Commits on Jul 26, 2009
Commits on Jul 22, 2009
  1. Explain the likely cause and solution for the "Child process (pid)

    reaped" messages, and why it's important for programs to avoid leaking
    committed Jul 22, 2009
  2. Add error number to signal pipe errors. Windows doesn't always have a…

    …n error string for some of winsock's wacky error numbers.
    committed Jul 22, 2009
  3. Wait for longer for all children to exit

    Philip Gwyn committed Jul 22, 2009
  4. Try to speed up the tests by triggering the children to exit upon

    demand rather than after a certain amount of time.  Doesn't seem
    significant on my G4 laptop, but it might be on faster machines.
    committed Jul 22, 2009
  5. Run perltidy & convert all tabs to spaces. No actual code changed

    this time around.  It's bad to mix semantic changes with stylistic
    ones.  The semantic changes get lost.
    committed Jul 22, 2009
  6. Added rt47966-sigchld.t which tortures the signal-pipe with CHLD, for…

    … testing
      on win32.  Maybe this test should be moved to POE::Test::Loops
    Philip Gwyn committed Jul 22, 2009
Commits on Jul 21, 2009
  1. Win32 fixes :

    - Assign fake numbers (>=128) to signals that don't have a SIG* constant
    - Don't call sigprocmask() and related
    Philip Gwyn committed Jul 21, 2009
  2. @bingos

    Make the somni-poco-server-tcp test explicitly bind to the loopback a…

    It is hit and miss if the OS supports binding client connections to
    bingos committed Jul 21, 2009
  3. Apply Philip Gwyn's second patch from #47966. This one

    builds upon his earlier patch on the same ticket.  It reinitializes
    POE's signal handlers in child processes after has_forked() is called.
    It documents has_forked() and USE_SIGNAL_PIPE.  It renames the _top
    methods to _bottom to follow interrupt handler semantics.  It avoids a
    race condition by blocking signals for the duration of pipe rebuilding.
    committed Jul 21, 2009
Commits on Jul 20, 2009
  1. Gcola on IRC discovered that the internal signals socket wasn't

    compatible with select() on Windows.  This patch makes it a plain
    socket when POE is running on Windows.
    committed Jul 20, 2009
  2. These tests have gone into POE::Test::Loops, where they may be useful

    for other POE::Loop writers.
    committed Jul 20, 2009
Commits on Jul 19, 2009
  1. Apply Philip Gwyn's patch to deliver signals via pipe rather than

    directly into the queue.  This avoids queue re-entrancy during find or
    modification operations, which has previously caused inconsistencies
    in the queue, event enqueuing and event dispatch.
    Philip's patch may be found along with a great problem description at
    committed Jul 19, 2009
Commits on Jul 15, 2009
  1. @bingos
  2. @bingos
  3. @bingos

    RT #47853: document Inline/Package/Object states as applying to clien…

    …t sessions not the listening session, by Michael Folwer
    bingos committed Jul 15, 2009
Commits on Jul 14, 2009
  1. @bingos

    RT #47852 'POE::Component::Server::TCP - ClientArgs is passed to the …

    …listener, too' applied patch from Michael Fowler, signed off by bingos
    bingos committed Jul 14, 2009
Commits on Jul 9, 2009
  1. @bingos
Commits on Jul 3, 2009
Commits on Jun 29, 2009
  1. Comment out garbage collection testing at the end of call(). Neither

    the caller's nor the callee's reference count is affected by the
    call() itself.  Anything that does affect the count should be GC
    tested separately.  All tests pass, which probably means we need new
    tests.  Thanks to DrGeek on IRC for asking an insightful question
    about call()'s GC.
    committed Jun 29, 2009
Commits on Jun 4, 2009
Commits on May 31, 2009
  1. Time for 1.006.

    committed May 31, 2009
Commits on May 30, 2009
  1. This test had a race condition when run on SunOS. It is fixed, with

    thanks to Gérard Henry forhis donation of time on a SunOS machine.
    committed May 30, 2009
  2. Some systems, especially SunOS, may return nothing on a sysread() whi…

    …le a
    pipe still contains data.  Subsequent reads get the remaining data, however.
    Resolve ticket 43252 by retrying a limited number of times when
    sysread() reports that the pipe is empty.
    committed May 30, 2009
Commits on Apr 21, 2009
  1. Time for a new release.

    committed Apr 21, 2009
  2. Skip the IO::Tty dependency on MSWin32. Windows doesn't have

    pseudo-ttys.  Resolves bugs reported by Chris Williams, Adam Kennedy,
    Offer Kaye ( #44931), and many more.
    committed Apr 21, 2009
Commits on Apr 16, 2009
  1. @bingos
Commits on Mar 31, 2009
  1. @bingos
Commits on Mar 29, 2009
  1. At long last, a major revision.

    committed Mar 29, 2009
Commits on Mar 28, 2009
  1. New dev release to address issues with Andreas Koenig's smoker. He's …

    …setting a DISPLAY that cannot be reached.
    committed Mar 28, 2009
Commits on Mar 27, 2009