Permalink
Commits on Sep 23, 2009
Commits on Sep 21, 2009
  1. Wheels inherit from POE::Wheel. I think someone asked me about this,

    but they didn't open a ticket to remind me.
    committed Sep 21, 2009
Commits on Sep 19, 2009
  1. Only warn about USE_SIGNAL_PIPE when it's deliberately set on MSWin32.

    Thans to Torsten Raudssus and gcola on irc.perl.org #poe for
    mentioning the problem.
    committed Sep 19, 2009
Commits on Sep 16, 2009
  1. Bump the version for a release.

    committed Sep 16, 2009
  2. Use the IO::Handle and FileHandle modules, so that blocking() is

    available at the time the signal pipe is enabled.
    committed Sep 16, 2009
Commits on Sep 12, 2009
  1. Clarify POE::Kernel->run() semantics when called from within some

    other event loop, thanks to criticism from "remlin" on a blog comment.
    committed Sep 12, 2009
Commits on Sep 8, 2009
  1. Unit test for r2687

    Philip Gwyn committed Sep 8, 2009
  2. Reset the idle queue size in Kernel->stop()

    Philip Gwyn committed Sep 8, 2009
  3. Prevent the signal pipe from being used during signal finalization

    Added a note about EIO in POE::Wheel::Run
    Philip Gwyn committed Sep 8, 2009
  4. This patch sucks less. _warn not defined at that point. Teach me to c…

    …ommit
    
    before make test.
    Philip Gwyn committed Sep 8, 2009
  5. Document the fact that the signal pipe isn't used on Windows.

    Warn if we are ignoring the signal pipe on Windows.
    Philip Gwyn committed Sep 8, 2009
  6. POE::Wheel::Run : Mask signals before fork(). Unmask when safe again.

    POE::Kernel Tweak doco.
    Philip Gwyn committed Sep 8, 2009
Commits on Sep 7, 2009
  1. POE::Kernel->stop(), when called from a child process, must imply

    POE::Kernel->has_forked().  POE::Wheel::Run recommends calling stop()
    to clear out the parent process' sessions before starting new ones in
    the child process.  Users should not also need to call has_forked() in
    this case.  Fixes a bug found by Marius Kjeldahl in irc.perl.org #poe.
    committed Sep 7, 2009
Commits on Sep 6, 2009
  1. [rt 36648] Give POE::Component::Server::TCP a ClientPreConnect

    callback.  Give POE::Component::Client::TCP a PreConnect callback.
    These callbacks are invoked just before POE::Wheel::ReadWrite is
    created in their respective components, allowing component users to
    reject connections or enable SSL on the sockets before beginning
    application I/O.
    committed Sep 6, 2009
Commits on Sep 5, 2009
Commits on Sep 4, 2009
  1. Defer resetting $SIG{CHLD} and $SIG{CLD} until AFTER the waitpid()

    loop reaps all outstanding children.  Resolves a deep recursion issue
    in the CLD/CHLD signal handlers on HP-UX and probably every other SysV
    based OS.  Resolves rt.cpan.org 45109 reported by, and graciously
    assisted by, Olivier 'dolmen' Mengué.
    committed Sep 4, 2009
  2. Skip the signals pipe when RUNNING_IN_HELL. Child "processes" in

    Windows take out the parent's sockets when they close, causing
    hilarity in the signal pipe code.  Probably in a lot of real
    applications, too.  Surely there's a better way?
    committed Sep 4, 2009
Commits on Sep 3, 2009
  1. Fix a typo in my last commit.

    committed Sep 3, 2009
  2. Fix POE::Wheel::Run stalling on MSWin32 when the Program is not valid.

    Resolves rt.cpan.org 49373, reported by Andrew Feren.
    committed Sep 3, 2009
Commits on Aug 29, 2009
  1. Changing the version to 1.266 to eliminate indexer and packager issue…

    …s. Thanks to Patterner on IRC for the reminder.
    committed Aug 29, 2009
Commits on Aug 28, 2009
  1. Set revisions for 1.020 release.

    committed Aug 28, 2009
Commits on Aug 27, 2009
  1. Fixes to POE::Filter::HTTPD:

        - RFC2616 says a body is implied if we have Content-Encoding or
          Content-Length;
        - C-E w/o C-L is an error for HTTP 1.1;
        - Added copious comments to justify the above;
        - Created ->_reset();
        - If possible, include the request in an error response object;
        - Set ->message on an error response object;
        - Updated documentation.
    Unit tests for above.
    Philip Gwyn committed Aug 27, 2009
Commits on Aug 26, 2009
  1. Comment on the likely cause of the deep recursion noted in rt.cpan.org

    ticket 45109.  Waiting for confirmation that Philip Gwyn's
    USE_SIGNAL_PIPE changes haven't fixed this particular problem.
    committed Aug 26, 2009
  2. Explain why wheels can't be altered outside their creator sessions.

    Potentially explains an oblique side effect reported by Sean Pieper in
    rt.cpan.org ticket 45408.
    committed Aug 26, 2009
Commits on Aug 25, 2009
  1. Create new project directories for POE::Loop::{Event,Tk,Gtk}. Moved

    the loops to their new project directories.  Configured the new
    projects for CPAN distribution.  Removed the event loops from POE's
    distro configuration.
    committed Aug 25, 2009
  2. Document POE::Wheel::Run coderef execution side effects. Disable

    the POE::Kernel->run() not-called warning if POE::Kernel->stop() is
    called.
    committed Aug 25, 2009
Commits on Aug 22, 2009
  1. Mainly rewrite POE::Filter::HTTPD's parser. May support get_one()

    now, so some tests also had to change.  I hope this will resolve
    rt.cpan.org ticket 48802, reported by a very patient zerhash.
    committed Aug 22, 2009
Commits on Aug 15, 2009
  1. Remove $Id$ and $Revision$ tags. No significant code changed despite

    most of the distribution being touched.
    committed Aug 15, 2009
Commits on Aug 13, 2009
  1. Added experimental code to allow postbacks and callbacks to be

    weakened, explicitly reducing their reference counts.  Untested.
    Resolves rt.cpan.org #48670, submitted by Adam Kennedy.
    committed Aug 13, 2009