
version 0.1008 final
windows support tweaks and a UDP test
minor fixes for testing and coverage reporting
added Filter::Block and Dieter's FollowTail patches; finishing touches for 0.1009
followtail and block tests and fixes
added Filter::Block to the documentation rewrite queue
final documentation tweaks for 0.1009
Removed $kernel->fork(); added new tests to atone
test tweaks and followtail fix
tweaked the version number
cosmetic tweak
Filter::Reference test and ascilliary files
cosmetic tweak
Fix garbage collection.
0.1011 release
fixed SocketFactory's getpeername problem?
tweak changes and bump version to 0.11
final Changes tweak
final README tweak
fix Filter::HTTP to use network newlines
Doc tweak; fixed Filter::HTTPD to send better headers
mainly tk fixes
comment tweaks; bump version to 0.1101
gutted and rewrote Filter::Line's innards for extra options
autodetect newlines in Filter::Line
newline detection in Filter::Line; fixed mailer breakage in filterchange; added a filter-change test; tweaked some tests
added test 19 and cleaned up some temp files
tweaked timing in the signals tests
tweak signals tests' timeouts
fixed t/19_filterchange.t when using socketpair
finished README tweaks for 0.1103
small last minute changes
minor preprocessor tweak to make win32 joyous again
TestPipe.pm; 20_accept.t; better select management
select badness
undo fileno; fix TestPipe to not use dup
final doc tweaks for 0.1104 release
doc patch and 19_filterchange.t Storable tweak
several win32 test skips
pre-commit to test on freebsd
fixed a typo in tk.t
updated documentation
wheel IDs
bump up the version
readme tweak
fix SIGCHLD tests
fix SIGINT in the SIGCHLD tests, but not all the way... move back to OS/2 for better editor
fix SIGCHLD tests
initial Gtk support
gtk debugged... moving to os/2 for testing
tweak README
added compile-time code inclusion
minor README tweaks and a missing inclusion
coverage and conditionals completed
test and coverage for README
Optimized POE::Preprocessor a bit, and added POE::NFA
more test results to README
samples cleanup pre-0.12
bump version to 0.1110
samples cleanup and intermezzo patches
final cleanup for 0.1110
enhance tests and coverage report. add heap parameter.
fix a test on FreeBSD
tweak readme
minor version tweak
more tests
big gobs of testage
test fixups
readme fixes
move kernel structures to lexicals
speed up kernel; fix bad filehandle in test 4
oops... add queue sample back
documented the object layer depreciation
depreciated sample programs
add dependency checking
add dependency checking
fix spurious error messages during dependency checking
rewrite dependency checker to use proper trees
improve the dependency summaries' grammar
usual test results tweaks before releasing
added an all-clear message
Set a date for 0.1201's release
oops... forgot a few things
revised documentation
tweak SEE ALSO/COPYRIGHT across the board
speed up a rather important (and inefficient) regexp
tweak the Preprocessor a little faster... add wheel IDs to ListenAccept's events... change hashrefs to listrefs... revise all but two manpages
Initial POE::Session documentation revision. Minor tweaks elsewhere.
POE::NFA initial revision; fix =over tags and constructor examples
revise README
Tweak README and Makefile.PL
Tweak Changes for distribution
fix noisiness when resuming unpaused handles in Gtk
fix noisiness when resuming unpaused handles in Gtk
fix noisiness when resuming unpaused handles in Gtk
touch-up distribution files for Wheel::Run and pipes
change the tests to use Pipe::(Uni|Bi)directional
rename unidirectional -> oneway; bidirectional -> twoway
remove TestPipe
tweak testsetup and properly skip tests in 22_wheel_run.t
fix Pipe::TwoWay's socketpair code
document the pipes and wheel::run
tweak README for next release; fix t/06_tk.t to use the right type pipe
final README/Changes tweaks for 0.1203
tweak readme again
forgot Wheel::Run in the MANIFEST. d'oh!
small error in pod
partial pty code in wheel::run... need to move to another machine
fix Wheel::Run's pty code
bad newlines in 2 files; document future depreciations
add state names to NFA parameters/methods
tweak controlling terminal fetch in Wheel::Run... broken
fix controlling terminal code for 4.3+BSD
add new tescase for POE::NFA
minor Changes and test fixes
add return tracing and asserting
updated hfb test system info and OS/2 test status
added ways to get a machine's current state
added Address parameter
documented changes for 0.1204
tweak postback docs
test Component::Server::TCP's new Address parameter
don't test Wheel::Run in pty mode if there's no IO::Pty
don't test Wheel::Run in pty mode if there's no IO::Pty
remove some debugging code that was leaving files around
to help automate tests, try to get gtk to close itself
to help automate tests, try to get gtk to close itself
to help automate tests, try to get gtk to close itself
finally get the gtk test not to wait for a user
remove warning caused by ASSERT_RETURNS
make ASSERT_RETURNS fatal where it wasn't
update with latest OS/2 test results
prepare for 0.1204 release
tweak gtk test for window auto-size
finally figure out the gtk window position thing
fix readme and changes for 0.1204 release
add cpan release numbers to depreciation schedule
add a way to trace test results
fix Wheel::Run pty conduits on Solaris
make Wheel::Run pty conduits run in raw/cbreak mode
clean up POD and whitespace
add a tk test to exercise fileevent bug
add a tk test to exercise fileevent bug
make the tk test failures work... or not, depending on how you look at it
initial fixed tk support... untested til it goes to the bsd box
tested tk support
comment change; postbacks are official
mainly documentation changes
update for 0.1205 release
fix date on Changes
break the monolithic kernel into personality modules
debug Event, Tk, and Gtk personality modules
macro-ize personalities
fix macro-ized event loop substrates
optimize substrate modules
moved object layer to a separate component
moved to a separate component
removed object layer documentation
split select logic into Kernel/Select.pm
debug test 8 (errors) and event substrate
clean up final warnings in test 8
add ReadLine wheel
add formal bits
fix newlines in banner
Added put() and modes to ReadLine wheel.
dissociate input cursor from display cursor
fix a race where the prompt was written needlessly
bad documentation must die
tweak regexp optimizer
remove some leftover debugging code
tiny tweaks for next release
minor tweaks
tweak OS/2 information
reduce the amount of time this test takes to run
remove os/2 from the testing information
reduce the amount of time spent in this test
stop pausing alarm watchers unnecessarily
document changes
add coverage statistics
oops! fix the release date for 0.1206
add a signal_ui_destroy method, and have the substrate_init_main_loop
remove the top-level window creation so people can write multi-paned
fix hostname to use POSIX::uname when available; remove UIDESTROY
Tk support relies on things after version 800.015. Add a version
update to stop using the built-in $poe_main_window, which has gone
add an alias parameter and a shutdown event handler
allow the Program parameter to hold a code reference
depreciate coderef state names
tweak the signal waiting time for a fine balance of robustness and
tweak the signal waiting time for a fine balance of robustness and
test fork/coderef/exit code
back-port to older File::Spec libraries that don't support splitdir
document cumulative changes
document other resources early, so people can find them sooner
clean up a new symbol from the POE::Kernel namespace between tests
add a stub signal_ui_destroy function so the interfaces are consistent
check the Tk version here... should go in NEEDS/deptest instead
add preliminary dependency version checking
add a signal_ui_destroy function to homogenize substrate interfaces
add ASSERT_USAGE to check parameters
fix a bad usage flushed out by ASSERT_USAGE
speed up tests by reducing delays
test wheels' event() methods
document changes
finally add Dieter Pearcey's stackable filters
add POE::Wheel::Curses and upgrade dscouch's status in the docs
added mouse events to Wheel::Curses
tweak README with latest test results
clarify what gets constructor args
bump version to 0.13
shotgun commit to catch stragglers before 0.13 release
fix dependency version checking in lib/deptest.perl; prepare for another cpan commit
extend debugging output on solaris
begin depreciating XyzState in favor of XyzEvent
begin depreciating XyzState in favor of XyzEvent
skip tk tests instead of failing them, if Tk is too old
update test results
update Windows support docs, and a few other bits
document depreciation steps done/to do and other things
bump poe to 0.1401 for internal testing
fix minor breakage when depreciating State for Event. now the tests pass
have the dependency checker be noisy on solaris
organize the depreciation schedules in Changes
add XyzEvent support and depreciate XyzState
ported wheel use from XyzState to XyzEvent
documented AS PPM install
documented latest changes
document CHANGES and TODO
rename Changes to CHANGES
add more details to the signal handler changes
add test for new feature
don't call select() without at least one filedescriptor on win32 (use sleep instead)
add detach_myself and detach_child methods (undocumented)
document session detaching
added a hook to change postback return values depending on event loop/toolkit, but haven't found a good reason to change them yet
document more plans for changing the interface
tweak a comment for no good reason
document new alarm interface and idle-time GC
tweak alarms TODO to reflect suggestions
bleeding-edge perl dislikes \_ in regexps. fix the warnings now.
document newer test/coverage results
document many more requests; update alarms schedule
document newer test/coverage results
add "June 2001" alarms functions, and document them
if POSIX::_exit fails, try kill KILL => 9 before falling back to plain exit()
add tests for June 2001 alarm functions
update coverage stats
update things done and plans changed
fix a %kr_alarm_ids leak
fix leaks in %kr_alarm_ids and %kr_signals;
insignificant tweaks
bump version number
document the request for exported constant prefixes
update input flow control notes with Torvald's suggestions
add and document alias_list()
add tests for alias_list()
change Changes case; add TODO
remove 0-width negative look-behind assertions
bump version to 0.1403
clean up regexp; remove commented out zero-width negative look-behind assertion so grep would stop seeing it
turn off TRACE_GATHER now that testing is done
reduce the number of events per session in each test
add new POE::Kernel constants to the symbol table cleanup between tests
update test info
add ASSERT_ALARMS to test alarm queue order; fix alarm_adjust to maintain queue order; minor optimization in _enqueue_alarm
add ASSERT_ALARMS to test alarm queue order
exercise the alarm queue's order extra hard
grammar nit in the documentation
remove ./lib from PMLIBDIRS. MakeMaker was making makefiles that installed some private files that ought not have been installed
document signals problems and the need for a batch preprocessor
document filter changing problem and possible solution
document the need for meta-documentation for guiding other developers
try to be clearer about the document's general form and purpose
document the need for better inheritance; yield's bad name
remove Exporter usage; turn off tracing and debugging options if they're set to 0
squelch diagnostics when optional modules fail to load
bump the version
fix the test results
bump the version te 0.15
tweak the testing and coverage results; edit for readability
Rewrite the introduction. Hopefully it'll be easier to understand.
all i did was fix a comment, honest!
apply Philip Gwyn's patch for better external data marshaller support
rewrite the input flow control notes
document get() method, and apply a patch from Anton Berezin
document implicit sessions
add ASSERT_DEFAULT and ASSERT_STATES
bump version to 0.1501
document the need for a POE that can facilitate one-liners
comment and documentation tweaks; no new functionality
fix tests so Kernel::ASSERT_DEFAULT doesn't cause missing state warnings. I'd used missing states on purpose, and now they're noisy!
documented an alternative signal propagation and handling scheme; documented an alternative to yield()
fix Solaris 8 test information
document that the event handler parameter offsets need to match Session's
move ARG0..ARG9 so they match NFA's, and document that they need to be the same in both kinds of session
remove an extra _stop in child processes
add Windows testing information
Enhance runtime error messages so they show the absolute line number
socketpair() doesn't seem to do the right thing for Windows systems; force pipes to use IO::Socket::INET instead on Windows
Wrap getpriority() and setpriority() calls in an eval block so they
bypass the signals tests on Windows
Add Windows compatibility constants to the symbols that must be
Revert const and enum constants back to Perl constant subs. Profiling
make the diagnostic more meaningful when a module cannot be found in %INC
fixed skipped compiled regexp tests in t/18_filter_line.t
develop some of the ideas herein
document change dependencies
document change dependencies
Add get_one_start() and get_one() for the alternate ReadWrite wheel
Add support for get_one_start() and get_one() to Wheel::ReadWrite. It
Document fixing filter changing, and fix depreciation language.
Fix some indenting. Whee!
Comment why this filter doesn't use get_one() etc.
Fix some comments, but nothing important has changed.
Made bad constructor parameter errors more descriptive.
Increased severity of using unsupported get_pending() method.
Using unsupported or unimplemented methods is now an error.
Added several new tests.
Added a test for the shutdown event.
Added tests for get_one_start() and get_one() methods.
Added tests for get_one_start().
Added get_one_start() and get_one() tests for discrete filters: Map
update coverage results
updated dynweb's test results. damn, he's fast.
fixed error reporting, breaking the filter's public interface (see manpage)
updated the sample server to use the changed Filter::HTTPD return values
try forward and back slashes just in case File::Spec->catdir() disagrees with %INC
Develop the filter changing fix.
document get_one_start() and get_one() and why they exist
tweak a comment
make some changes, and then back them out. all that's left are some whitespace changes.
Apply Anton Berezin's patches to Filter::HTTPD.
Apply more of Anton Berezin's patches to Filter::HTTPD.
add helper functions for POE::Pipe::OneWay and ::TwoWay
Make the code which sets filehandles nonblocking more robust. Retry
Use POE::Pipe's version of socketpair if pipe() and socketpair() fail.
Turn on ASSERT_STATES and add dummy states for events which didn't
add POE::Pipe
commented some of the strange data-driven features
try to avoid the windows equivalent of a coredump
Document Mac OS X test results.
Change the -e delimiter on Windows to a double quote. Booger.
added Philip Gwyn's debugging patches
update freebsd test results for Gtk 0.7008
add a kill() method
freebsd tests with Event 0.83 now
Pass filehandles by reference, or things tend to break.
add uncategorized user requests
add a clear() method to clear the terminal
Update signal ideas. Document plans for the successor to
s/depreciat/deprecat/g;
s/depreciat/deprecat/g; Update status of wheels, signals, and other
Update set_filter and set_input_filter to transfer pending input
Add streaming filter change tests.
set the version number to 0.1502
update test and coverage information
Add dynweb's most recent test results.
Add notes about stream paradigms. Document Artur's request for
Fix the way alarms interact with event loops.
Add substrate_reset_alarm_watcher to reset the next-alarm timer.
bump the version to 0.16
Added POE::Monitor. Added ideas to simplify POE::Kernel. Added ideas
added a gotchas/faq section to the docs
remove the double-run check from POE::Kernel
document coral's idea for timesliced code
Compared packed socket addresses on both ends of the socketpair
added dynweb's solaris and titanic's linux test results
forced filehandle flushes before exiting a child process (coderef execution)
bump the version to 0.17
update test results, FreeBSD, Linux, Solaris
Update the following sections: signal dispatching; safe signals; how
Sky's doc patch and content unmangling patch
TonyC's doc and carp patch
dummy states to prevent warnings
experimental SIGCHLD poll patch
patch SIGCHLD poll loop to only spawn one timer if CHLD and CLD are present
make warnings more understandable
recombine alarm and event queues
remove obsolete idle watchers
change 'state' to 'event' in places
comment on some duplicated code
fix a confusing thinko in the docs
set the minimum select() timeout on Linux to 1 microsecond. avoids a bug in Linux's select() syscall.
add capability to detect log resets
work around a DESTROY-time undef warning
fix alarm-not-found per Kirill
fix select assertions per Kirill
make posted events keep sessions alive
move unessential information to the web
fix FollowTail to pass freeside's test case
fix reference loop in filehandles watched by Tk
bump version to 0.1701
apostrophic jihad strikes again
fix some symbol table cleanup for new symbols in POE::Kernel
Move the client count to a constant. Add dummy event handlers to
fix a wheel id leak
remove an unnecessary level of indirection from the input state
fix a doc typo
add -> back into event names
Change %kr_handles to @kr_filenos and update *all* the logic to
Refactor macro interfaces and guts to work from the new @kr_filenos
Replace symbols associated with %kr_handles... with symbols associated
add (and comment out) some more handy debugging flags
Replace $self as the unique identifier in Wheel/*.pm. Instead it's
add newer wheels
Remove enqueued events from filehandle watchers that are being
add an ASSERT_DEFAULT for fun
Fix a syntax error from my previous commit. Removed most mention of
Removed some of the mention of synchronous or immediate events from
Add pause_input() and resume_input() methods and corresponding
Add a (commented out) TRACE_SIGNALS symbol for testing new SIGCHLD
Clean up the SIGCHLD polling logic. Mostly meaningless source and
Consolidate and clean up comments in the SIGCHLD polling logic. Add
fix a reference count mistake
apply Ben Laurie's signal handling algorithm
applied freeside's patch to allow hyphens in hostnames
fix test failure when HTTP::Status is missing
fix test failure fixes
Add default Driver and Filter to ReadWrite and FollowTail. Fix tests
add MyOtherFreezer to Filter::Reference instantiations
turn off tracing
Updated CHANGES. Added new files to MANIFEST. Added ReadWrite code
Apply Miroslav Madzarevic's fix for improper proxy session shutdown.
Add the ability to define extra states.
add parameters to signal()
patch Preprocessor to die on an undefined macro
credit more people
fix a typo in a skip message
fix Ann's e-mail address
fixed an unimplemented constant in Win32 POSIX
fix a POD-o
Fix the double-protocol bug Jason Boxman found.
updates so it doesn't crash on Windows
Filter parameter should not be a coderef
Fix the cause of the signal() redefinition warnings. Was being
apply Jason Boxman's URI patch
just skip the tests on win32
just skip the tests on win32
just skip the tests on win32
just skip the tests on win32
just skip the tests on win32
fix the bad file descriptor error in test 14
make Makefile.PL work better on MacOS
document flush() error status
add the default error handler (wasn't there)
add new dependent modules
add more Windows compatibility
improve an internal error
improve win32 support
fix shutdown semantics
document _stop problems
garbage collect event senders as well as their receivers
don't bother shutting down components that already are shut down
shut down wheels on errors
Skip some things that ActiveState Perl's POSIX doesn't export.
clear a new POE::Kernel constant between tests
fix POD syntax errors
warn if misusing POE's built-in events
-1 is not a process ID on Win32
remove undef warnings in new ASSERT_DEFAULT code
set slave pty's window size
add more traces to signals stuff
fix messages
re-fix messages
increase default block sizes
flag a server connection as shut down on error
Fixed a stupid closure trick that made every Server::TCP session use
fix bad parameter scarfing in shutdown
add a comment so something isn't erroneously corrected
remove a debugging warn
minor nit fix
remove experimental synchronous stuff
revise for imminent 0.1702 release
fix some minor typos
fix e-mail address
Bump the version number and halfheartedly revise the man page. The
increase the timeout to make tests more robust
fix my e-mail address
increase LARGE_QUEUE_SIZE to 512
Add a $VERSION number.
propagate cvs version number as $VERSION
t/00_coverage.t test skips looked an awful lot like errors. Make this
Add meta(1) so that international characters may be entered.
Tweak alarm docs.
Bump version to 0.1703. Document cookbook.
update for 0.18 release
Bump the version to 0.18
Call the Disconnect callback on forced shutdown.
Experimental revision of curses initialization order.
Make missing termcap warnings less ugly.
Redocument internal data structures.
Remove some deadwood. Apply a typo fix that Sungo spotted.
Remove a blank line.
Add CALLER_FILE and CALLER_LINE parameters. Add a get_runstate()
Made ReadLine use a private copy of STDIN and STDOUT so that
Wrap binmode() in eval{} because tied filehandle classes may die in it
Wrap binmode() in eval so that it won't die if a tied filehandle
Updated Windows and MacOS support info. in POE's man page.
Apply TonyC's suggestion for improved Win32 compatibility.
Turn blocking back on for STDIN, or some curses libraries won't
Rocco found and fixed a circular reference that prevented it (and thus
A half-hearted attempt at getting the test working on Win32.
Tweak the language in a comment. Nothing functional has changed.
Add blib and blib/lib to lib/deptest.perl, making it more useful for
Increase version to 0.19
Implement three-phase signal dispatch.
Three-phase signal dispatch added new constants to POE::Kernel. Clean
Add debugging information so that signal dispatch could be monitored.
Sessions just happened to be destroyed in the proper order for the
Make TRACE_ and ASSERT_FOO output more friendly.
Begin the next phase of FooState parameter deprecation in the POE
Remove mandatory tracing on Solaris since nobody has reported the
Remove implicit signal handling from the tests.
Removed use of deprecated SuccessState and FailureState. Made them
Alter ssid macro so it won't expand in a comment.
Deprecate _signal event, and by association using _default to handle
Various work-arounds for fork() not being supported under MacOS 9.x.
Added sungo's makefile magic.
Change the deprecation warning back to pre-5.005_03. It was at
Tone down the old-Perl warning. POE hasn't deprecated 5.004 just yet.
Peter Barabas was going through POE's source and uncovered a logic
Thanks to sungo, I was able to add magic to automatically generate
Fixed the filename in the auto-generated CHANGES magic. That's still
Ekkis found a system where POSIX::Termios::getospeed() on STDIN
While writing a sample program for POE's cookbook, I found that
I confirmed that Tk's fileevent watchers are broken on Win32. POE's
Make the "Windows don't do dat!" message a little more professional.
Moved ExtUtils::AutoInstall to a location where it may be found with a
Made some of the TRACE_SELECT warnings a little more descriptive.
Differentiate between package and object methods in an error message.
Remove a spurious $|=1 leftover from testing.
I added CloseEvent to Wheel::Run. This should address a long-standing
Added ARG4 to Wheel::Run's ErrorEvent. This extra parameter indicates
Increase the required version for ExtUtils::AutoInstall.
After spending several days beating my head against the fork/exec
Scott Beck discovered a long-standing (but rarely, if ever heretofore
ttyS on IRC was confused about the "Optional Full-Screen Child Process
Fixed a typo that Kane found in the docs.
Standardized the semantics of Filter and ClientFilter. Both may now
Document how easy it is to use YAML with Filter::Reference. Damn,
ReadLine now assumes vt100 if a TERM environment variable isn't set.
Re-revised the module categories. I know it's silly to think there
Make the Win32 skip messages a little friendlier.
The new select bug test failed on Win32 because it assumed select()
Skipped more Wheel::Run based tests.
My previous "fixes" for FollowTail weren't very good. This set of
I patched POE::Wheel::Run to use POE::Filter::Line by default. This
Applied some of the feedback from Android18, errr, and hdm on #poe.
Moved the C<use POE::Preprocessor;> statement to the top of each
Moved C<use POE::Preprocessor;> to the top of C<package POE::Kernel;>
Added the ability to dump files with expanded macros. This should let
Try to make the skip message so friendly that Windows users don't
Matt Sergeant contributed POE::Kernel::Poll, an event loop substrate
Properly attributed Matt Sergeant (not Scott Beck) for IO::Poll
Fixed the bug Ted Suzman reported on MSWin32 (re: SIGCHLD returning
Increased the amount of debugging information when TRACE_SELECT is
Jos Boumans discovered some segfault-like behavior when using
Prettified another message pertaining to a missing optional
Disable t/27_poll.t if IO::Poll is not installed.
Fixed a missing symbol warning when expanded source dumping is off.
Bump version number to 0.1901
Remove IPC::Run (we're not using it yet).
Added the second bit of sungo's Makefile magic. This part generates
Philip Gwyn spotted that TestSetup wasn't being used here. That's
Philip Gwyn discovered a "modification of a read-only value" error in
Perl 5.005_03 didn't like bareword IO::Poll on the left side of the =>
Perl 5.004_03 didn't like AUTHOR and ABSTRACT, so I removed them.
Have Makefile.PL touch the CHANGES file so that "make dist" will find
Final preparations for releasing POE 0.20.
Fix the cvs2cl invocations so they really include CHANGES files in
feelicks had a problem where his program never ran. It was traced
This test never called POE::Kernel's run() method, so we got a warning
Peter Chen noticed that the NEEDS file did not list Storable as one of
Update the instructions for using POE::Processor with PerlApp. Thanks
When the HTTP parser writes a hex dump, make it use a fixed font so
Justify the offsets in error dumps so things line up better.
Prettied up the error dump some more.
Extend the description of SuccessEvent's INET address field. This is
(!!!) Change the ClientFilter and Filter parameters in Server::TCP and
Increase the version number to 0.21.
Skip socketpair() and pipe() on MacOS pre-X. Even with a vastly
Richard Soderberg pointed out a couple things, which are now fixed.
Iain Chalmers reported 0.20 test failures under MacOS 9.1. These
(!!!) POE::Wheel::Run's Filter parameter was broken as designed. A
Update the documentation and carps and croaks to guide people away
Worked around a memory leak in Tk 800.023. Thanks to Zoltan Kandi for
Clean up the overview of Filter::Reference after someone (I forget!
Matt Sergeant noticed that POE::Wheel::Run sometimes creates defunct
As I once learned long ago, IO::Foo tends to be slow. In the old
Document why SysRW does not use BlockSize for syswrite(), and fix the
Updated to ExtUtils::AutoInstall 0.33.
Todd Caine requested a "Driver" parameter for Wheel::Run. While I was
Bump version to 0.2101 for release.
Cleared up a warning about Filter being deprecated. Thanks to ignatz
All the mentioned problems in 0.2101 were fixed. Looks like it's
Make the program runnable from lib/ or /, and make it fetch the tests
Added testreport.pl so it will be included in tarballs.
Move C<use lib> higher so it catches Test::Harness and friends. Work
Replaced these files with pointers to their counterparts in the POE
Document that the samples are moving to the POE Cookbook.
Don't accuse MacOS Classic (and other user-less operating systems) of
Tidy up a skip message for the test scoreboard.
PerlPanda on dalnet was disturbed that POE exports $poe_kernel into
Some people find the dependency prompts annoying. This change adds
Randal Schwartz discovered a problem in POE::Kernel::Event where it
Fix a documentation nit pointed out by Randal Schwartz.
Standardize the skip messages. See the style guide at poe.perl.org
Standardize skip messages, and use Symbol's delete_package() function
In a tight yield() loop, POE::Kernel::Gtk would fire and reset timers
Added tentative (untested) YAML support. It was a trivial extension
James March noted that Client::TCP and Server::TCP don't mention how
(!!!) Peter Chen is writing a SOAP server that exercises Server::TCP's
Changes in Client::TCP and server::TCP caused errors in the tests. I
"When three people tell you you're sick, lie down." Matt Cashner,
Sergey Skvortsov reported that Driver::SysRW uses length() to count
Invoke inline states with idiomatically correct Perl. This is
James March is trying to tweak a POE program for performance. He was
James March found some debugging prints leftover from testing. They
LordVorp discovered that Wheel/FollowTail does not document its
Cristiano Lincoln Mattos reported that SocketFactory was not emitting
L. Leo reported that run() could not be called twice in the same
Michael Stevens discovered a situation where signals could trigger a
Applied a modified version of Averell's patch to let a Client::TCP
Applied jinzougen's patch to have alias_list return an arbitrary alias
Added initial IPv6 support.
Fixed SocketFactory to support ipv6 host names rather than
Replace cvs2cl with a home-made program that creates much nicer
Added documentation and a note at the end where submitters can find
Add an AUTHOR section. Pedantic, yes, but credit where it's due.
Skip all tests if Socket6 is not loadable.
Added BindAddress and BindPort support, allowing client connections to
Replaced three sample programs with pointers to their updated
Added a heap flag for these components: shutdown_on_error. When
Bump the version to 0.2201 and start a feature freeze.
sungo spotted a warning in this test. It turns out that the test was
Fixed an undefined warning in a seldom-used block of code.
Added cvs-log.perl; added the report upload utility to MANIFEST.
Last minute tweaks.
Fix paths to cvs-log.perl.
Rebuild the XML test report if Makefile has changed.
It is possible to load Socket6 yet still not have IPv6 configured on
Applied James M.'s patch to correct some bad example code in the
Multi-file commits on slow links are sometimes smeared across several
Remove the link to the deprecated stinky hand-maintained test results.
Publish ClientShutdownOnError, which sets the defalut value of all
Apply JamesCOU's doc patch that stops saying things are optional when
Compatibility enhancements for running test reports on Windows
Peter Chen discovered and helped solve a problem where the subroutine
A previous change to Server::TCP "optimized" some code by moving the
Skip tests rather than fail if AF_INET6 gethostbyname2() works but
Skip a third test which was failing when AF_INET6 was not configured
Bump the version number to 0.23.
Bring it up to date with what's going on.
Fix a misleading skip message (Tk *is* supported, just not fully).
Erick Calder's clarifications on a couple points regarding delay().
Retry on EWOULDBLOCK or EINPROGRESS rather than dying outright. James
Explanation of the ErrorEvent parameters. Patch by Erick Calder.
Removed the use of POE::Preprocessor, except in ReadLine.pm which
Added two environment variables to set ASSERT_DEFAULT and
Tatsuhiko Miyagawa pointed out that this module uses HTTP::Response
Renamed the event loop "substrate" functions to "loop" functions,
Remove a public interface symbol from the event loop bridges. Add it
More code cleanup. Simplified Select.pm and Poll.pm by removing a
Fix the URL of POE's RFCs.
Increase the size of the "large queue" tests. They were never made
Extract POE's queue code into POE::Queue::Array.
Finish separating POE::Kernel from the event loop bridges.
Apply sungo's fix for tcaine's Wheel::Run test failures on MacOS X.
How could I have forgotten to add this before? Duh!
Beginning of the refactoring that will move all POE data structure
Rename POE::Kernel::* to POE::Loop::*. People following along with
Deprecated and moved. The entire program is replaced by a brief
Attempt to binmode() tied files, although do it in a block eval in
Split FollowTail into two modes: plain-file mode which uses "blocking"
Be more conscientious about cleaning up when FollowTail is destroyed.
Wheel::Run contains its own exit code, which tries to thwart object
We're going to try a different version numbering scheme. Odd
I copped out, sorry. I removed some parameter syntax test which
Split Kernel.pm into several subsystems. Although they are in the
Chop up all the long (>79 chars) lines. Inline a function that was
Undo the inline "optimization" from the last commit. The function
Implement POE_TRACE_FOO and POE_ASSERT_FOO environment variables.
Normalized the parameters for the enqueue and dequeue functions in
Add TRACE_RELATIONS, and move extra-reference garbage collection so
Reset all signal handlers to DEFAULT in the child process. Leaving
Clean up the trace/assert messages a little bit. Rename VEC to MODE
Fixed file descriptor tracking. The first rewrite was broken because
Renamed the VEC_* constants to MODE_* because they no longer are tied
While just admiring the cvs logs this generates, I noticed that I had
Classified the unclassified accessors as helpers, and tidied up their
Moved POE::Queue::Array documentation to POE::Queue and made it
Moved POE::Queue::Array documentation to POE::Queue and made it
Applied James March's patch to move the final leak checking out of END
This little ditty runs all the /^POE::/ modules on the CPAN against
Applied Casey Zacek's patch adding a ConnectTimeout to Client::TCP.
Removed support for /.*State/ parameters and upgraded their use from
Added a mandatory warning whenever a signal is handled by returning
Only import what we need.
Fixed a problem where the newest versions of a module were deleted
[May Break Code] Clean up the ASSERT_ and TRACE_ messages, and
Clean up an error message regarding Wheel::ReadLine and Term::Cap.
Minor edits before 0.2302 is tarred up.
Bump the version to 0.2302 for release. Feature freeze. Only commit
Update the ithreads warning, and make the delay for it a little longer
Bump the version up to 0.2303. Feature freeze is over, la la la.
I discovered the podchecker utility, and ran it on POE's
Addi mentioned using event counts to short-circuit queue sweeps. I
Add examples of Filter and ClientFilter parameters, thanks to a
Fixed a typo reported by Jeffz.
Described the timer features needed in more detail, per Stef's
Applied Garrett Goebel's doc fixes.
Fixed some of the Kernel traces that were confessing instead of
Jos Boumans provided a testcase for Wheel::Run not sending CloseEvent.
Applied Garrett Goebel's SocketFactory patch to support non-blocking
Apply Garrett Goebel's patch to make dynamic constants inline-able.
Fixed explained return values when using TRACE_RETVALS.
Found and fixed the error in TRACE_DEFAULT: _dispatch_event() was
Support le and ri in addition to LE and RI. We'll need to support
Carthag on efnet #poe pointed out that the last patch (also a response
Fixed a paste error in the documentation.
Added and documented pause_accept() and resume_accept() methods.
Applied Pedro Melo Cunha's patch to add a Started callback and an Args
Cygwin requires the exception flag be checked on sockets in order to
Fix the documentation for Started (was documeted as Start, which is
Add Makefile targets for PPM-compatible tarballs.
Applied Garrett Goebel's patch to support EWOULDBLOCK as a recoverable
Include MANIFEST.SKIP.
Pre-release tweaks.
Bump version to 0.24.
Remove some Win32 warnings about constant prototypes.
Bump POE's version to 0.2301 for development.
Damien Neil pointed out that Session.pm destroys its package_states
Apply Rob Bloodgood's patch to support 5.005_03 in POE::Kernel.
Removed an unneeded Fcntl module use that was also causing trouble
Applied Matt Sergeant's globref patch.
Applied Garrett Goebel's patch to clean up calls to
Applied Garrett Goebel's patch that allows call() to honor wantarray
Replace the homegrown blocking and nonblocking handle code with
Damien Neil discovered a temporary Session reference leak in the new
Scott Beck pointed out that statement-modifier foreach isn't supported
Jason Stillwell discovered that event handlers can't return undef.
Applied Jason Stillwell's patch to finish several features in this
Doc patch for Averell.
Document Started in the SYNOPSIS.
Halfjack discovered six subtests that failed. We tracked them down to
Mark each "resource" section with the name of the module it will
Bob Maccione discovered that ECONNABORTED (a connection was closed
The introduction of $handle->blocking() for ActivePerl 5.8.0 breaks
Wheel::Run can change input/output filters now, so remove the tests
One cannot assume that filehandles have methods such as clearerr().
The old HP-UX SIGRTMIN bug in Perl bit us again on Gentoo Linux. It
Bad Rocco! Always verify copied/pasted/tweaked code has actually been
Nancy Grady pointed out that Randal Schwartz's listing at
Collateral refactoring while trying to work around ActivePerl's
It is unfortunate, but a bug in ActivePerl 5.8.0's version of Tk
This fix is against against revision 1.14. Forcing that new test to
Finalize things for the imminent 0.25 release.
Bump the version to the next "CVS" number.
Cygwin doesn't give us warnings when creating this bogus
<sky> what happens is that the code is compiled twice, and the first
Clear the Client::TCP component's alias on shutdown. Otherwise, if
Hachi's double-free fix leaves the code in this file, but it's
The two-second timeout was too slow for the 486 I'm using to test
It seems impossible to avoid duplicate tags (identical timestamp) for
Applied and documented Dmitri Tikhonov's patch for the ProgramArgs
Patch the documentation for the Reuse parameter. Also support another
Apply Alan Hastings' Cygwin support patch.
Commit Alan Hastings' more proper patch for Cygwin support.
Fixed a leak discovered by dynweb. In client mode, SocketFactory
Update to version 0.50.
Applied Brandon Beamer's POD patch for $poe_main_window when using Tk
Work around the persistent double-GC problem by avoiding a second GC on a defunct session.
Fix POE::NFA's stop() method. Previously it was using POE::Kernel's
Applied Nick Williams' patch to avoid internal inconsistency problems
Added the ability to shutdown() input/output on a ReadWrite wheel
Add ExtUtils::AutoInstall's postamble() so the automatic installation
Require ExtUtils::AutoInstall version 0.50.
A previous patch to shutdown client sockets before they had closed
Increasing the version from 0.2501 (CVS after 0.25) to 0.2502 (first
Increase version from 0.2502 (first release candidate for 0.26) to
Add a warning to explain the most likely cause of this module hanging.
Increase the version number to 0.26 for release. Revise the README
Typographical change in a status warning. No big deal.
Split out the "resource" functions from POE::Kernel. The major data
Upgrade the mandatory signal deprecation warnings to errors.
Clear up some of the terminology in the session warnings/errors.
Expand the tests to include one-level subdirectories.
Comment observations made while writing new tests.
Add Resource.pm to document what we know so far about POE::Resource
Added an accessor to get at a session's reference count without
Aliases also affect sessions' master reference counts. Add tests to
Make sure that sessions' master reference counts properly reflect the
Add accessors to POE::Resource::Aliases so we can test the module's
Remove the previous testing-only accessors from Aliases.pm. Instead,
Add tests for session ID management.
Added the finalized-ok return value to _data_ev_finalize. Added some
Added the detailed tests for POE::Resource::Events.
Rename the initialize function _data_ses_initialize to follow the
Add the finalized-ok return value to _data_ses_finalize, so we can
Add the tests for POE::Resource::Sessions.
Added the finalized-ok return value to _data_sig_finalize().
Migrate a function call to the outside of a loop. That might cut down
Add some internals accessors so the tests can get at necessary values.
Removed some junk comments.
Add another sixty or so tests. :)
Added the finalized-ok flag and return value for testing. Corrected a
Add accessors so that tests can verify the subsystem's internal state.
Add t/res/filehandles.t (200-ish new tests! gods, that was tedious).
Enable TRACE_DEFAULT, and direct STDERR to a file so the tests remain
Some TRACE_FOO output was sent to STDOUT. Replace print() with warn()
Remove the code that blockes signals while tracking runtime
Ignore several things that falsely show up as uninstrumented code.
Add new junk files, sort, and generally tidy up.
Update initializers to reflect the current naming conventions.
Add some of our test results files to the ones that get unlinked on
Matt Harrington pointed out that Client:: and Server::TCP did not
Moved the plain-Perl signal handling code into POE::Loop::PerlSignals.
Forgot to add the new file in my previous commit.
Aw, heck. I forgot to update this to a development version.
Support Tk under ActivePerl by polling for socket activity every 1/10
Of course we can't really begin to support ActivePerl+Tk without the
Debugged the initial ActiveState Tk support on a real live Windows
Assimilate the latest (0.52) version of ExtUtils::AutoInstall. This
Remove a bare debugging warn.
"kaffee" reported warnings in loop_attach_uidestroy(). He suggested
Bruno Boettcher was banging his head on an ASSERT_USAGE message that
Clear up some error messages while documenting them.
Fixed a documentation typo.
Clean up some code structure. Nothing gets fixed.
Bruno Boettcher discovered an obscure, yet hideously fatal bug in
So apparently the reason ActiveState hasn't been building
Fix a test which was invalidated by the change in Resource::Extrefs.
Avi Finkel noticed a lack of a POE::Wheel::Write, or a write-only
Allow internal traces to be shunted to a file without using STDERR.
Nick discovered that sessions sometimes aren't garbage collected when
Correct localized STDERR for Perl 5.8.0.
The tests had some timing issues that only show up on the 486. This
Bruno Boettcher discovered the Gtk event loop didn't work with the
Remove the dependency on $^T and the assumption that it will never
Applied a modified version of jrieks' Seek patch. This lets you seek
jdeluise discovered a longstanding bug in cygwin's socketpair and/or
Merge in Bruno Boettcher's Gnome.pm as part of Gtk.pm since they're
Applied Rich Morin's documentation edits.
Leif Gustafson discovered a nasty little bug in POE::Pipe::TwoWay that
Fixed a subtle bug in the new pipe selection code.
Apply Leif Gustafson's patch to pass conduit types on to
jdeluise suggested that we document how to scrub a child process'
Carl suggested that the TCP client and server not report "error 0"
After many requests, I have finally added delay_adjust(). This resets
Silence the "connection reset by peer" errors, which indicate that a
Minor syntax cleanups as a result of trying (and failing) to determine
Azhrarn discovered that Linux only reports file shrinkage (nothing
Perform a proper fix for the file size shrinkage. Rather than reopen
Move the size-saving code so that it always moves the current size to
Reformatted a line to be less than 80 columns.
Dorian suggested moving the setgid code before the setuid code, so
Avoid ExtUtil::AutoInstall prompts if STDIN isn't a terminal. Bypass
James March's "UNIX Clients" recipe uses POE::Wheel::Readline.
Andrew Chen pointed out that the -9 signal doesn't work. Corrected it
Add a stub for Gtk2 support. Need a way that eliminates POE::Kernel
Include the tilde (~) as part of the legal character class for HTTP
Properly set the high watermark state when put() exceeds it. Also
Increase default block sizes to 65536. This improves performance in
Rethrow any errors caught in POE::Resource modules.
Add POE::Kernel::_trap for errors internal to POE's core. This is a
Replace _confess calls with _trap calls to increase the severity of
Replace _confess with _trap to increase some error severities.
Supply a B38400 symbol if the local version of POSIX.pm won't.
Prepare for 0.27 release.
There are no Gnome equivalents for Gtk->main() and Gtk->main_quit(),
Bline pointed out that kill(0) sent SIGTERM. Added a defined() to
Note a logic error in the cleanup code. The error isn't damaging, but
Add another no-END/no-DESTROY method for terminating a child process.
Increase version to 0.2701, to indicate we're in between-release CVS mode.
Various Disconnect callbacks weren't actually called when a remote end
Generalize POE's environment variable handling so you can specify
Propagate the POE::Kernel TRACE and ASSERT constants into any other
"rmah" and "czth" reported that Solaris and some BSD variants define a
Make TRACE_DESTROY more useful.
Document TRACE_DESTROY, and modify it to use POE::Kernel::_warn() so
Re-establish aliases on reconnect. Nick Williams discovered that
Applied a patch from Eric (l2kashe) that lets you specify different
Applied Dan McCormicks's callback() patch.
Add explanations to some of the _trap() calls, and comment on why we
Neyuki discovered that detach_myself() was not working from _start,
Move some delete() calls to the finalize stage of resource cleanup.
Add shutdown_stdin() to POE::Wheel::Run. Refactor some code in
Nick Williams at Morgan Stanley reported that the new "Please mail the
Reimplement _stop, mostly by moving the side effects out of
Exploratory commit of the new $VERSION code.
Ok. So I'm braindead from staring into SOAP modules all day. I
Deferring deletion of %kr_extra_refs items caused inconsistencies in
Allow POE::Kernel->run() to be called multiple times in Gtk programs.
Move the parent GC tests out of _data_ses_free and into
Replace POSIX errno constants with ones from Errno. The Errno module
Applied Scott Beck's patch to add a hard POE::Kernel stop.
Applied a POD typo correction from Alex.
Add a (commented out) option to go back farther in time. What's
Remove the hardcoding for POE::Loop classes. Now POE::Kernel will
Event loops. Look for POE::XS::Loop::$something before
Support auto-loading POE::XS::Resource::Foo if it's available. Fall
Reorder the timeout calculation to reduce latency in select().
Improve error detection in the POE::XS::Loop loader. Add autoloading
Applied Gerald Richter's patch to allow setting the child process'
Applied Alex Avriette's documentation patch. This mentions the return
While discussing libevent, I realized that a POE::XS::Loop::Libevent
Matt Cashner pointed out that ^D on an empty line should throw an
Quiet some spurious warnings generated by testing the errors that
Wrap postbacks in unblessed, anonymous subs under Tk. Tk blesses its
Apply Graham Barr's patch to REALLY disable SIGCHLD polling when
Define POLLRDBAND if it's not available. This occurs on MacOSX quite
Meaningless style tweaks.
Second attempt to define POLLRDBAND if it wasn't already.
Apply Nick Williams' POE::Resource::Performance class and the
Document run_one_timeslice().
Commit Nick Williams' patches to add documentation and fix some of the
Documentation, test fixes for Nick Williams' performance monitoring
Hardcode some keystroke character sequences for a few common
I accidentally rewrote the documentation for this module.
Minor style changes as a result of adding (then removing) a first stab
Apply Gerald Richter's fix to his earlier patch. This corrects
Renamed POE::Resource::Performance to POE::Resource::Statistics. The
Implement Alexander Zangerl's suggestion that Filter::Block have
Optimize the semantics of Filter::Block's custom length codecs---no
Apocalypse found a bug in the documentation for put(). This is such a
Bump POE.pm up to 0.28. META.yml doesn't change, but cvs still
The last commit DID modify META.yml. The new MakeMaker generates a
Add more development artifacts.
Bump version to 0.2801 for cvs development between CPAN releases.
The 5.004-and-earlier dependency checker picks up spurious uses of the
Hopefully this patch fixes all the paths to point to mylib instead of
Update to version 0.56.
Clean up paths to point to the new lib/ directory.
Work around "Too early to specify a build action 'Build'. Do 'Build
We're generating META.yml at make dist time from now on.
Make Build.PL and Makefile.PL play nicely together. The distribution
The generated META.yml must be touched early in Makefile.PL, or the
Bump the version to 0.2802. This will be a CPAN-released update for
Apply Corwin's doc patch exposing the "connect" event to the public.
I wrote a UDP cookbook example in response to someone on IRC. This
Jay Strauss pointed out that we're using InputLiteral instead of
Applied Scott's patch to speed up call(), which in turn makes wheels
Philip Gwyn pointed out that the signal() documentation doesn't match
Removed several of the sample programs. These are ones with
The redo statement is excessively clever. Commented it a little, per
Remove the dead files.
Increase version number for release candidate.
Add command line switches, and normalize whitespace so the results are
Tweak for the impending release.
Removed a workaround for older builds of ActivePerl.
Fix the coverage reporting tool to find the source files in their new
Turn off the debugging.
Fix the version number so CPAN won't index it.
Ton Hospel discovered that POE::Loop::Event wasn't terminating when a
Worked around t/30_filter_httpd.t failing on outdated versions of
Migrated the Socket6 code from the deprecated gethostbyname2() to the
Lance Braswell made some specific recommendations for using
Specify the address family and socket type to getaddrinfo() calls.
Set the PERL_DL_NONLAZY environment variable for the test report
Clarify the description of session lifespans when aliases are in
Bump version to 0.24_06 for another release candidate.
Anonymous Coward reported a bug in the code that skips testing a
Bump the version for the next release candidate.
Fletch pointed out that POE::Kernel doesn't discuss the POE_*
This commit localizes $SIG{__DIE__} while loading POE::Resources.
Improve filehandle tracing while verifying a last minute bug report.
POE.pm was moved during a recent directory restructuring. The test
Bump version to 0.29 for release.
Bump the version to indicate internal CVS development.
PAR understands POE::Preprocessor very well. Document this.
Lars Theglar writes: The ancient (0.47) version of Test::More in CPAN
Patched a doc nit noticed by Andrew A. Chen. Thanks, dude!
Include the _stop event handler's return value in the _child "lose"
Remove a race condition in the event queue resource tests. This fixes
Switch from our homegrown Devel::Trace-based test coverage reporter to
Removed the deprecated gethostbyname2() call. I had made this fix in
_trap_death() was not dying with the correct message. Now it should
Correct a typo in an error message. Also tidied up some whitespace.
Replaced the use of some public API functions with the private API
Note some TODOs in the source. One's an ugly bit of logging---a
The Tk event loop wouldn't end with TRACE_DEFAULT on (as in the
The SIGCHLD signal handler was not actually used. Instead we've been
Changed the way wheel IDs are allocated, so the first one evar is 1
The description of aliases was horrible. Now it's rewritten.
Docs: Clarify the role of the Args constructor parameters in both
Comment some code to increment event reference counts. It's only used
Add messages to the _trap() calls for various forms of internal
Moved parameter test assertions before their corresponding traces.
Clarify what "blocked" events are in some comments. The code confused
Moved more assertions before their corresponding traces. This
Replaced some silent returns with full-on croaks when a bad number of
Note a potential error condition in the source, and pose the question
Plug at least three memory leaks that only occur when ASSERT_DATA is
Improve Windows (and non-UNIXy) operating system compatibility. Using
Replaced the previous commit with File::Spec. This makes the event
An anoymous tipster pointed out that Acceptor is not valid when using
Mark Lawrence discovered that POE::Wheel::Run's shutdown_stdin()
Ton Hospel discovered that POE::Loop::Event does not exit when an
Nick Williams made extensive revisions to this wheel, adding most
Nick Williams fixed a bug where pressing Return (^M) from within a
Tweaked the POE::Loop loader to load things like Event::Lib as
Renamed POE::Loop::Poll to POE::Loop::IO-Poll, to normalize the way
Ignore modules within the test directory. Also allow coverage of
Remove the existing tests to make room for the new ones.
Commit the new tests, the new utilities to support them, and necessary
Migrate the prompt bypassing code out of the 5.5+ Makefile generator
Replaced by new tests.
Replaced by another test.
Replaced by other tests.
These are the wonderful tests we're adding to replace the stinky old
This utility derives tests based on templates and different
This test contained a hardcoded path to t/19_filterchange.t, which I
Stop excluding tests themselves. The reports don't cover all of the POE modules when they are excluded.
This commit moves signal deprecations forward. It incorporates a
Ton Hospel made a couple interesting observations about POE's event
Ton Hospel discovered that stop() from some event handlers causes the
Turn on ASSERT_DEFAULT, like we're supposed to in tests.
In retrospect, it was rather dumb to rename Poll.pm to IO-Poll.pm.
17:35 <autrijus> dngor: hey. you may wish to start shipping POE with
Fluffle said the documentation didn't explain how sessions are reaped
Swapped a couple words is a prompt, possibly making it easier to read.
Fixed some awkward indenting. This change is meaningless.
Removed the '-tests' parameters for each of the optional modules. The
God kills a kitten every time someone doesn't use POE. This change
Move some RUNNING_IN_HELL code to an appropriate file. Namely, the
Three-arg open is not supported in perl 5.005. This massages it back
Remove "use warnings" throughout the tests. It's amazing what flushes
Applying a patch by Martijn van Beers. This change adds syntax for
Removed the tk.perl sample program. It is superseded by the Cookbook
This sample program is covered in the "Evolution of a POE Server"
After a lengthy study into the reasons for Term::ReadKey failing, it
Ever since the experimental stop() was added, and run() was resumable,
Clear $! if syswrite() returns successfully.
Jonathan Steinert pointed out that the "can't select a handle 2+
We don't have tests that trigger the error I just "fixed", so the
Ok. This time for sure. Jonathan verified it before I committed it.
Removed the CAVEATS section, as the one caveat was ill thought out.
Rhesa Rozendaal reported a problem where XML::Simple uses a serializer
Philip Gwyn said my changes to his test invalidated it. That isn't
Allow users to register $kernel->sig("WINCH"). POE does not register
Scott Beck pointed out that the qq{} wasn't future proof. The eval
Get rid of redefined subroutine warnings when POE::Filter::Reference
This commit resolves rt.cpan.org ticket 8281: Attached patch brings
Fixed a typo in the POD, resolving rt.cpan.org ticket 8358 by Joost
Applied LARRYSH's massive spellchecking patch. This resolves
Applied the patch in rt.cpan.org ticket 7787. This adds a Started
Resolve rt.cpan.org ticket 6319 by asking people to "make
Use fileglobs to shorten the list of test files passed to MakeMaker.
In these enlightened times of ActivePerl 5.8.4, $handle->blocking()
Fixed the DISPLAY checking code so that test templates expand
So, recent versions of Tk trap exit() when it's called from BEGIN{} or
After expanding and refactoring POE's test suite, the command line to
Hurk burk! Why didn't I include this in the last commit?
Fix the skip message so Test::Harness can report the reason.
SourceForge has moved t to tests. This commit should point everything
Work around an intermittent failure in Tk when starting its main loop
5.005_04 exists, so the Makefile.PL deprecation warning has changed to
Perl 5.005_04 was not properly setting $! on important sysread()
IO::Poll went retrograde on CPAN. Back the required version down to
Use Carp explicitly rather than assume it's loaded by another module.
File::Spec::Functions isn't available in Perl 5.004_05. Downgrade to
Two-argument binmode() isn't supported in very very extremely old
Fix temporary STDERR redirection around tests for error and/or warning
Data::Dumper doesn't build on---to put it nicely---"legacy" versions
CROM COUNT THE TESTS! Uh... add a test count to the Test::More
The more versions of Perl we test on, the more new and interesting
Upgrade dependencies to versions that are known to pass tests.
Remove a vestige of debugging the issue where STDERR wouldn't come
Garbage collection behaves differently under 5.005_04.
Cosmetic change to wrap a line at 80 columns.
Disable the XS tests, seeing as we don't have any XS to test yet. I
The two-arg binmode() call wasn't needed. It was a vestige of
Skip the event loop restarting test if we're using Tk and a
Add an Id tag.
Put back some code that was removed in an overzealous change.
Remove the old dependency checker. It was reporting all kinds of
Upgrade to 0.61.
Downgrade requirements because ActiveState's repository is behind the
Combine the core requirements for ExtUtils::AutoInstall and MakeMaker
POE::Filter::Reference wasn't loading Compress::Zlib due to require() semantics changes somewhere between 5.6.1 and 5.7.4. Thanks to Nilson Santos (nilsonsfj on the CPAN) for reporting the problem and providing a suggested fix.
Bump the version to 0.30 for release.
Module::Build seems to enable YAML support at install time. If you
I broke the test report tool when I renamed the test directories.
Clean cup! Clean cup! Move down! Move down!
The test reports are generated based on which perl is running. The
Remove Build.PL. CPANPLUS spots it and tries to build POE with it,
Moved the META.yml generation to a separate utility. This lets us
Moved prerequisites, recommended modules, file cleanup configuration,
Use PoeBuildInfo.pm rather than hardcode common data in both versions
Add some comments.
Fix a syntax error, and have "make dist" fail all the time under old
Switch from variables to "constant" subroutines. Just because it
Oh, and we'll need to distribute PoeBuildInfo.pm, so put it in
Increase minor version for another 0.30 fix release. Hopefully the
POE 0.3003 seems to be stable. The first couple CPAN testers' reports
Resolve rt.cpan.org #1648. Tied STDOUT, STDERR would cause
Point the user at the RT ticket 8462, which is the umbrella for all
Tom Klein pointed out a typo in one of the POE::NFA error messages.
Mike Castle found references to the old POE::Loop::Poll in the
Khisanth on IRC reported a problem with the signals test. Somni
Because there's a behavioral difference between Test::More releases,
Bump version to 0.3005 for release.
Bump the distro version to 0.3006 to differentiate the CVS version
Richard Clamp pointed out that Filter::HTTPD doesn't import all the
This regression test requires fork(), but it didn't skip if the system
Bump the version to 0.3009 for a new CPAN bugfix release.
Martin Ferrari pointed out that there's an annoying warning when using
Fix a typo in the documentation, thanks to an anonymous tip in
Jay Kim reported that setsid() wasn't being effectively called. It
Applying Jay Kim's patch from RT ticket 8807. This adds checks to
Remove link to ADT::PriorityQueue per Sam Carmalt's report.
Remove new() from documentation samples. This resolves rt.cpan.org
Remove POE::Session->new() from tests. This wasn't an RT ticket, but
Remove POE::Session->new() from the tarball samples. Resolves
Add a tell() method so programs can resume tailing a log where they
Remove the minimum timeslice constant. Nobody complained while it was
Only poll for SIGCHLD when someone actually requests sig(CHLD) events.
Apply Matt Trout's NoSetSid patch to POE::Wheel::Run. When set, it
Fix the call() example in the docs. Thanks to Tim Klein for spotting
Tk::After timers don't use stop(). Rather, they use cancel(). I'm
Applied Martijn van Beers' patch to allow more than one request on a
Applied Martijn van Beers' patch to make callbacks add reference
Make the wantarray code in call() more explicit, but don't change the
Update the code style, but nothing else is changed. This was while
Part of ExtUtils::AutoInstall's automated install procedure involves
Nick Williams and Merijn Broeren at MSDW did the impossible and
Add Test::Harness to the dependencies in an effort to solve
This guy named Jack has written a WebDAV server with POE and needs
Replace the hardcoded "make" utility with $Config::Config{make} so
A cheap indent fix.
Uhlarik Ond�ej suggested that POE::Wheel::FollowTail should be allowed
Remove Test::More as a dependency. It has a tendency to want lots of
Applied Nick Williams' patch to add Hostname to Server::TCP. This is
Replaced {STDIN,STDOUT,STDERR}_FILENO with checks for fileno() in the
Replace exit(0) in the test templates with POSIX::_exit(0). Some
Documentation tweaks. Bump the version up to 0.31 for release.
Re-entering Tk::MainLoop too many times, plus signals, seems to
Apply Ed W's advice regarding non-blocking connect() in ActiveState
Resolve rt.cpan.org ticket 11056: Socket6.pm can be installed under
Add some demo comments about how to focus on just one event loop's
ActiveState Perl + Tk + POE::Wheel::Run + coderefs = Pain.
Enable support for $kernel->run_one_timeslice() under Tk. I had to
Remove tabs from source code. People (myself included) have been
Document that POE::Wheel::Run and POE do not reap child processes
Put TkActiveState.pm's $VERSION in the correct package. Module::Build
Compensate for machines with variable CPU rates. They cause
Clear up the purpose and activities of ClientShutdownOnError, at least
Fix some cosmetic problems in the end-of-run statistics report.
Moved the POE::Pipe::{One,Two}Way test code out of the POE::Wheel::Run
Skip the pty tests when IO::Poll or Event are used on Mac OS X. It
Use better choices for the poll(2) flags when watching filehandles for
Fix SIGCHLD and SIGPIPE so they are IGNOREd by default. In SIGCHLD's
The restructured tests cause one of the sessions to receive a SIGCHLD
General cleanups during a Devel::Cover debugging session. Be more
Fix a miscount in the number of skipped tests under Windows. Also
Fall back to a single POE::Kernel->run() in this test because multiple
Bump up the version. Dance! Dance!
Add pass-through parameters to POE::Kernel's select methods. Values
Add a regression test for hachi's last commit. His commit breaks
Add an Id tag. D'oh.
Modified the test for the internal _data_handle_add()'s new
Expose and document the new select() methods' arguments pass-through
Put Jonathan Steinert's patch back. Fix recursive signal dispatch to
Add a regression test to ensure recursive signal dispatch doesn't
The last two regression tests won't do much good if they aren't in the
Remove the unused ET_SIGNAL_COMPATIBLE and rename the explicit
Make the POE::Filter::* classes inherit from POE::Filter despite that
Modify get_one() to not destroy the reference's header in the buffer,
Fix tab damage I introduced in the last patch to Filter::Reference.
Deprecate the stinky old TestSetup library in favor of the shiny, and
Document new()'s return value, closing rt.cpan.org ticket 9373.
Add a mandatory carp() to POE::Session->new() per rt.cpan.org ticket
Lift POE::Filter::Reference's get() to POE::Filter, and modify most of
Arturas Slajus (with diacriticals, however) reported a bug in POE's
Explain how POE finds a POE::Loop module. This closes rt.cpan.org
Preliminary fixes for coredumps when calling POE::Kernel->run() more
Correct a typo in a $^O check. Cygwin's ID is "cygwin", all
Work around Cygwin quirks in the tests. 1. syswrite() blocks on
Windows XP under Virtual PC is functional but teh sl0. Increase the
Cygwin under Virtual PC is just too many layers of emulation.
Apply Merijn Broeren's patch to SocketFactory for ActiveState Perl.
Shorten and sweeten the documentation for postback() and callback().
Averell pointed out that callbacks via POE::Session->callback() do not
Increment the version for a new release.
Fix an amazing amount of tab damage. Yes, this commit is entirely to
Invoke event handlers in void context if the dispatcher is called in
Abhishek Jain discovered a heretofore obscure socket leakage issue in
Bump the version for a bugfix release.
Joel Bernstein pointed out a few warnings when POE is run under -W
Remove a $kernel->select() call that was accidentally left in during a
Bump up the version slightly for a new release.
Removed a delete() that was used to close a connection. Server::TCP
Philip Gwyn added a Concurrency parameter to PoCo::Server::TCP. It
Philip Gwyn discovered a slight problem with his previous patch.
Applied Philip Gwyn's patch to skip IO::Poll if it doesn't define a
Applied Philip Gwyn's patch to avoid errors on old Perl versions.
"Elly" in IRC suggested that the POE::Loop documentation explicitly
Add Jonathan Steinert's test to the MANIFEST.
Toy around with the idea of auto-generating MANIFEST at distribution
Resolved a conflict between two event types. The recursive signal
MSIE bites early and often. In this case, it's sending "\x0D\x0A"
Version update for release.
This release is different enough to warrant version 0.33 even though
Replace string evals with typeglob tricks to define various constants.
Whitespace cleanup.
These tests fail under Win32, so I'm skipping them. It seems that
Use the WNOHANG constant rather than the magic 0. Skip child reaping
Add the ability to skip certain classes of tests on certain platforms,
The end-of-run child process reaper should block if there are unreaped
Make TRACE_STATISTICS (and probably others) strict-safe. Thanks, Nick
Fixed a really dumb error... if (condition) { ... } else { do_stuff()
Bump up the version for a bugfix.
Remove ExtUtils::AutoInstall from the installer to support
Apply Dan McCormick's patch to carp on unknown constructor parameters.
ExtUtils::AutoInstall is no longer needed.
Bump up the version and throw in an underscore. Some people like to
Replace {% use_bytes %} with a Makefile.PL-time source filter, and
Clean up the preprocessor's backup files.
Remove POE::Preprocessor. Replaced it with a simple, almost one-liner
Remove some more vestiges of POE::Preprocessor. We'll get 'em all
Whitespace fix... I think. I can't actually see a difference when I
Remove mention of our homebrew test report system. It's been down
Switch from the cvs changelog tool to the svn one. I'm also testing
Change the VERSION strings from cvs style to subversion.
Ignore .svn directories.
Try to load POE::XS::Queue::Array before POE::Queue::Array. This
Make some minor syntactical changes to some comments to support a side
Commit MANIFEST changes.
Applied Jeff Pinyan's patch to fix tab completion when the only match
Remove some files that shouldn't be there.
Add a no_index entry for the mylib directory.
Applied Lars' patch to avoid division by zero errors in
Upgrade carp() to croak() in the new() constructor. This constructor
Get the regression tester finding and downloading POE component
Eliminate a memory leak using Tk with ActiveState Perl.
Chris Fedde, in rt.cpan.org ticket #16691, accidentally discovered a
Chris Fedde discovered that the "harmless" practice of closing STDIN,
Change the test timing to work around Virtual PC's
rt.cpan.org squashes indents in source code and patches, and some code
Use the emacs input mode by default. Everybody expects the arrow keys
Matt Sickler added clone() methods to POE::Filter and its subclasses.
Matt Sickler patched POE::Component::Client::TCP and the server to use
Explain a test result. Sometimes this test fails. It's not
Clear the we're-do-one-looping flag when a UIDESTROY signal is
Slow machines weren't seeing all the flushes. Turns out that the
Beef up error reporting for a sometimes-failing test.
Increase the timing tolerances on some tests.
We're 0.34 now. Make some pre-release README tweaks, too.
Now at version 0.34_01 so that cpan test reports from svn aren't
Add the tests directory to the no-index directive. This should stop
Fix a bug Rob Bloodgood discovered when installing on 5.5-ish systems.
Chris Fedde found a bug in the way Server::TCP creates filters. He
Make 0.3401 official for release.
Next development version.
Add more diagnostic information to tests that are failing under
Applied Marco Antonio Manzo's patch to simplify clone().
Add a test case for rt14444 based on test code by Matt Sickler.
Make an intermittently failing test's result message more informative.
Apply Matt Sickler's patches to fix POE::Session (and POE::NFA) and
Apply a superset of buu's patch, resolving rt.cpan.org ticket #18762.
Don't append a message to the HTTP headers if there isn't one. Avoids
Bump the version for release.
Minor release to fix bugs introduced in last major release.
Skip these two regression tests if HTTP::Request isn't available.
Move the bots over to extras. Remove them from MANIFEST. They're not
Allow POE::Queue::Array to be subclassed. Resolves rt.cpan.org ticket
Resolve Stuart Kendrick's bug, rt.cpan.org 19529, in which POE resets
Move alarm_adjust() after alarm_set(), per the recommendation of
Apply Randal Schwartz's support for pty-pipe, a POE::Wheel::Run hybrid
Attempt to resolve a documentation incongruity, per Tim Otten's report
Apply a fix described by Eben Rauhut. He pointed out a place where we
Bump the version for release.
Detabify the test generator. More importantly, turn on
Use POSIX::_exit(0) for the event loop tests when they're
Replace $accept_session with $accept_session_id, since that's what it
Improve the CAVEATS/BUGS documentation in these components.
Move the bytes.pm preprocessing from Makefile.PL to
Domain wasn't defaulting to AF_INET for some reason, even though it's
Commit Pravus' patch in rt.cpan.org ticket 19366.
Update POE::Wheel::Run's SYNOPSIS per Mike Baas' feedback. Also fixed
Avoid an undef warning in a test message.
Peter Farmer pointed out that POE_TRACE_FILENAME was being wrapped in
Remove a discussion of StderrEvent's parameters from the constructor
Applied Nick Williams' patch to remove newlines from history read from
Add signal handlers to the TRACE_REFCNT display.
Apply Nick Williams' (Morgan Stanley) patch to restore the currently
Remove some cryptic diag() statements that were cluttering up the test
Don't park the initial timer callback. Doing so causes Event to throw
Relegate some diag()s to DEBUG status, and then turn them off.
Skip this regression test if Win32::Console is not installed. Some
Skip the Wheel::Run tests if $^O eq "MSWin32" and Win32::Console is
Stop deleting the freezer module's path from from %INC. While this
Make a couple PEBKAC messages explain themselves more clearly and with
Differentiate between two kinds of watching-the-same-descriptor
Add a test for Randal Schwartz's bug, rt.cpan.org ticket 19908. Also
The previous leak fix broke some tests. Shaaaaame! This fixes the
Fix the comment that triggers vim's perl mode.
If not on MSWin32, skip because of that rather than due to the absense
Add an Id tag. Whee.
Don't throw exceptions when errors are caught in _stop. Resolves
Die, don't exit, when filehandle based PEBKAC errors occur. Otherwise
Can't run a .perl script as an executable on Windows? Bah! Execute it via
Make the MSWin32 non-blocking ioctl() call a little more informative.
Stuff works way better when you use variables in the scopes in which
Time for a new release. 0.36 is on its way out the door.
Make the test "name" more descriptive in case of error.
If Wheel::ReadLine must die because it cannot open a private copy of STDIN or STDOUT, at least have it explain WHY the open failed!
Apply Nick Williams's patch to only enable raw input when
Some CPAN testers failures have been addressed. Time for 0.3601.
Don't rely on STDIN and STDOUT being file descriptors 0 and 1,
Yuval Kogman discovered two problems in exception throwing for die()
Shut down signals in the tests that watch for them. Six tests would
Don't allow _data_ses_stop() to stop a session currently being
Remove tabs. Curse indentation!
Fix a method name. One was named _data_ses_... instead of
Re-instate exception handling from _stop, since we need it or
Improve the test by counting _stop invocations and making sure that
After dispatching a timeslice, check whether an unhandled exception
Save unhandled exceptions so they'll be rethrown from a safe place.
Add Yuval Kogman's test case for edge conditions in rethrowing die()
Replace POE::Loop::Event's direct use of Event's main loop with a very
Tabs to spaces. Bleah.
Add a test case (and fix) for a session garbage collection timing
Disable a large chunk of tests under Tk because seek() fails. Tk
Program => "string" may be double forking when run on some platforms,
Tests were failing under Tk because Tk traps exit() and throws a
_default no longer needs to worry about its return value if it's
Switch from exit() to CORE::exit() for testing under Tk. Have the
Disallow users calling put() on a POE::Wheel::Run instance that has
Skip the signals rerun test when using a combination of Event.pm and
Clear the SIGCHLD handler when we're done, otherwise the test never exits.
It really seems that Event doesn't catch SIGCHLD, or at least
POE::Wheel::ReadLine isn't supported on Windows. ActivePerl can only
Increase the child process timeout. It timed out accidentially when
Fix a Tk busy loop, reported by Michael Hartmann in rt.cpan.org ticket
Explain that read error 0 happens when the child closes one of its
Skip Wheel::ReadLine tests in ActivePerl. ActivePerl can't multiplex
Resolve rt.cpan.org ticket 20990. XS modules define stub functions
IO::Poll I/O callbacks were in a slightly different order than
Can't test Wheel::ReadLine without Term::ReadKey.
POE::Loop::Gtk doesn't perform I/O callbacks in exactly the same order
Instrument some code in POE's import, especially the code for explicit
Rename some of the tests. There were two 14s.
Add two new tests to the manifest. Skip more coverage databases.
A cheezy hack to avoid destroying Wheel::ReadLine instances from
Create proper data-driven tests for Wheel::ReadLine. You can tweak
Comment out the intrflush() call. It flushes to the terminal, which
Make this into a proper test. Keystrokes are pushed down to a
Put the intrflush() call back, and skip all my shiny new tests because
POE::Wheel::Readline would delete twice as many characters as
Start testing different editor commands.
Cover more conditions. Now with 27.5% coverage.
Move the POD tests to 10_units/01_pod.
Skip the POE::Wheel::Curses tests entirely if IO::Pty isn't available.
Pedro Melo suggested on the mailing list that clearing the components
Remove the disclaimer about needing system's tail for robustness.
Rename the samples files to examples files.
Rename samples to examples.
Updated Chris Williams' fine POD tests to more explicitly treat
Fix a vim modeline. Nothing here to see, folks.
Version 0.37! Rejoice!
Skip POE::Wheel::ReadLine testing if the module can't be loaded due to
Remove the sig() reference count. After this commit, registering a
Replace eval{use} with eval"use". Should resolve the test failure in
Added sig_chlid(), test case, and documentation. Resolves rt.cpan.org
More workarounds for Vanilla Perl not differentiating itself from
Skip the signals rerun test when running under MSWin32 and Event.
Time for 0.38.
Remove the cheezy hack from r2098 that prevents terminals from being
Document why this module refuses to run on MSWin32. At least two
Replace an unpack() containing a post-5.6 template character with a
Skip coverage tests for modules that fail to load. They're most
I found this (unused) module in a Pod::Coverage sweep. I'm committing
Matt Trout strongly suggests that version numbers should have a
Apply Florian Zumbiehl's docs patches. Resolves rt.cpan.org 21794.
The text of this error has confused multiple people. Try to make it
Avoid warnings if the result of a sysread() is undefined. Resolves
Removed the isa("POE::Filter") check from POE::Filter::Stackable.
Improve the prompts for skipping network tests.
The IO::Poll loop would busy loop without reporting errors on handles
Skip Wheel::ReadLine tests when poll() is in play on MacOS X. For
Get ready to rumble.
Remove references to POE::Preprocessor. Point to Filter::Template
Add a nonfatal mode to the resolver explainer function, so that alias
Philip Gwyn discovered that sig_child($pid) without a corresponding
Fixed a typo.
Add a test case and apply a patch by Whitney Jackson. FollowTail
Add Whitney Jackson's test case to the MANIFEST. D'oh!
Clear the $heap->{connected} flag on shutdown. Reported by Paul Evans
Rob Bloodgood supplied a test for an error when shutting down Tk with
Release the hounds! Or at least the code...
Bump the version to test sourceforge's new subversion server.
Skip this test if the underlying OS can't rename open files.
Rafael Kitover found out that some .inputrc entries can break the
Only enable this test when Test::Pod::Coverage 1.08 or later is
Stop setting $^W globally. Fix a couple warnings as a result. Thanks
Fix a doc bug found by Randal Schwartz.
An anonymous poster to POE's mailing list pointed out that put()'s
Lukas Mai discovered a bug in the include .inputrc directive and was
Some package tools expect Makefile.PL to return 1. And since this is
Eric Busto discovered that POE::Wheel::Run depended on $/ to be the
Time for a new release!
Cosmetic fixes in the high/low water mark code.
blocking() appears to be a FileHandle method, not an IO::Handle
I read on perl5-porters this trick to conditionally "use bytes"
Marlon Bailey discovered that POE::Filter::HTTPD threw incorrect
Shevek suggested that Fcntl be used to import F_GETFL, F_SETFL and O_NONBLOCK since POSIX doesn't always do it. Tests here show that using both doesn't throw warnings, so there doesn't seem to be harm in it.
Rob Bloodgood found a new deprecation. This change works around it.
Yuval Kogman discovered that sig_child() events were not triggering
Per STRO's suggestion, test and provide dummies for the IPv6 constants
Oops. Remove tabs.
Eric Hacker found a case where the IPv6 tests failed when AF_INET6
Resolve rt.cpan.org ticket 27833. Philip Gwyn found a case where the
Mark Jason Dominus' binary insert example (Higher Order Perl, p. 292)
Add a brief sleep to the child program to allow pipes to be flushed
Apply Gareth's patch for an optional synchronous put() and a flush()
Avoid a warning about ambiguous use of select().
Resolve a CPU spinning issue when using the Tk event loop without any
Bump up the version.
Attempt to make the timing more stable in t/30_loops/*/k_alarms.t
Applied Markus Jansen's patch to prevent a memory leak when Socket6 is
According to the CPAN testers (bless their souls), POE 0.9989 failed
According to at least one CPAN tester we cannot assume Storable
Set proper skip messages.
Revise documentation in a major way.
Need to redocument these. Also added configuration info. for the common loop tests when they are distributed separately.
Mark files that need to be redocumented.
Technically it is not a TODO.
Replace the symbol of Eris with the usual TODO.
Document "use POE" magic.
Move the POE.pm use lines to POE::Kernel. Mark that contributors need to be documented.
Partial revision. Documentation is currently complete but the structure is broken.
Make a little more progress on POE::Kernel before going to work.
Make a little more progress on the docs.
Rewrite the I/O watcher methods.
Clean up the prompts a little.
Make editorial decisions on Matt Sickler's doc contributions (thanks,
Rewrite state() docs.
Fix a couple typos.
Fix a typo.
Redocument reference counters.
Redocument Kernel state accessors.
Redocument Kernel accessors and lesser-used utility methods.
Document why sessions remain active.
Finish first pass through POE::Kernel. There are a lot of TODOs however.
Add a print() to the SYNOPSIS.
Remove POE::Api::Ctl per the warning on the mailing list.
Remove POE::API::ResLoader.
Remove some obsolete tests from the manifest.
Oops... reversed the sense of the want-to-run-tests question.
Minor edits to POE and POE::Kernel. Major revisions to POE::Session.
Make more progress on POE::Session
Rewrite option() documentation.
Rewrite docs for postback(), callback() and get_heap().
Incorporate some older event parameter constant documentation into the
Initial stab at _default docs.
Finish the first redocumentation pass on POE::Session. Migrate some
Redocument!
Discuss how filters and drivers interact with wheels.
See also: POE::Driver and POE::Filter.
Redocument.
Comment on sub-second accuracy requiring Time::HiRes::time(). Based on a discussion with and test case from Sjors Gielen.
Add prototypes to F_GETFL and F_SETFL.
Make SIGDIE asynchronous. It will no longer interrupt long chains of
Socket6 and Socket both export AF_INET6, but their prototypes don't match. Ask
Redocument POE::Loop.
Minor tweaks.
Add a reference to POE's subversion repository, per Zoffix's
Use IO::Pty helper methods to condition the slave pty rather than the
Apply RDB's patches to update tracing. Closes rt.cpan.org 30342.
Apply SREZIC's patch to work around an issue on 64-bit FreeBSD 6.2.
Remove the AF_INET6 warnings when running Makefile.PL on MSWin32.
Packaging up a release for the testers.
Solve POD test coverage issues.
Tweak the debugging output in a vain attempt to address rt 28804.
Bump up the version for the 2nd pre-release test.
Fall back to an 80x25 terminal size if Term::ReadKey::GetTerminalSize
Contrary to the docs, IO::Pty's clone_winsize_from() seems to die
Can't use AF_INET6 directly. Needed to fully qualify its package.
Looks like Perl 5.8.7 mistreats my @x = $o->method() =~ m//g; as an
Simulate a driver get() error by reading from a write-only filehandle.
Addressed several of the CPAN testers failures. Releasing a new
At long last, it is time.
LARGE_QUEUE_SIZE isn't needed anymore. Linear scanning wasn't that
Change the name of Setpgrp to SetPgrp to follow the form of SetSid.
First stab at a doc rewrite.
First draft of the new docs.
Removal of POE::API::ResLoader broke POE::API::Peek. This change
Resolve rt.cpan.org ticket 34803. Apocalypse pointed out that
Bump for release.
Use a much faster svn-log.perl from the latest repo-tools.
Increase the timeout in a test to avoid false errors.
Document how to avoid the run-wasn't-called warning. Thanks to Marc
Per Marc Lehmann's feedback, I have documented how to avoid a
Link the sig_child() documentation to the USE_SIGCHLD documentation
Actually, we must skip 4 tests if /bin/true does not exist.
Increase timeout for my slow VirtualPC test system.
Increase the timeout a little. Seeing some test failures due to
Clean up the F_GETFL/F_SETFL mess a little.
Bump version for release.
Redocument.
Fix a typo.
Redocument the specific POE::Loop classes.
Point out POE::Test::Loops.
Redocument.
Redocument.
Fix Jos Boumans' credit.
Redocument.
Redocument.
Redocument.
Redocument.
Redocument.
Redocument.
Fix a bug introduced during redocumenting.
Eliminate a call to get_next_event_time().
Fix a test that relied on unqualified stacked filter class names.
Support XS loops in POE::Kernel's import() syntax.
Redocument.
Redocument POE::Wheel::ListenAccept. Add examples for
Partially redocument POE::Wheel::SocketFactory.
Rename POE_LOOP to POE_EVENT_LOOP.
Redocument.
Redocument POE::Wheel::ReadLine. Remove tabs from
Apply Tony Cook's patch to fix the updated time watcher methods. Avoids a nasty loop spin when there are no pending events.
Make the same fix to POE::Loop::IO_Poll as was done in
Apply Chris Fedde's massive doc patch, resolving rt.cpan.org ticket
Minor NAME update.
Note some deprecated method names in the POD coverage tests.
Bump the version for release.
Remove the removed files from the MANIFEST.
Use the bleeding-edge POE::Test::Loops.
Time for a new release.
Add POE::Test::Loops to config_requires, so modern systems can install
Add build_requires and rename config_requires to configure_requires.
Move test generation to PL_FILES. Requires a separate "make" for the
Touch mylib/gen-tests.perl during Makefile.PL execution. This
Let's try this again.
Omit POE::Test::Loops from coverage output.
Revise contributed docs regarding circular references and the safer ways to refer to sessions.
Partial redocumentation.
Add Yuval Kogman's test to ensure proper signal dispatch order.
Fix a wee error pointed out by Zoffix.
Partial documentation rewrite.
More redocumentation.
Fix a POD error.
Note a TODO regarding extra refcounts and signal-based session termination.
Add run_while(). Untested.
Finish first draft rewrite of documentation.
Fix sig_child() documentation. There is not ARGS_LIST for that method.
Fix a minor example typo discovered by Apocalypse.
Fix style on Apocalypse's last commit.
Begin rewriting POE::Component::Server::TCP's documentation. Made a
Rewrote another chunk of documentation.
Skip POE::Wheel::ReadLine tests when using Event on darwin. Event
Document another chunk of POE::Component::Server::TCP, and deprecate
Do a bit more rewriting.
Fix tab damage. No significant changes occurred here.
Update Larwan's last patch for style.
Returns a hash, not a reference.
POE::Session would clear $^W unconditionally, which we consider to be
Redocument some more, and put in an =over to quiet POD tests for the
Fix a plethora of warnings that were uncovered by lexically scoping
Note that POE::Filter::Line strips newlines from input.
Document through new()'s paramters.
Finish the first redocumentation pass for POE::Component::Server::TCP.
POD doesn't allow =head5?!
Finish initial drafting of the last POE module. Woo-hoo!
Everything needs an edit pass. Add TODOs to every module.
Resolve rt.cpan.org #37938, reported by Rob Bloodgood.
Resolve rt.cpan.org #38974, reported by Jari Bonte. If TERM is
Bypass errors when trying to ignore filehandles whose fileno() are not defined.
Reformat some contributed code.
Try to clarify the _trap() message so that people look above the trap
Time for a developer release, to test all the changes that have accumulated so far.
Improve fcntl() error messages to help track down a 5.6.2 failure.
Avoid a lot of warnings, which are the inevitable result of loading all POE modules, even the ones which are alternate implementations of the same methods.
New test release.
Apply Philip Gwyn's testcase and patch for an end-time hang when
Fix a SYNOPSIS typo, discovered on IRC by Gcola.
Add even more fcntl() diagnostic information for a failure the cpan testers are seeing on 5.6.2.
Another developer release for 5.6.2 testing.
Move around some POSIX constants so they're used where they're needed.
Restrict what Fcntl exports. Bump up the versions for a new test
Gcola discovered that we don't explain POE::Filter get() and get_one()
Add Philip Gwyn's test for a failure for POE::Kernel to stop if
We need POE::Test::Loops 1.004 or later now.
Another development release. This one relies on the latest POE::Test::Loops changes.
Fix a doc typo reported by Abhijit Menon-Sen.
Fix Apocalypse's last patch's style to match the rest of POE.
Remove a new test that segfaults for some people, and bump the version for another (hopefully final) dev release.
Skip Tk tests if Tk cannot be started.
Skip Gtk tests if Gtk cannot be initialized.
Avoid IO::Pty crashes on fbsd 6.4 with perl 5.8.8.
New dev release to address issues with Andreas Koenig's smoker. He's setting a DISPLAY that cannot be reached.
Clean up some large loose ends in the POD.
At long last, a major revision.
Skip the IO::Tty dependency on MSWin32. Windows doesn't have
Time for a new release.
Some systems, especially SunOS, may return nothing on a sysread() while a
This test had a race condition when run on SunOS. It is fixed, with
Time for 1.006.
Fix a minor doc typo, thanks to Hinrik.
Comment out garbage collection testing at the end of call(). Neither
Resolve a test failure where the error message cannot be recognized on certain systems.
Apply Philip Gwyn's patch to deliver signals via pipe rather than
These tests have gone into POE::Test::Loops, where they may be useful
Gcola on IRC discovered that the internal signals socket wasn't
Apply Philip Gwyn's second patch from rt.cpan.org #47966. This one
Run perltidy & convert all tabs to spaces. No actual code changed
Try to speed up the tests by triggering the children to exit upon
Add rt47966-sigchld.t to the MANIFEST.
Add error number to signal pipe errors. Windows doesn't always have an error string for some of winsock's wacky error numbers.
Explain the likely cause and solution for the "Child process (pid)
Set the POE::Test::Loops dependency to 1.020, and bump the version to
Roll the version back slightly so CPAN can test before an official release.
Documented the bug tracker, repository, and other resource URLs.
Rely on POE::Test::Loops 1.021, which should be everywhere by now.
Some people may be using POE without actually reading the docs.
Initialize the output masks for sigprocmask. Undefined output masks
!!! This change breaks backward compatibility on a relatively unused
Fix case where POE::Wheel::Run CloseEvent doesn't fire on Linux.
Move Philip Gwyn's signals test to POE::Test::Loops, so it can
Describe Event watchers to aid in debugging.
Increase filehandle trace verbosity. Add more diagnostic information
Quiet Event's exception handler.
Defer rethrowing exceptions until after run() cleans up. Previously
Added experimental code to allow postbacks and callbacks to be
Remove $Id$ and $Revision$ tags. No significant code changed despite
For some reason, this wheel was not inheriting from POE::Wheel.
Mainly rewrite POE::Filter::HTTPD's parser. May support get_one()
Document POE::Wheel::Run coderef execution side effects. Disable
Create new project directories for POE::Loop::{Event,Tk,Gtk}. Moved
Explain why wheels can't be altered outside their creator sessions.
Comment on the likely cause of the deep recursion noted in rt.cpan.org
Document that it's good practice to scrap the connection when there's
Set revisions for 1.020 release.
Found and removed a lingering metion of the Event, Gtk and Tk loops.
Changing the version to 1.266 to eliminate indexer and packager issues. Thanks to Patterner on IRC for the reminder.
Fix POE::Wheel::Run stalling on MSWin32 when the Program is not valid.
Fix a typo in my last commit.
Skip the signals pipe when RUNNING_IN_HELL. Child "processes" in
Defer resetting $SIG{CHLD} and $SIG{CLD} until AFTER the waitpid()
[rt 49384] Explicitly bind to 127.0.0.1 in tests, reported anonymously.
[rt 36648] Give POE::Component::Server::TCP a ClientPreConnect
POE::Kernel->stop(), when called from a child process, must imply
Bump version for new release. Rely upon latest POE::Test::Loops.
Clarify POE::Kernel->run() semantics when called from within some
Use the IO::Handle and FileHandle modules, so that blocking() is
Bump the version for a release.
Only warn about USE_SIGNAL_PIPE when it's deliberately set on MSWin32.
Fix a typo in a warning, reported in IRC by Abhijit Menon-Sen.
Wheels inherit from POE::Wheel. I think someone asked me about this,
Fix a hanging problem on MSWin32, and bump the version for release.
Dispatch I/O synchronously rather than enqueuing and dequeuing events.
Reduce the amout of work the IO::Poll and Select loops do.
!!! Add mark-and-sweep testing for session stopping.
[fix rt.cpan.org 49904] Fixed a doc typo noted by Andrew Feren.
[rt.cpan.org 50025] Apply Andrew Feren's patch to incorporate Chris
[rt.cpan.org 50071] rt1648-tied-stderr.t seems to work on sufficiently
[ticket 49915] Make the "DONE" test whitespace-insensitive.
Clean up some comments just before some contested code.
Initial refactor of the POE::Wheel::Run::Win32 code.
Enable autoflush for the semaphore pipe.
Correct bugs introduced during recent Win32 changes.
Add in a forgotten semaphore pipe write. fork/exec on sane systems
Fixed a stop() bug while working on Marius Kjeldahl's test. stop()
Make Marius Kjeldahl's regression test work under ASSERT_DEFAULT.
Avoid garbage collecting the POE::Kernel singleton.
Split kjeldahl-stop-start into three separate tests: signal polling, SIGCHLD without signal pipe, SIGCHLD with signal pipe.
[rt.cpan.org 49915] Turn off debugging in the new tests.
[rt.cpan.org 49915] Inc/decrement the event count before the session count, so that the TRACE_REFCNT reference counts add up.
[rt.cpan.org 49915] Promote the experimental mark/sweep GC debugging output to TRACE_REFCNT. Cluck on increment/decrement so we can trace the execution that leads up to reference count changes.
I don't care what people say, I like test plans.
[rt.cpan.org 49915] Turn off debugging.
Fix the timing of calling has_forked() from stop(). Also fix the idle
Note in the comments a potential micro-optimization to try.
Fix some new mark-and-sweep GC debugging output. It was originally specialty debugging, and I forgot to convert print() to _warn() when I promoted it to TRACE_REFCNT debugging.
Make _data_sig_reset_procs() also stop polling for children. Used as
Add an option to enable all the broken MSWin32 tests.
Only stop watching the handles in the modes we've already watched them
Document the POE_DANTIC environment variable.
Dump reference counts on session-scoped increment and decrement. Just doing so on decrement told only half the story.
Only install __WARN__ and __DIE__ handlers to redirect messages when TRACE_FILENAME is set.
Work around circular references on $self. Uses Scalar::Util::weaken() until someone comes up with a better plan.
Redirect STDERR while testing some errors. ASSERT_DEFAULT is on and making noise when things are called wrong, even when we do it on purpose.
Adjust event and/or reference counts when Kernel options that use internal timers are in effect. PID polling and statistics collection affect the baseline counts, throwing off tests.
Make the tests generally ASSERT_DEFAULT clean. POE_TRACE_FILENAME hid a multitude of sins in the test output file. Update the vim options lines while I'm in there.
Set version for dev release.
Oh, right, depend on the dev POE::Test::Loops.
Display the POE::Test::Loops version, the better to tell when a CPAN
Fix the dependency on POE::Test::Loops, and bump the version for release.
Include Win32::Job, Win32::Process and Win32 dependencies when running
New version for a new release.
Update POE::Test::Loops dependency to latest CPAN release.
mousemask() is very not portable. Remove it.
Eliminate some temporary variables.
Don't bother propagating wantarray() and tracking return values if we're not dispatching ET_CALL.
Clarify session multitasking documentation per IRC feedback from a
Fix a typo in a code example reported by Hinrik Örn Sigurðsson on IRC.
Improve reset detection on Windows. Reduce polling when a file is idle.
Detangle filehandle vs. filename timer callbacks.
Fix filename-based reset detection in Windows.
Make this test line-ending neutral.
Fix regression in the new file rotation code.
Change the timing of file reset detection to match previous logic.
[rt.cpan.org 51772] Don't fire EN_CHILD,CHILD_LOSE when
Upgrade the detachment test to use Test::More.
[rt.cpan.org 51171] Olivier Mengué noticed that kill() returns a
[rt.cpan.org 52452] Attempt to make the destruction order in the test
[rt.cpan.org 24956] Set STDIN blocking. Large amounts of STDOUT data
Bump up the version for release.
Clarify what's going on in two Session References examples. Reported
Several doc changes to fix issues reported by garu in irc.perl.org
POE::Wheel::ReadLine appears to hang on AIX when STDIN is set blocking(1). Reported (and graciously debugged) by Criveti Mihai from the Ministry of Economy and Finance in Romania.
Bump the version for a new release.
[rt.cpan.org 53352] Correct spelling mistakes found by Jonathan Yu. Simplifies Debian package maintenance since that's one less downstream patch to worrry about.
Hinrik Örn Sigurðsson discovered that the new POE::Wheel::FollowTail
Fix POE::Wheel::FollowTail, at least as far as Hinrik's test and the rest of the test suite are concerned.
Bump revision for release.
[rt.cpan.org 53519] The issue was a disagreement in the interpretation of "\n" on Windows (DOS newline semantics). Changed the newline charater to "!", because I can.
Bump the version for release.
Fix a typo in a trace message, reported by "jussive" in irc.perl.org #poe.
Correct a documentation typo reported by jussive in IRC.
[rt.cpan.org 53777] Restore _start & _stop return values. Mikko Ehto
Define a _stop handler to mollify assertions.
[rt.cpan.org 53757] Apply G. Allen Morris III's spelling patch but not
Fix a cpantesters.org link that gcola reported as 404 in irc.perl.org #poe.
Correct a type reported by Hinrik Örn Sigurðsson in irc.perl.org #poe.
Resolve rt.cpan.org #54319. In polling mode, the wheel wasn't
Rely upon the latest POE::Test::Loops.
Remove dependency on Test::Exception (it appears to be unused). Fix a
Applied Andrew Feren's patch to avoid a POE::Wheel::Run memory leak
Replace the detach-during-start test with an enhanced test that
Fix state bleed between POE::NFA instances.
Some code cleanup in POE::NFA.
Bump the revision number for a new release.
POE::Pipe and its subclasses work best using Cygwin's pipe(), not IO::Socket.
Bump for yet another release.
Fix an infinite file rollover when open() succeeds but stat() fails.
Fix a sig_child() race condition discovered by Tom Lanyon and reported
Bump revision for new release.
Document how POE::Wheel::ReadWrite reports EOF.
Avoid undefined value warnings when trying to tab-complete
Fix dependencies and bump POE revision for release.
Bump revision for a new release. No significant changes, other than to build a new dist with the latest EUMM.
Add support for POE_CATCH_EXCEPTIONS environment variable.
Fix a method typo at Hinrik Örn Sigurðsson's prompting. Also improve
Apply Hinrik's patch to catch die() and his test case for proper
Apply Hinrik's test changes from http://gist.github.com/449231
Small optimizations. Avoid _data_stat_add() calls when not needed.
At long last, apply Ton Hospel's micro-optimization for get_next_priority().
Fix tests related to recent POE::Resource::Events optimizations.
Reduce calls to get_item_count() by one per _data_ev_dispatch_due().
Move non-alarms out of POE::Queue::Array.
Finish a sentence. Incomplete docs noted by Hinrik.
Remove some logic that stopped being used.
Clarify the refcount_{inc,dec}rement() return values per Olivier
Nick Perez pointed out that we don't absolutely need HighEvent, since
Version bump for release.
New version to spackle over failed one.
Fix a method name typo in TRACE_REFCNT code.
Resolve rt.cpan.org 59925, wherein sig_child() use would increase some
Bump revision for release.
Apply Hinrik's patch to increase test failure verbosity.
Chris Benedict reported that nmap will crash Wheel::SocketFactory
Chris Benedict reported that Mac OS X returns ESPIPE when accept() is
Add user continuation parameters to sig() and sig_child().
Bump version for release.
New version for a new release. Require a new POE::Test::Loops.
Convert the changelog and repository metadata to Git.
Allow call to POE::Kernel methods by class instead of by object.
Merge branch 'master' of github.com:rcaputo/poe
Remove alias_resolve()'s warning if the resolution failed.
Deprecate Perl 5.4. Next step will be Dist::Zilla for building distros.
Upgrade IPv6 support from Socket6 to Socket::GetAddrInfo.
Merge branch 'master' of ssh://poe.git.sourceforge.net/gitroot/poe/poe into wtf
Work around the possibility that Socket::GetAddrInfo isn't loaded when someone needs IPv6. Reported by Apocalypse in irc.perl.org.
Fix Apocalypse's MSWin32 TODO for non-Windows machines. Unless I've secretly broken it. :)
Bump dependency.
Try Sucket6 if Socket doesn't provide IPv6 address families.
Bump the version. Dist::Zilla would make this a lot less noisy in the changelog.
Bump version for release.
Add CLONE to POE::Kernel, for better fork() support in Windows.
Fix a "panic: top_env" occurring during CLONE on MSWin32.
Move session IDs into session objects.
Key POE::Resource::Aliases on session ID, not object. Makes alias data iThread-safe.
Move POE::Resource::Extrefs to session IDs.
Add a sequence number to POE::Kernel's ID.
Remove Session references from POE::Resource::SIDs hash keys.
Migrate from Session to SID keys for POE::Resource::Events.
Remove POE::Resource::Statistics. Move POE::Resource::FileHandles, POE::Resource::Sessions and POE::Resource::Signals data structures from session references to session IDs.
During Kernel ID recalculation, walk POE::Resource::*'s data, fixing up the old IDs.
Honor the POE_DANTIC environment variable in POE::Loop::IO_Poll.
SIGCHLD appears not to work in hell (MSWin32).
Standardize non-blocking on IO::Handle's blocking(0) call. This may cause problems for systems using Perl older than 5.8.1.
Require a version of IO documented to support blocking(0) on Windows.
Add process IDs to _warn(), _die(), etc.
Don't always require IO::Tty's TIOCSWINSZ.
Clear out the event queue by session ID rather than reference.
Rekey POE::Resource::FileHandles on file descriptor instead of handle.
Fix an indefinite hang in non-blocking connect() failures on Windows.
Resolve a GC timing issue with call(), discovered by flw and diagnosed
Merge branch 'master' of github.com:rcaputo/poe
Begin modernizing.
Silence AF_INET6 warnings, per Apocalypse's research.
Make Time::HiRes a mandatory dependency. Perl has included it for
Quiet a warning when calling getsockname() on an unopened socket.
Rework Apocalypse's blocking(0) change to avoid code BLOCKs.
Rework Apocalypse's blocking(0) change to avoid code BLOCKs.
Bump version & dependency for release.
Bump version & dependency for release.
Merge branch 'master' of gitorious.org:poe/poe into up
Merge branch 'master' of github.com:rcaputo/poe
Merge branch 'master' of ssh://poe.git.sourceforge.net/gitroot/poe/poe
Avoid silent POE::Wheel::Run failures when stdio has been relocated.
Document why POE doesn't use Dist::Zilla, so I have something to reference when someone else suggests that it would be totally rad.
Version bump for release.
Rely on new POE::Test::Loops.
Comment out a well-meaning warning that breaks all related tests.
Don't fail t/00_info.t if POE::Test::Loops is unavailable.
Localize $SIG{__DIE__} when rethrowing userspace exceptions from a safe point. Avoids shenanigans at a distance when reporting on userspace errors. May thwart someone's exception handling, but I don't know a good way to satisfy everyone here. Suggestions?
Add stack traces to _data_handle_remove() failures.
getnameinfo() returns host and port in the opposite order of
Apply Stephen Oberholtzer's patch to fix a test freeze.
Add stack traces to _data_handle_remove() failures.
getnameinfo() returns host and port in the opposite order of
Apply Stephen Oberholtzer's patch to fix a test freeze.
Merge branch 'master' of ssh://poe.git.sourceforge.net/gitroot/poe/poe into sf-up
Bump the version.
Allow event handlers to persistently alter $SIG{__DIE__}.
Avoid trying to dispatch SIGDIE to sessions in a POE::Kernel that has
Don't block on active child processes at endtime.
Bump revision to 1.350.
Remove Module::Build usage.
Work around a consistent timeout issue in Hinrik's POE::Wheel::Run test.
Merge pull request #11 from mnunberg/master
Merge branch 'master' of github.com:rcaputo/poe into merge
[rt.cpan.org 67680] Use IO::Handle::blocking() rather than $fh->blocking().
Bump versions for release.
Very large select() timeouts can trigger EINVAL on Mac OSX.
Bump version for release. *sigh*
Remove test reliance on deterministic _stop order.
Include the unknown or illegal alias in error messages about them.
Merge pull request #12 from hubie/master
[rt.cpan.org 76314] Untaint externally supplied domains.
[rt.cpan.org 76314] Untaint externally supplied domains.
Merge branch 'master' of ssh://poe.git.sourceforge.net/gitroot/poe/poe into blah-sf
Merge branch 'master' of gitorious.org:poe/poe into blah-go
Fixing up desynchronization among repositories.
Allow POE::Filter::Reference to not die on bad wire data.
Add Paul Webster's test for edge-case YAML input.
Bump the version for release.
Fix address untainting error.
Bump the version for release.
Docment that waitpid() calls interfere with sig_child().
Add more signal handling diagnostics when asserting data at shutdown.
Optionally export FIRST_UNUSED so subclasses know where in @$self to start.
[rt.cpan.org 79379] Document repositories and online change log browsers.
Don't confess on the use of AF_INET6 or PF_INET6 in SocketFactory.
Don't bother with IO::Tty on Cygwin.
[rt.cpan.org 77233] Apply Lukas Mai's suggested POD fix.
[rt.cpan.org 77907] Apply Alexander Becker's documentation suggestion.
Replace Z<comment> with =for comment, per perlpod.
Note how to get ClosedEvent when killing processes in Windows.
Fix iThread-safety issue in POE::Component docs.
Consistently stat file names or handles.
Merge remote-tracking branch 'remotes/sf/master'
Turn off debugging.
Fix a typo introduced in an earlier commit.
[rt.cpan.org 81550] Untaint %ENV as it's examined for TRACE/ASSERTs.
Add lib/POE/Resource/Clock.pm to MANIFEST.
Update sample programs.
Move the signal information out of the signal pipe.
Update copyright dates.
Replace POE::Pipe::* guts with IO::Pipely.
Add t/10_units/05_filters/51_reference_die.t to MANIFEST.
Eliminate an uninitialized value warning in the tests.
Make a new POE::Queue::Array method private.
Remove unused code from POE::Resource::Clock.
Skip POD::Coverage checks for POE::Pipe::OneWay and TwoWay.
[rt.cpan.org 87928] POE::Queue::Array didn't need POE::Resource::Clock?
[rt.cpan.org 87931] Redo _data_ev_adjust in terms of peek_items().
Bump version and dependencies for release.
I think it should be _filespec_to_fh() since it returns a handle.
Remove trailing whitespace.
[rt.cpan.org 1648] Tk may also tie STDIN to an object.
Another day, another release. Or at least another four hours. Bump version.
Merge pull request #13 from dmw397/master
Add David Webb's regression test to the MANIFEST.
Fail better if asked to serialize a non-reference.
Make _trap(), _warn(), etc. a little more consistent.
Refactor _dispatch_event(), extracting _dispatch_signal_event().
Remove some redundant C<no strict 'refs'>.
Switch some signal regression tests from ok() to is() for more detailed failures.
!!! CATCH_EXCEPTIONS revamp.
Fix a "Kernel has child processes" warning even when it doesn't.
[rt.cpan.org 89627] Documentation typo fixes.
Merge pull request #15 from nkukard/tcp-server-component-stopped-event
Merge pull request #16 from dsteinbrunner/patch-1
Merge branch 'master' of gitorious.org:poe/poe
Bump version for release.
Rethrow exceptions that occur during call().
Bump version for release.
Socket.pm supports unpack_sockaddr_in6() now.
Crudely benchmark the machine under test.
Add parens to method invocations. Comment some internal methods.
Convert POE::Resource::SIDs into a proper class.
Convert POE::Resource::Extrefs into a proper class.
Convert POE::Resource::Aliases into a proper class.
Revert "Convert POE::Resource::Aliases into a proper class."
Revert "Convert POE::Resource::Extrefs into a proper class."
Revert "Convert POE::Resource::SIDs into a proper class."
Windows reports local address 0 as 0.0.0.0, and then the test canot connect to itself.
Merge remote-tracking branch 'remotes/gh/master'
rt.cpan.org 91374. Clarify a warning when an optional dependency is needed but absent.
Version bump.
Update POE::Test::Loops dependency.
Version bump to trigger a new release with Chris' regression fix.
Fix a test hang on Windows.
Version bump for release.
Remove a harmless(?) debugging print() from a test.
Avoid uninitialized value warnings when testing in development.
Speed up a test by adding an idle event to POE::Wheel::FollowTail.
Abstract the regression test sequence helper into POE::Test::Sequence.
Don't fire undefined FollowTail idle events.
Reduce the time for another FollowTail test from ~5sec to ~0.2sec.
Bump version for release.
Shave 2.5 seconds off a regression test.
Shave about 3sec off a regression test.
Add POE::Test::Sequence::create_generic_session().
Take about 1sec off a regression test.
Bump version for release.
Replace // with || for that CPAN smoke box running Perl 5.8.9.
Stop loading Data::Dump. It wasn't even being used.
Version bump for release.
Merge pull request #17 from Hugmeir/master
Import carp() into POE::Filter::HTTPD.
SeekBack of 0 is fine on special files.
Add a failing test case for reporting die() from within SIGDIE.
Expose exceptions thrown from SIGDIE handlers.
Bump the version for release.
Tom Feist (1):
Modified Wheel::Run to accept winsize argument again, using Termios TIOCSWINSZ
Yuval Kogman (8):
Demonstrate usage of POE::Kernel->stop in L<POE::Wheel::Run/TIPS AND TRICKS>
Accept values from POE_([A-Z]+) environment variables for any defined constant, not just ASSERT/TRACE
Introduce the CHILD_POLLING_INTERVAL constant
Document all the "other" constants in their own section of POE::Kernel
Introduce optional (see USE_SIGCHLD) $SIG{CHLD} based child polling in addition to interval based polling
clarify trace message when polling for signals under USE_SIGCHLD
replace C<> with C<< >> when the formatted text contains a > symbol
Fixed instances of the package_states incorrectly referred to as class_states in docs
gregor herrmann (1):
Imported Upstream version 1.3650
jmadler (1):
Adding table to compare delay/alarm and related methods.
-----------------------------------------------------------------------
No new revisions were added by this update.
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-perl/packages/libpoe-perl.git
More information about the Pkg-perl-cvs-commits
mailing list