Skip to content
Commits on Dec 23, 2005
  1. Add the ability to skip certain classes of tests on certain platforms,

    committed
    then use it to skip the IO::Poll tests on MSWin32.  First, IO::Poll is
    kinda broken on MSWin32 and falls back to select() anyway.  Second,
    some of the tests break outright in the IO::Poll fallback tests.
  2. Use the WNOHANG constant rather than the magic 0. Skip child reaping

    committed
    on Win32.  It seems to cause problems of the "segfault" variety.
Commits on Dec 22, 2005
  1. These tests fail under Win32, so I'm skipping them. It seems that

    committed
    Win32 socket creation isn't as concurrent as we'd expect, and
    deadlocks occur.
Commits on Dec 19, 2005
  1. Whitespace cleanup.

    committed
  2. Replace string evals with typeglob tricks to define various constants.

    committed
    It weren't exactly broken, but I fixed it anyway.
Commits on Dec 13, 2005
  1. Change exceptions to propagate out and kill the whole process if they…

    Jonathan Steinert committed
    … are unhandled.
    
    All tests pass Debian Linux 2.6.12.3something with perl 5.8.7
Commits on Dec 11, 2005
  1. Better testage for signal recursion causing corruption of the bookkee…

    Jonathan Steinert committed
    …ping values.
Commits on Dec 10, 2005
  1. Change exceptions test to better determine if POE has shut down prope…

    Jonathan Steinert committed
    …rly when the DIE signal is not handled
  2. This patch fixes the recursive signal testing added earlier. It uses …

    Jonathan Steinert committed
    …package globals instead of lexicals for the bookkeeping vars in POE/Resource/Signals.pm, this should be examined for a faster way if possible.
    
    All tests pass Linux Debian x86
  3. Forgot to add new test to MANIFEST

    Jonathan Steinert committed
  4. Welcome to recursive signals, this test makes sure that the sig_handl…

    Jonathan Steinert committed
    …ed() flag does not affect outer signals during recursive dispatch.
    
    This is currently only occuring when a signal handler causes an exception, and the DIE signal is called in response to this. This test fails on cvspoe currently, patch forthcoming to fix it.
Commits on Dec 9, 2005
  1. This release is different enough to warrant version 0.33 even though

    committed
    we haven't made our stated goals for this release.  I'll probably
    break up the milestones in rt.cpan.org and stop trying to adhere to
    specific goals for specific versions.
Commits on Dec 5, 2005
  1. Version update for release.

    committed
Commits on Dec 4, 2005
  1. update manifest for new tests. apparently the manifest got rearranged…

    Matt Cashner committed
    … internally so the diff looks way larger than it really is. add run_network_tests file to MANIFEST.SKIP.
  2. tests for new exception handling code

    Matt Cashner committed
  3. add documentation for the new exception handling code

    Matt Cashner committed
Commits on Nov 30, 2005
  1. roll exception handling (ala POE::Exceptions) into the core. this pat…

    Matt Cashner committed
    …ch 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 of _data_alias_loggable, for later logging, and it needed to check to see if the new session survived _start and _child dispatches. currently this new functionality is undocumented and does not have its own unit tests. this oversight will be corrected soon. right now, all current tests pass without modification on gentoo linux running kernel version 2.6.14 and perl 5.8.7.
  2. change a raft full of ok($string eq "string") tests to is($string, "s…

    Matt Cashner committed
    …tring") for easier test debugging
Commits on Nov 26, 2005
  1. MSIE bites early and often. In this case, it's sending "\x0D\x0A"

    committed
    after Content-Length octets of POST request, which is clearly in
    violation of w3.org statute number mumble.  Here we see the intrepid
    Philip Gwyn patch Filter::HTTPD to ignore whitespace after a full
    request.
Commits on Nov 21, 2005
  1. Mandatory warning for reaped processes to pair with sigchld semantics…

    Jonathan Steinert committed
    … changing. This loop will block the Kernel shutdown until all processes are reaped. All tests pass.
  2. Probably should test after removing the signal watcher again.

    Jonathan Steinert committed
Commits on Nov 7, 2005
  1. Change signal watchers so they keep sessions alive.

    Jonathan Steinert committed
    WARNING: This is a major semantics change in POE. It has the potential to make code 'hang' in places where it formerly did not.
    
    This change is necessary so sessions expressing an interest in SIG CH?LD do not die prematurely. (There is a planned mandatory warning for reaped children that were not being watched.)
    This change fixes RT 15215.
  2. Disable use of $SIG{CH?LD} = "IGNORE". This cannot be done in perl wi…

    Jonathan Steinert committed
    …thout breaking system() and friends. This fixes a failing tests/90_regression/merijn-sigchld-system.t
  3. Updated to check for number of sessions in run() and skip the event l…

    Larwan Berke committed
    …oop if sessions == 0
Commits on Oct 22, 2005
  1. Resolved a conflict between two event types. The recursive signal

    committed
    type shared a value with a statistics gathering event type.  Now they
    don't overlap.
Commits on Oct 19, 2005
  1. We need to watch for the child process ending in this test so we are …

    Jonathan Steinert committed
    …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.
Commits on Oct 14, 2005
  1. Toy around with the idea of auto-generating MANIFEST at distribution

    committed
    time.  I'm not satisfied with the opt-out nature of MANIFEST.SKIP.
    It's too easy to create a temporary file and forget to add it to the
    skip file.  Next thing I know, crap's being distributed in the
    tarball.  Possibly proprietary crap.  I can't have that.
  2. Start making this test more portable, someone please suggest somethin…

    Jonathan Steinert committed
    …g I can use on Windows
  3. SIGCHLD bug reported by Merijn Broeren on October 13 2005 on the mail…

    Jonathan Steinert committed
    …ing 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.
Commits on Oct 6, 2005
  1. "Elly" in IRC suggested that the POE::Loop documentation explicitly

    committed
    point out that POE::Loop classes aren't used directly.  And that the
    DESCRIPTION point to more detail about how the loops are actually
    used.  So here's a fix for that.
Commits on Sep 8, 2005
  1. Philip Gwyn discovered a slight problem with his previous patch.

    committed
    Increasing the concurrency limit wouldn't resume the listener.  This
    applies his fix.
  2. Philip Gwyn added a Concurrency parameter to PoCo::Server::TCP. It

    committed
    limits the number of concurrent connections the server can handle.
Something went wrong with that request. Please try again.