Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Commits on Jun 2, 2002
  1. Bump version number to 0.1901

  2. Prettified another message pertaining to a missing optional

    dependency.  Made an old one more cleare (gods, I hope) about the
    missing module being optional.
  3. Jos Boumans discovered some segfault-like behavior when using

    POE::Wheel::ReadLine on MSWin32.  ActiveState Perl does not gracefully
    handle BEGIN{eval{}}, so the BEGIN{} parts were successfully removed.
  4. Increased the amount of debugging information when TRACE_SELECT is

    enabled.  Part of the effort to make IO::Poll work on MSWin32.
  5. Fixed the bug Ted Suzman reported on MSWin32 (re: SIGCHLD returning

    positive PIDs).  Worked around the fact that IO::Poll is broken (or at
    least behaving inconsistent with its documentation) on MSWin32.
  6. Properly attributed Matt Sergeant (not Scott Beck) for IO::Poll

    support.  Fixed the bug that Ted Suzman reported in's
    SIGCHLD handling.  True processes return positive PIDs, which were
    being treated as inconsistencies in POE::Kernel.  Added a regression
    test for this condition.
Commits on May 30, 2002
  1. Matt Sergeant contributed POE::Kernel::Poll, an event loop substrate

    based on IO::Poll.  IO::Poll is potentially faster than select() in
    large scale servers (and other multi-connection programs).  I also
    tweaked a description in the 5005 Makefile so the intent of some
    modules is more clear.  Test 27_poll.t is neat, and I'm glad that it
    works.  It reruns test 04_selects.t with IO::Poll loaded, forcing
    POE::Kernel::Poll to be used instead of POE::Kernel::Select.
  2. Try to make the skip message so friendly that Windows users don't

    wonder whether they should install
Commits on May 29, 2002
  1. Added the ability to dump files with expanded macros. This should let

    perlapp and perl2exe "compile" POE programs now.
  2. Moved C<use POE::Preprocessor;> to the top of C<package POE::Kernel;>

    so it would see all of the code therein.  The "# include" directives
    are incompatible with the new POE::Preprocessor dump feature, so I
    removed them.
  3. Moved the C<use POE::Preprocessor;> statement to the top of each

    package it was in.  This lets the preprocessor (and the resulting
    expanded files) include everything in the package.
  4. Applied some of the feedback from Android18, errr, and hdm on #poe.

    These are minor doc patches to clarify some finer points.
  5. I patched POE::Wheel::Run to use POE::Filter::Line by default. This

    broke a subtest in t/08_errors.t when previously illegal usage became
    okay.  Heh!
  6. My previous "fixes" for FollowTail weren't very good. This set of

    fixes seems to work much better in tests.  The problem?  Not properly
    detecting when log files were reset.  In the process of fixing it,
    more debugging information was added, and the BUGS were updated.
Commits on May 28, 2002
  1. The new select bug test failed on Win32 because it assumed select()

    and non-blocking I/O worked on plain files.  Fixed it to use a socket
    instead, which Windows can cope with.
  2. Re-revised the module categories. I know it's silly to think there

    ever will be a wording that will satisfy everyone, but still I try. :)
  3. Standardized the semantics of Filter and ClientFilter. Both may now

    take a POE::Filter class instance or name.  For class names, they try
    building a standard (no parameters) instance of the class internally.
Commits on May 27, 2002
  1. ttyS on IRC was confused about the "Optional Full-Screen Child Process

    Support" message, so I revised them all to be a little less dense.
Commits on May 25, 2002
  1. Scott Beck discovered a long-standing (but rarely, if ever heretofore

    tickled) bug in POE::Kernel's _internal_select().  I tracked it down
    to a leftover hash key that wasn't removed when necessary.
    POE::Kernel now deletes it, and t/04_selects.t has a new test.
  2. After spending several days beating my head against the fork/exec

    paradigm on Win32, I have finally concluded that it is not
    consistently implemented from one Windows version to the next, making
    it a veritable nightmare of twisty variations to work around.  Neither
    IPC::Open2, IPC::Open3, nor IPC::Run work across the myriad flavors of
    "Win32", and I have neither the time nor inclination to attempt
    success where three such modules-- all dedicated to performing this
    very task, and it alone-- have failed.  Arf fthagn poit!  (Anyone who
    does have the time or inclination to fix it is welcome to, though.)
Commits on May 23, 2002
  1. Added ARG4 to Wheel::Run's ErrorEvent. This extra parameter indicates

    the child filehandle that generated the error.  For example, "read"
    error "0" on "STDOUT" means that the child's STDOUT was closed.  I
    also updated the Wheel::Run test to use CloseEvent instead of signals
    to detect when a child was done.
  2. I added CloseEvent to Wheel::Run. This should address a long-standing

    issue where it was difficult to tell when child processes were done
    sending information.
Commits on May 22, 2002
  1. Moved ExtUtils::AutoInstall to a location where it may be found with a

    simple "use" statement.  Also upgraded it to 0.32.
Something went wrong with that request. Please try again.