Skip to content
Commits on Sep 12, 2002
  1. Bump the version number to 0.23.

    committed Sep 12, 2002
  2. Skip tests rather than fail if AF_INET6 gethostbyname2() works but

    committed Sep 12, 2002
    none of the socket calls do.  This might indicate that while localhost
    (::1) is resolvable, the interface is not actually configured.  It may
    also indicate while the interface is configured, an unconfigured ipv6
    firewall's "deny all" may be blocking it.
  3. A previous change to Server::TCP "optimized" some code by moving the

    committed Sep 12, 2002
    ClientDisconnected code to _stop.  However, since very little can be
    done from _stop, this prevented post() calls from working there.  I
    discovered the problem when one of the POE Cookbook recipes stopped
    exiting as it should.  This change moves the ClientDisconnected code
    back out of _stop.  It also pretties the default error handlers for
    Client::TCP and Server::TCP.
  4. Peter Chen discovered and helped solve a problem where the subroutine

    committed Sep 11, 2002
    form of Program caused file handles to bleed into the child process.
    The result is that sockets linger open beyond the time they're
    needed-- at least until the child exits.  This patch introduces
    CloseOnCall, which emulates close-on-exec as Perl uses it.
Commits on Sep 10, 2002
Commits on Sep 9, 2002
Commits on Sep 8, 2002
  1. Publish ClientShutdownOnError, which sets the defalut value of all

    committed Sep 8, 2002
    client sessions' $heap->{shutdown_on_error} flags.
  2. Remove the link to the deprecated stinky hand-maintained test results.

    committed Sep 8, 2002
    The shiny, pretty automated one is much better.  Go there, or else!
Commits on Sep 5, 2002
  1. Multi-file commits on slow links are sometimes smeared across several

    committed Sep 5, 2002
    timestamps.  Previously the report would present them as separate
    commits with identical descriptions.  Now it detects and combines
Commits on Sep 4, 2002
Commits on Sep 3, 2002
  1. It is possible to load Socket6 yet still not have IPv6 configured on

    committed Sep 3, 2002
    the system.  Added a check for IPv6 localhost ("::1") and skip the
    tests if one isn't available.
  2. Fix paths to cvs-log.perl.

    committed Sep 3, 2002
  3. Last minute tweaks.

    committed Sep 3, 2002
Commits on Sep 1, 2002
  1. sungo spotted a warning in this test. It turns out that the test was

    committed Sep 1, 2002
    printing "ok 15" twice instead of 15 and 16.  Fixed.
  2. revert last patch. my mistake

    Matt Cashner committed Aug 31, 2002
  3. remove Macro/UseBytes from the manifest

    Matt Cashner committed Aug 31, 2002
  4. document the test server and the makefile targets

    Matt Cashner committed Aug 31, 2002
Commits on Aug 31, 2002
  1. Added a heap flag for these components: shutdown_on_error. When

    committed Aug 30, 2002
    cleared, it becomes the end developer's responsibility to handle
    errors and shut down connections when they occur.  It defaults to
    true, which preserves the old behavior of the components shutting
    connections down.
Commits on Aug 30, 2002
  1. Replaced three sample programs with pointers to their updated

    committed Aug 30, 2002
    counterparts on the web.
  2. Added BindAddress and BindPort support, allowing client connections to

    committed Aug 30, 2002
    come from specific interfaces on a multi-host system.
Commits on Aug 29, 2002
  1. Fixed SocketFactory to support ipv6 host names rather than

    committed Aug 29, 2002
    presentation (human readable numeric) addresses.  Added the test
    program I forgot when the initial support went in.
  2. Added initial IPv6 support.

    committed Aug 29, 2002
Commits on Aug 27, 2002
  1. Applied a modified version of Averell's patch to let a Client::TCP

    committed Aug 27, 2002
    component connect to a different address and/or port.  It first
    disconnects from any established server.  This feature is experimental
    pending documentation and tests.
  2. Michael Stevens discovered a situation where signals could trigger a

    committed Aug 26, 2002
    session's destruction twice.  The symptom is a "session does not
    exist" error in  This patch adds a simplified version of
    his test case to t/11_signals_poe.t, clarifies the "session does not
    exist" error message, and prevents what is essentially a double-free
    error in POE::Kernel.  Other files, such as and
    were touched during the course of this fix but essentially remain
Something went wrong with that request. Please try again.