Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Commits on Oct 25, 2002
  1. [May Break Code] Clean up the ASSERT_ and TRACE_ messages, and

    authored
    standardize their categories.  The _DEFAULT ones still work. :)
  2. Fixed a problem where the newest versions of a module were deleted

    authored
    rather than the oldest ones.  Minor aesthetic changes, too.
Commits on Oct 21, 2002
  1. Only import what we need.

    authored
  2. Added a mandatory warning whenever a signal is handled by returning

    authored
    true from its event handler.  POE versions since 0.20 have recommended
    using sig_handled() instead.
    
    Added a mandatory warning whenever a "_signal" event handler is
    registered.  POE versions since 0.20 have recommended using sig() to
    register explicit handlers rather than using catch-all _signal event.
    
    Added a mandatory warning when calling sig_handled() in a "_signal"
    handler.  This includes handling signals from "_default".
    
    Removed _signal handlers from components and tests, since they now
    cause warnings.  For the tests which need to handle signals, we now
    call sig() to register those handlers.
    
    Removed _signal from the list of events which generate the "... but
    session foo has neither that state nor a _default state to handle it"
    warnings.  Otherwise removing _signal handlers would trade one warning
    for another.
Commits on Oct 19, 2002
Commits on Oct 15, 2002
  1. This little ditty runs all the /^POE::/ modules on the CPAN against

    authored
    the local CVS version of POE.  Useful for regression testing and
    uncovering problems in components.
Commits on Oct 14, 2002
  1. Applied James March's patch to move the final leak checking out of END

    authored
    and into explicitly called finalize() functions.  This prevents the
    checks from being triggered on crashes (which always leave some
    cruft behind).
Commits on Oct 12, 2002
  1. Moved POE::Queue::Array documentation to POE::Queue and made it

    authored
    generic (only minor edits). Replaced POE::Queue::Array documentation
    with a pointer to POE::Queue's.
  2. Moved POE::Queue::Array documentation to POE::Queue and made it

    authored
    generic (only minor edits).  Replaced POE::Queue::Array documentation
    with a pointer to POE::Queue's.
Commits on Oct 9, 2002
  1. Classified the unclassified accessors as helpers, and tidied up their

    authored
    names to match their new category.  Documented the abstract POE::Loop
    interface as POE/Loop.pm, and documented the specific loops with
    pointers back to it.
Commits on Oct 8, 2002
  1. While just admiring the cvs logs this generates, I noticed that I had

    authored
    badly formatted the dates.  This swaps month and day in the log so
    that the dates make sense.
Commits on Oct 4, 2002
  1. Renamed the VEC_* constants to MODE_* because they no longer are tied

    authored
    to select() bit vectors.  Also renamed the FVC_* constants to FMO_*
    for the same reason.  Tweaked the way _dispatch_event() is called as
    part of the voodoo of fixing Gtk and Event (see earlier commit).
  2. Fixed file descriptor tracking. The first rewrite was broken because

    authored
    different access modes (read, write) would overwrite themselves for
    the same file descriptor.  This version separates watchers by
    descriptor and mode, fixing LordVorp's problem and probaly also
    jerakeen's.
    
    Also cleaned up some of the warning messages, renamed VEC to MODE
    throughout, and fixed the _dispatch_event calls to match Kernel.pm's
    definition for that function.
  3. Clean up the trace/assert messages a little bit. Rename VEC to MODE

    authored
    throughout.  Standardize the _dispatch_event() calls to match changes
    in Kernel.pm.
Commits on Oct 3, 2002
  1. Reset all signal handlers to DEFAULT in the child process. Leaving

    authored
    POE's signal handlers active was the wrong thing to do.  Thanks to
    JamesCOU for finding this bug.
  2. Add TRACE_RELATIONS, and move extra-reference garbage collection so

    authored
    that it doesn't trigger a duplicate _stop event *from* a _stop event.
    This clears up the last of the known bugs introduced by accessorizing
    POE::Kernel's internal structures.
  3. Normalized the parameters for the enqueue and dequeue functions in

    authored
    POE::Kernel, and their use throughout all the files.  Tracked down and
    stomped a file descriptor reference count inconsistency.  Tested POE
    against its own tests, and tested my components against POE.  All
    tests successful.  However, more work needs to be done: A lot of
    warnings come up when SIGINT'ing a program.
  4. Implement POE_TRACE_FOO and POE_ASSERT_FOO environment variables.

    authored
    These let you turn on traces and assertions without modifying any
    code.  Also cleaned up a leak in the extra refcount management, and
    tidied up some error messages.
Commits on Oct 2, 2002
  1. Undo the inline "optimization" from the last commit. The function

    authored
    needs to exist for scoping reasons.
  2. Chop up all the long (>79 chars) lines. Inline a function that was

    authored
    only called from one place.  Fix a misspelled method name, which
    caused Rob Bloodgood's program to fail.
  3. Split Kernel.pm into several subsystems. Although they are in the

    authored
    same file, their data structures have been separated and hidden as
    lexicals in private scopes.  The only way to modify them is through
    public accessors.  The next step in the refactor will be to clean up
    the accessors, but I wanted to commit things while it still passes
    "make test".
  4. I copped out, sorry. I removed some parameter syntax test which

    authored
    created wheels outside of any session.  The resulting good syntax came
    out as errors because POE freaked when it tried to put the wheels
    somewhere that didn't exist.  I hope to make up for tonight's lameness
    with a decent suite of unit tests later.
  5. We're going to try a different version numbering scheme. Odd

    authored
    subversions will be for CVS stuff.  Even subversions will be web-only
    releases.  As always, plain versions (which only go out to the
    hundredths place) are reserved for stable CPAN releases.  Bump the
    version to 0.2301 to kick off the CVS phase (kind of late, though).
  6. Wheel::Run contains its own exit code, which tries to thwart object

    authored
    destruction so POE's own destructors aren't triggered.  The exit
    statement in the coderef test was thwarting it, causing all kinds of
    false leak detection.
Commits on Sep 30, 2002
  1. @sungo
Commits on Sep 28, 2002
  1. Split FollowTail into two modes: plain-file mode which uses "blocking"

    authored
    I/O and timer based polling on disk files, and fifo/socket mode which
    uses non-blocking I/O and select-based polling on special files.  This
    should do the right things for sockets, pipes, and plain files.
    Because it avoids NBIO and select() on disk files, it should also work
    on Windows and VMS.
Commits on Sep 27, 2002
  1. Attempt to binmode() tied files, although do it in a block eval in

    authored
    case it fails.  Don't bother making plain files (-f $handle)
    nonblocking, since it rarely makes a difference and sometimes really
    honks off the operating system (aka: POE dies in the attempt).  Oh,
    and much code movement (aka: refactoring).  Still not done with that.
Commits on Sep 26, 2002
  1. Deprecated and moved. The entire program is replaced by a brief

    authored
    message pointing to its new location on the web.
Commits on Sep 25, 2002
  1. Rename POE::Kernel::* to POE::Loop::*. People following along with

    authored
    CVS will need to "cvs update -dP" or check the whole thing out again.
Something went wrong with that request. Please try again.