1.28 2021-04-29
- remove C-only restriction on ->Inline method - thanks @oodler577

1.27 2018-11-17
  - only Zero(Polld) if not NULL (reported by @andk, thanks!)

1.26 2015-06-13
  - fix doc typos (Lucas Kanashiro)

1.25 2015-06-06
  - Fix Event::PRIO_NORMAL call in Watcher.pm (Lucas Kanashiro)
  - update Changes

1.24 2015-02-27
  - Don't distribute MANIFEST.SKIP
  - Update docs & code to require perl 5.008
  - Define MIN_PERL_VERSION to pass Kwalitee checks in Makefile.PL
    (Doran L. Barton)
  - Added =encoding utf8 to pass Test::Pod/Kwalitee tests. Added t/pod.t
    (Doran L. Barton)

1.23 2014-07-09
  - Add git repo
  - Rely on Inline "with" hook error message
  - use "our" not "use vars"

1.22 2014-06-30
  - Use updated Inline "with" hook API

1.21 2012-12-14
  - http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2012-12/msg00424.html
    (C warnings avoidance)

1.20 2011-07-27
  - Repair test again

1.19 2011-07-24
  - Really skip the bad fd test on platforms that don't support it
    Considering how much trouble this test has caused, I question whether
    it is worth keeping it.

1.18 2011-06-25
  - Repair test to use a truly read-only variable
    Fix suggested by Father Chrysostomos <sprout@cpan.org>

1.17 2011-05-18
  - Skip bogus fd test for GNU Hurd
    A test report from Nigel Horne suggested that GNU Hurd either does not
    support detection of bogus fds or supports it differently than Linux.
    I do not have Hurd available for testing so I will simply skip the test.
  - On windows, skip tests correctly
    APOCAL@cpan.org wrote:
    The module failed to pass the tests because it emitted incorrect TAP
    plan. After consulting with some people on IRC and looking at the TAP
    manual, I found the solution. It seems you were using the "old-style"
    TAP printouts and with the latest Test::Harness, it complains!
    The fix is to change those 4 tests: hup, io, reenter, and signal. You
    need to print one line for the skip, not two! The TAP doc can be found
    here: http://search.cpan.org/~petdance/TAP-1.00/TAP.pm#Skipping_tests

1.15 2011-05-03
  - Clarify conditions for skipping test
    Spotted by George Hartzell <hartzell@alerce.com>

1.14 2011-05-03
  - change skip condition

1.13 2009-09-30
  - Freshen MANIFEST
  - Note new GIT repo location: http://gitorious.org/perl-event/mainline
  - Add errata for tutorial

1.12 2009-08-26
  - Point people to Marc Lehmann's stuff
  - Skip bogus fd test for armv5tejl

1.11 2008-05-17
  - s/double/NV/g to support uselongdouble
  - Make debug code compile again
  - Increase minimum version of perl to 5.006_003
  - This test is too sensitive to load on the machine causing
    random failures. Better to remove it from the regression test

1.10 2008-01-15
  - min perl 5.006
  - Skip bogus_fd test on darwin. Reported by schwern@pobox.com.
  - Jerry D. Hedden <jdhedden@cpan.org> and Ævar Arnfjörð
    Bjarmason <avarab@gmail.com> correctly diagnosed that a recent
    change in perl (#31130, FETCH/STORE/LENGTH callbacks for numbered capture
    variables) caused $1 to loose its read-only-ness hence breaking test 6.
    Zefram <zefram@fysh.org> suggested using $$ instead. I have not tested
    this patch but it seems trivial enough.

1.09 2007-05-22
  - doc git location git://nirmalvihar.info/home/joshua/event.git

1.08 2006-11-01
  - Add casts to quell warnings as suggested by JDHEDDEN@cpan.org.

1.07 2006-10-10
  - Marc Lehmann <schmorp@schmorp.de> suggests using mg_ptr instead
    of mg_obj. Sounds good to me.

1.06 2005-05-14
  - Fix 64 bit issue.  Encouraged by Eric Rybski <rybskej@yahoo.com>.

1.05 2005-04-24
  - Fix ancient memory leak in loop() noticed by Nikita Savin

1.04 2005-03-29
  - Preliminary /dev/poll support from Michael Pergament
    <ESeifert@t-online.de>.  As of yet, there is no config test to
    turn on this code.  Patches welcome.

1.03 2005-01-31
  - Apply Win32 patch from Graciliano M. P. <gmpowers@terra.com.br>.

1.02 2004-12-08
  - Generic watchers: implementation, documentation, and tests.
  - Documentation refinements.

1.01 2004-11-29
  - Event.xs: Due to popular demand, queue_pending() is now a public
    API, for better or worse.  Beware of race conditions.
  - A var watcher can currently be created with no variable to
    watch, provided of course that it is parked.  However, once a
    variable has been set, the watcher cannot be returned to the
    no-variable state.  The variable to watch can be changed, but the
    watcher rejects any attempt to set the reference to undef.
    Applied a patch to allow the variable reference of a var watcher
    to be set to undef.
  - Applied patch to clarify the documentation (with tweaks from

1.00 2004-05-14
  - c/watcher.c (pe_watcher_now): Check for a missing callback, not
    only a missing perl callback.  Marc A. Lehmann <pcg@goof.com>
  - lib/Event.pm (import): Make NO_TIME_HIRES actually work. Jerry
    D. Hedden <jerry@hedden.us>
  - c/signal.c (pe_signal_stop): When a signal watcher stops, the
    signal counter is cleared.  Otherwise the signal watcher can
    generate one last event after being stopped, which is a
    bug. (Pointed out by Zefram <zefram@fysh.org>)

0.88 2004-04-21
  - lib/Event.pm: Simplify hooking into Time::HiRes.  Add a
    NO_HIRES_TIME export_ok symbol to turn off the probing.
  - Event.xs: Remove U2time. Remove install_time_api().
  - lib/Event.pod: More refinement from Zefram <zefram@fysh.org>.
  - lib/Event.pod: ($watcher->pending & signal watchers): More
    specific description by Zefram <zefram@fysh.org>.
  - c/watcher.c (pe_watcher_now): Fix SEGV reported by Zefram
    <zefram@fysh.org>.  Add test.
  - lib/Event.pod (timer): Correction by Randal L. Schwartz
  - lib/Event.pod (timer): Doc clarification (jdhedden@1979.usna.com).

0.87 2003-02-18
  - c/var.c (tracevar_r, tracevar_w): Fix declaration for recent
    versions of perl (patch from Nick Eggleston <nick@dccinc.com>).

0.86 2002-05-25
  - c/unix.c: Solaris works better without POLLWRBAND (Clemens
    Schrimpe <csch@Kiez.NET>).

0.85 2002-02-01
  - Apply patch from Allen Smith <easmith@beatrice.rutgers.edu>
    to avoid infinite polling loop on some select implementations.
    Also reported by Marc Lehmann <pcg@goof.com>.

0.84 2002-01-23
  - Doc StarvePrio (Allen Smith <easmith@beatrice.rutgers.edu>).
  - Remove IRIX special casing.  Reports indicate that new versions
    of IRIX don't need any hacks. (Allen Smith <easmith@beatrice.rutgers.edu>)

0.83 2001-06-22
  - Waste less time in timeable.c. (Peter Lombard <lombard@seismo.berkeley.edu>)

0.82 2001-06-21
  - Stop watcher if $w->cb(undef) (Damien Neil <neild@misago.org>)

0.81 2001-01-31
  - Fix callback.t fail without Time::HiRes.
  - Fix typo in EventAPI.h.

0.80 2000-12-07
  - Update Tutorial.pdf to Jochen's latest.
  - Remove silly warning on $io->poll(T).
  - Fix $DIED doc. (paul@miraclefish.com)
  - Change C callback prototype to (void(*)(pe_event*))
    at request of Jochen.
  - Warn if loop without active watchers.
  - Slightly more reliable default_exception_handler.
  - Inline 0.30 integration (briani@activestate.com)

0.79 2000-11-17
  - Removed Event/inactivity.pm
  - #undef HAS_POLL for sgi (treygraves@yahoo.com)
  - Moved REENTRANT & HARD flags to EventAPI.h. (Jochen)
  - Removed obsolete CBTIME flag. (Jochen)
  - EventAPI += sv_2interval. (Jochen)
  - Applied patch for IRIS poll() from R.de.Vries@fokkerspace.nl
  - EventAPI += unloop_all.

0.78 2000-05-24
  - Invoke Carp::carp for internal warnings.
  - Fix SEGV triggered when Event-Stats is enabled while loop is

0.77 2000-05-12
  - Fixed an obscure SEGV discovered by uri@sysarch.com.
  - EventAPI += NVtime.

0.76 2000-04-26
  - Fixed a horribly subtle bug in the handling of reentrant
    INVOKE1-style watchers (e.g. a timer) in nested loops.  Added
    test.  This change also removes at least one conditional from a
    hot code path.
  - EventAPI += unloop.

0.75 2000-04-24
  - Attributes called in "set" mode now return the new value.
  - Renamed data() to private() and added back the old data().
    Added tests & updated the pod.

0.74 2000-04-19
  - Resolved data() quandary.  Added test.
  - Updated perlqt.t example.
  - Spell checked pod.

0.73 2000-03-24
  - Fix for 5.6.0 with implicit context enabled.
  - Rename unix_io.c -> unix.c in anticipation of Win32 support.
  - Offer assurance about future compatibility (docs).

0.72 2000-03-10
  - Rename NetServer::ProcessTop -> NetServer::Portal.
  - Removed unused QUEUED & RUNNING flags.
  - Made pending() aware of context.  Added tests.

0.71 2000-03-03
  - Moved LoopLevel & ExitLevel from Perl to C, squeezing out some
    potential bugs.
  - Fixed SEGV triggered if $DebugLevel >= 4.
  - Warn on unmatched unloop.
  - Added Tutorial.pdf (by Jochen).
  - Moved semaphore & msg into demo directory.  No one has ever
    asked for them to be finished, so they aren't.

0.70 2000-02-25
  - Fix bug in select() engine spotted by perl@jochen-stenzel.de
  - Prune demo directory.
  - Add repeat.t demo. (perl@jochen-stenzel.de)
  - Add perlqt.t demo.

0.69 2000-02-18
  - Tweak documentation.

0.68 2000-02-15
  - More careful checking of
    intervals. (suggested by perl@jochen-stenzel.de)
  - Fix yet another typemap bug. (perl@jochen-stenzel.de)
  - Make io default to poll=r; more stringent start test;
    fix demo/stdin.t (suggested by perl@jochen-stenzel.de)

0.67 2000-02-09
  - Fix bomb in group watchers. (perl@jochen-stenzel.de)
  - Fixed another bug introduced in 0.61.  (Insure++)
  - Tighten typemap.
  - Warn about non-existant methods.

0.66 2000-02-08
  - Fix horrendously subtle bug uncovered by Event-tcp.
  - Eliminate (hopefully all) silent failures.
  - Restrict var watchers to
    scalars. (perl@jochen-stenzel.de is too creative)

0.65 2000-02-07
  - Group watchers for watching watchers.  See demo/group.t
  - Fix got() documentation, spotted by
  - Make callback a prerequisite for starting watchers.
  - Add test checking FIFO dispatch of equal priority events.

0.64 2000-02-04
  - Applied patch from gbarr@pobox.com that allows one to obtain the
    number of events a watcher has pending in the event queue.  Also,
    included was a cosmetic fix for the typemap.
  - Load the basic watcher types by default.
  - Adjust hup.t test (conrad@fringehead.org).

0.63 2000-02-03
  - Allow var watchers to watch the same variable.
  - Deprecated AUTOLOAD of Event->$type.  (Nick & Graham)
  - Reworked priority documentation based on comments from

0.62 2000-02-01
  - Allow implied stop() if required attributes are explicitly unset.

0.61 2000-01-31
  - is_cancelled() method (requested by
  - Doc updates.
  - Prevent starting unconfigured or badly configured
    watchers. Added tests. (perl@jochen-stenzel.de)
  - Refuse to start cancelled watchers.  Added test & updated doc.
  - If parked=>1 then the constructor will not invoke
    $watcher->start().  (uri@sysarch.com)

0.60 2000-01-26
  - Added $watcher->data() attribute. (This is Uri Guttman
    <uri@sysarch.com>'s fault.)
  - Fixed refcnt problem caused by extra invocations of cancel()
    (Discovered by uri@sysarch.com.)  Added test.

0.59 2000-01-20
  - Fixed serious memory leak (isolated by
    perl@jochen-stenzel.de). Added t/leak2.t.
  - Doc for timeout_cb (demanded by uri@sysarch.com :-).
  - Added demo/variable_timer.t along with minor doc
    update. (jsalmon@gw.thesalmons.org)

0.58 1999-11-22
  - timeout_cb for io watchers.  Added test.  (suggested by Uri
    Guttman <uri@sysarch.com>)
  - Copy the callback specification from watcher to event for
  - Removed support for clump.  The performance trade-off for this
    feature was probably unmeasurable.  Now it's definitely
    unmeasurable.  :-)
  - Rename Ev* macros to Wa* to make room for new Ev macros.
  - Remove EvNOW macro.
  - Added demo/readline.t. (perl@jochen-stenzel.de)

0.57 1999-10-06
  - Memory leak fixed: ioevents were being retired to the wrong
    freelist. (thospel@mail.dma.be)
  - Document is_active/is_suspended. (perl@jochen-stenzel.de)
  - Skip hup.t on Win32. (Paul.Moore@uk.origin-it.com)

0.56 1999-09-22
  - Fixed detection of EOF.  Added test. (Gisle Aas)
  - Figured out the strange pipe symantics on AIX and modified the
    io.t test to cope. (dfavor@austin.ibm.com)

0.55 1999-09-13
  - Changes for clean threaded build against 5.5.61.  Unfortunately,
    this breaks threaded 5.5.3.  If this is a problem for you then
    send patches.  The threading variations in the devel track are
    driving me crazy. (dfavor@austin.ibm.com)
  - Fixed coredump trigger by attach_to test (cpan-testers@perl.org).

0.54 1999-09-10
  - Improved cb(...) diagnostics.  Added tests.
  - Devel-PPPort (schinder@pobox.com of cpan-testers@perl.org).
  - If a file is unexpectedly closed, the io watcher is no longer
    stopped.  Instead the file handle is reset.  This symantic allows
    io timeouts to continue to trigger and should make recovery
    simpler.  Added tests.
  - Prohibit attach_to blessed refs.  Added test.
  - Minor doc updates.

0.53 1999-09-09
  - Fix Event::MakeMaker to pick the correct install directory.
  - This release is NOT binary compatible with 0.52!
  - Obscure potential typemap bugs fixed.
  - Reworked the typemap to avoid the exotic 5.6 self-tied hash
    mechanism.  It should now be fairly easy to backport Event to
    older versions of perl.
  - In the tradition of adding innovative (or at least strange)
    features to Event, constructors now accept (attach_to=>$ref).
    This will cause the watcher to attach its information to the given
    $ref instead of to a fresh anonymous hash ref.  For backwards
    compatibility the default is still an anonymous hash.
  - Removed a bunch of deprecated features.
  - Reformatted the source to be more consistent (if not beautiful).

0.52 1999-08-18
  - Moved t/inactivity.t to demo/ because it fails occationally and
    I don't have time to figure it out.  You might say that it takes
    too much time to write a good test of inactivity.
  - Make mention of Time::Warp.
  - Don't complain about Time::HiRes (uri@sysarch.com)
  - Skip t/idle2.t if Time::HiRes is unavailable (lukka@iki.fi &

0.51 1999-07-23
  - Notice bogus fds in both poll & select.  Added test.
  - Make mention of the mailing list archive.
  - Don't request benchmark results (artur@vogon.se).

0.50 1999-07-15
  - A few Win32 tweaks to get working tests to pass (jan.dubois).
  - Minor cleanup in preparation for 0.50.

0.44 1999-06-30
  - Release 0.44 (thanks: jan.dubois@ibm.net).
  - Renamed abort() to scrub() to avoid clashing with Perl's abort.
  - Removed remaining dependency on gettimeofday.
  - C++ clean.

0.43 1999-06-22
  - Tweaks for threaded compile (dfavor@austin.ibm.com).

0.42 1999-06-19
  - Rename event.c -> ev.c (Jan Dubois).
  - Split flags into individual methods: is_running, is_active,
    is_suspended, is_queued.
  - Repaired default_exception_handler.

0.41 1999-05-19
  - Added attributes() and configure() methods.
  - When using eval {} inside a callback, $@ could fool Event into
    thinking that an exception is being thrown.
  - Moved attributes from hash entries to methods.
  - Fix unloop('').
  - Regularize event constructors.
  - Doc fixes catalyzed by perl@jochen-stenzel.de.
  - Fix broken t/eval.t.
  - More conservative refcnt policy for watchers.
  - Rename 'mom' to 'w'.
  - Deprecate the accessing of watcher attributes via events.
  - Fix PL_ macros again.  (*Argh!*)
  - Redesign typemap.
  - Make 'e_cbtime' available from perl.  Added test.
  - Return unloop_all($arg) from loop().
  - Make compile with threaded perl.
  - Documentation updates and reorganization.
  - Switch inactivity from e_interval to e_timeout.
  - Plug typemap holes.  Implementation still needs improvement.
  - Make io timeout respect e_repeat. (gisle@aas.no).

0.33 1999-03-30
  - Store 'e_max_cb_tm' and hand-off to stats engine.
  - Morph events into watchers when they go out of scope.
  - Refine typemap fix.

0.32 1999-03-15
  - Apply typemap fix as per gsar@activestate.com.

0.31 1999-03-09
  - Simplified typemap in an attempt to squash bugs.
  - Search @INC in Event::MakeMaker.
  - Make io timers non-repeating.
  - Relax io watchers more.  Now they can be started with either e_fd
    or e_timeout or both or neither.
  - Allow creation of io watchers without a valid e_fd.  Should
    probably do the same with the rest of the watcher types.
  - Added e_suspend.  Do we keep the 'suspend' and 'resume' methods?
  - Fixed refcnt problem triggered by deleting hash entries.  Added test.
  - Avoid qr// in keys.t.
  - Reroute C-API croak through Carp::croak. Added test.
  - Removed backward compatibility up to 0.24 (inclusive).
  - Factored event init/dtor code.
  - Added 'callback' hooks.  Updated docs.

0.30 1999-02-04
  - Constructors now populate the watcher hash with any extra
    unrecognized key-value pairs.
  - Better typemap diagnostics.  There seem to be bugs lurking but I
    can't construct a short test case...
  - Arrange for events to have the same careful typemap treatment as
    do watchers.  Added test.
  - Unplugged Time::Virtual (suggested mainly by Sarathy, Nick and
    Graham).  See patches/time-hires.
  - Added AllWatchers & typemap functions to the public API.
  - Moved Event::Stats to a separate tarball.  I wonder if the stats
    API is too restrictive?
  - Split Event::Watcher code into a separate file for intuitive
    importing via 'use' (suggested by artur@vogon-solutions.com).
  - Fix SEGV in keys %$event (ophir@internap.com &
    artur@vogon-solutions.com).  Added test.
  - Tweaks to diagnostics.
  - Event destruction now must be managed explicitly.  Use the
    'stop' method to disable a watcher and use 'cancel' to destroy a
    watcher.  This change is an unfortunate necessity because of
    optimizations in perl's reference counting.
  - Suspend & cancel now abort any pending events.  In previous
    releases, the callback could be called after cancel in some cases.
  - Replaced newHVhv.  Some versions of perl have a buggy implementation.
  - Fixed a few minor problems detected by Insure++.
  - Fixed typos in croak messages.
  - Generous updates to Event.pod.
  - Renamed (lots of) keys.
  - Added 'use_keys' method.  Comments?
  - (Try to) remove support for leading dashes (any left?).
  - Factor out code to Time::Virtual (& Time::HiRes).
  - Moved process.pm to demo directory.
  - Added discussion of event loops vs. threads (kudos to Mark.Mielke.markm@nt.com & fellow loop'ers).

0.25 1998-11-20
  - Move SAVETMPS/FREETMPS up to Event::_loop.
  - Split ACTIVE flag into two separate flags, ACTIVE & POLLING, to
    more accurately track watcher state.
  - Tweaks to io 'events' mask handling.
  - Move @EXPORT_OK list for qw(R W E T) to Event::Watcher.
  - Fix SEGV in all_watchers/all_idle if zero watchers present.
  - Nuke obsolete backward support code in io.c.  (Poll constants to
    set event mask.)
  - Rename event_vtbl->watcher.

0.24 1998-11-01
  - Reengineered hash-object implementation to use HVs.
  - Removed tailpoll support from io watchers.  File::Tail can just
    hook up with timers & a generic "block on read" API can be built
    on top of that if need arises.  Nick, this okay?
  - Split watchers & events into separate C structures.
  - Fixed misbehaving autoload code.
  - END hook to cancel all events before global destruction.  Still
    needs more work to be sure the memory is freed.
  - Fixed bad arithmetic in io timeout code.  Renamed interval
    epsilon and added better diagnostics (see Event::_timeout_too_early).
  - Timers cancelled within their callback weren't!  Fixed.
  - Fix typo in c/queue.c (thanks again Jan!).

0.23 1998-10-18
  - Unified Perl & C API for hooks.
  - Niggles for Win32 (thanks Jan!).

0.22 1998-10-17
  - Include copy of self-tie patch with dist.
  - Documentation for Event::MakeMaker.
  - tailpoll revamped.
  - MIN_PERL_DEFINE clean.
  - Improve sv_2interval typemap.

0.21 1998-10-08
  - Try to normalize benchmark.
  - 'now' was broken for inactive events.  Fixed & added test.
  - Fixed infinite loop triggered by botched exception recovery.

0.20 1998-10-06
  - Rename inactive to inactivity.
  - Use new self-tie to wrap watcher objects.  Leak test now passes
    for almost all watcher types.
  - Slight improvements to gettimeofday.c.
  - Merged c_callback & perl_callback into one pointer slot.
  - Added timeable methods to the public API.
  - Var watchers weren't working very well.  Set up events/got mask
    stuff mirroring io watchers.  Added tests.  Does anyone really
    want timeouts?

0.19 1998-10-03
  - Put in code for tailpoll, but I can't tell if it works because
    the tail-f behavior seems to already happen automatically.
    Comments?  See demo/tail.t.
  - Added inactive watchers (& tests).
  - timeable API changes to allow multiple timers per watcher.
  - Fixed non-ANSI function declaration in c/signal.c
  - Reduced C API.
  - Renamed queueEvent to queue.
  - Fixed refcnt problem in DELETE.
  - Dashes depreciated.
  - Added -timeout to process watchers.  Might be useful.
  - Fixed typo induced polling in one_event.
  - Moved R/W/E/T flags to Event from Event::io.
  - Changed the arguments to $Event::DIED in order to cope with
    $Event::DIED dieing.  Fixed sweep exception handling.  Added test.
  - Watchers created with the C API now should require their perl
    support code automatically.  Moved idle support to an autoloaded
  - Added sweep & loop($timeout).  Fixed sleep.

0.18 1998-10-02
  - Added non-working export_fail prototype.  Suggestions welcome.
  - Re-factored start & stop better (the preCB method is gone).
  - Finished up sleep().  Added tests.
  - Tied watchers are now based on pe_tmevent.
  - Applied patch from Gisle for pe_sys_sleep/select.
  - Renamed 'watchvar' to 'var'.
  - Added tests for 'now' (it was completely broken!).
  - Tweaked debug levels.  Someone needs to inventory debug warnings
    are decide what is reasonable.

0.17 1998-10-01
  - Added tests.
  - Added sleep().  Truly a tortuous experience.
  - Added min/max interval to idle watchers.
  - c_callback no longer prevents watcher destruction.  Use refcnt!
  - Renamed various C APIs for accuracy (unix_io.c).
  - Reentrant flag added.
  - loop() now terminates when there are no active watchers left.
  - Depreciated async.  Moved -priority to -nice & added warning.
  - Fixed minor stuff.
  - Make io timeouts work for real.  (Thanks Gisle :-)

0.15 1998-09-29
  - Improved the EventAPI.h version check.
  - Added -timeout for io watchers.  Untested.
  - Fix io watcher bugs (unix_io.c) pointed out by Gisle Aas.

0.14 1998-09-28
  - Gutted the internals to make events reentrant.
  - The $Now cache was not being refreshed and that was causing
    timers to be queued, well, never.
  - Fixed whitespace in Event.pod.  (Gisle Aas again.)
  - Optimized priority queue (hopefully).  I probably should have
    benchmarked it first.  :-P

0.13 1998-09-27
  - Moved and renamed lots of methods.  Yet, added code to ease backward compatibility.  (perl-loop suggested)
  - Integrated rewritten documentation contributed by Gisle Aas. Thanks!

0.12 1998-09-22
  - Event::MakeMaker & event_api.h.
  - Add 'total' stat.  Moved stat methods to Event::Loop package.

0.11 1998-09-11
  - Regularized exception handling.  $Event::DIED is now called if an event dies.
  - Allow timer interval to be a scalar ref.
  - Tweaked queue scheduling.  Events with priorities <= 5 now starve if higher priority events are available.
  - Changing the event priority of a queued event will now reque the event into the appropriate queue.
  - Split NetServer::ProcessTop into it's own tarball.

0.10 1998-09-03
  - Fixed serious bug in the event destructor.
  - Accept file descriptor numbers in addition to filehandles, etc.
    Maybe non-portable, but nice for unixen.
  - Fixed repeating timer logic.  Tweaks for idle events.
  - Completed re-write of almost everything in C.
  - NetServer::ProcessTop implements a top-style control panel
    available via telnet ($$ % 7000).  See util/top.pl.
  - Proxy "tied" event type that allows completely perl-side
    implementation of new event handlers.
  - Event::process is working again (implementation is entirely in perl).
  - Eliminated race conditions in signal handling.
  - Zombie events are still not being collected due to mysterious
    refcnt problems.  It is possible that this cannot be resolved
    until perl implements tied hashes properly (through a VTBL!).
  - Removed $Event::*::Count since it is superceded by

0.08 1998-08-14
  - C rewrite

0.07 1998-08-05
  - test callbacks
  - doc queuing
  - remove debugging

0.06 1998-07-24
  - parameterise register method so 'check' not fixed method but callback
  - doc priorities
  - doc callbacks

0.05 1998-07-20
  - queuing, waiting
  - Status method

0.04 1998-07-20
  - registering forces class to be child of Event
  - add waitForEvents

0.03 1998-07-18
  - Event::watchvar now passes $ref to callback
  - added use attrs qw(locked method) to all subs, and code to
    ensure this will work on pre-threaded perl
  - Event.xs Fixed so will compile with threaded perl
  - Event::timer Added -hard option
  - Tweaks to keys dispatchAsyncEvents()
  - added Time::HiRes
  - moved idle events into sub-module
  - Makefile.PL removed INSTALLDIRS => 'perl'

0.01 1998-01-26
  - Initial