ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.520
Committed: Thu Aug 20 21:38:51 2009 UTC (14 years, 9 months ago) by root
Branch: MAIN
Changes since 1.519: +2 -0 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 Revision history for Perl extension Coro.
2
3 TODO: should explore PerlIO::coroaio (perl leaks like hell).
4 TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7 TODO: unready_all
8 TODO: myhttpd header parsing
9 TODO: channel->maxsize(newsize)?
10
11 - work around a bug in the perl debugger causing crashes
12 when running under the debugger by marking _pool_handler as nodebug.
13 - speed up Coro::async considerably.
14
15 5.162 Tue Jul 28 04:04:03 CEST 2009
16 - perl 5.8.2 is now minimum requirement.
17 - skip t/19_handle.t on broken windows perls.
18
19 5.161 Wed Jul 22 04:47:38 CEST 2009
20 - Coro::AnyEvent::poll could have a different prototype when EV was
21 used as backend (analyzed by Tatsuhiko Miyagawa).
22 - Coro::AnyEvent errornously initialised the event loop when loaded,
23 not on demand.
24 - try to workaround rare */t/01_unblock.t failures.
25
26 5.16 Tue Jul 21 01:44:37 CEST 2009
27 - Coro::AnyEvent failed to hook into the event loop
28 when no threads had been readied between detecting
29 the event loop and actually running it.
30 - considerably speed up Coro::Select by taking avdantage
31 of AnyEvent > 4.8 and some other optimisations.
32 - implement paragraph readline mode in Coro::Handle
33 (based on patches by Zsbán Ambrus).
34 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
35 and analyzed by Yasuhiro MATSUMOTO).
36 - clarified libcoro license and copyright.
37 - someone stole my EXTRA_META!!!
38 - implement Coro::Select::patch_pp_sselect and it's brother,
39 for hardcode select overriding.
40
41 5.151 Mon Jul 6 05:41:57 CEST 2009
42 - backport to windows process emulation code again (patch by
43 Yasuhiro MATSUMOTO).
44 - slightly update Coro::MakeMaker.
45
46 5.15 Tue Jun 30 10:28:06 CEST 2009
47 - deprecate Coro::Socket, document how to get ipv6 support via
48 AnyEvent::Socket instead.
49 - implement signal->wait ($cb) interface, similar to semaphores.
50 - work around SvOK not supporting getmagic, so we have to getmagic
51 to test for undef :/ (reported by Matthias Waldorf).
52 - load Coro::AnyEvent in all modules using AnyEvent.
53 - work around perl corrupting our internal data structures,
54 reported by Tokuhiro Matsuno.
55 - enable per-coroutine real and cpu time gathering
56 (Coro::State::enable_times).
57
58 5.14 Wed Jun 24 01:37:48 CEST 2009
59 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
60 as this seems to have been a difficult concept (poll, sleep, idle,
61 idle_upto).
62 - add Coro::AnyEvent::readable/writable functions.
63 - clarify Coro::EV/Event/AnyEvent manpages.
64 - free per-thread global scalars in the thread calling ->cancel, to
65 avoid crashes when $_, $@ etc., are magical but some of those
66 had already been freed.
67 - "unexperimentalise" the callback interface for Coro::Semaphore.
68 - speed up ready queue management/context switching by using a linked
69 list instead of an array (~5-10%).
70 - implement "watch" command in Coro::Debug shells.
71 - for fun, implement time-slicing as an example in the manpage.
72 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
73 Coro::Event, use the normal AnyEvent handling instead of dieing
74 (the same is true for Coro::Handle).
75 - properly document Coro::EV::timed_io_once.
76 - avoid unneccessary ->cancel calls in Coro::Handle.
77 - maybe make it work on mingw32 with win32 backend
78 (based on patch by Yasuhiro Matsumoto).
79
80 5.132 Fri May 29 09:00:39 CEST 2009
81 - do not keep a reference to the argument itself in
82 Coro::Semaphore::guard, as it could change later.
83 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
84 should support a prepare callback.
85
86 5.131 Mon Mar 16 23:20:37 CET 2009
87 - implement and document Coro->suspend, Coro->resume.
88 - fix Coro::Select implementation to not (often) close
89 the passed file descriptors (testcase provided by pippijn).
90
91 5.13 Mon Dec 15 21:51:42 CET 2008
92 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
93 - don't set diehook to C<undef> but instead to NULL, to avoid
94 spurious warnings.
95 - fix a lot of bugs in Coro::SemaphoreSet.
96 - Coro::SemaphoreSet will less often create a semaphore needlessly.
97 - add Coro::SemaphoreSet::count and wait methods.
98 - take advantage of the new Guard module.
99 - deprecate Coro::guard.
100 - try to fix the dreaded 01_unblock tests once more. I hate it when
101 testsuites need more fixing than the code they are supposed to test.
102 - croak in more cases when a required callback isn't resolvable.
103 - fix some minor issues in Coro::State->call/eval.
104 - use current coroutine context instead of a temporary one
105 when temporarily switching to another coroutine.
106 - do not call C-level on_destroy handlers during global destruction,
107 to avoid needless segfaults.
108
109 5.12 Sun Dec 7 13:30:38 CET 2008
110 - add default config for MirOS, which seems to be a bug-to-bug
111 compatible fork of openbsd ("world domination by releasing
112 openbsd cvs before the openbsd folks do it" or so :).
113 - free_padlist did destroy the names pad, not good, but didn't
114 seem to bother perl - this could fix issues such as eval ""
115 inside a function called from multiple coroutines.
116 - use a different method to detect destruction time.
117 - be more careful when freeing padlists just before global
118 destruction.
119 - fixed and expanded the call/cc example.
120 - renamed _terminate to _coro_run.
121 - new method Coro::Channel->shutdown.
122 - try pthreads on openbsd <4.4 (broken sigaltstack, will
123 pthreads fare better?).
124 - be less picky about destroying "a" running coroutine.
125
126 5.11 Tue Nov 25 21:49:05 CET 2008
127 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
128 working code with failed assertions, introducing more bugs than
129 it fixes, requiring elaborate workarounds :(
130
131 5.1 Mon Nov 24 08:54:59 CET 2008
132 - wrote a small introductory tutorial - Coro::Intro.
133 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
134 - Coro::Select did errornously dup the file descriptors
135 and didn't work with all AnyEvent backends.
136 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
137 LWP data transfers.
138 - disassociate c contexts from coro objects - this is agruably more
139 correct, but mostly allows sharing of cctxs between coro and state
140 objects, for added memory savings and speed increases.
141 - bumped $Coro::POOL_RSS up to 32kb by default.
142 - no longer set the optype to OP_CUSTOM, as B::* understandably
143 doesn't like this very much (and we *are* a type of entersub).
144 - implement state cloning, just to prove that call/cc can be done.
145 - automatically load Coro::AnyEvent in Coro::Handle.
146 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
147 do not have this method (and don't need it either).
148 - speed up generic anyevent methods in Coro::Handle by using rouse
149 callbacks.
150 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
151 others. It also makes the debugger happier, as you can trace
152 through the idle threads now.
153 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
154 - eg/event was pretty broken.
155 - better 5.8.6 compatibility.
156
157 5.0 Thu Nov 20 10:35:05 CET 2008
158 - NEW ARCHITECTURE: use the latest 4.x version if you experience
159 stability issues.
160 - bump API version to 7 - all dependents must be recompiled.
161 - removed timed_* functions - they were not being used anyways
162 and should be replaced by a more generic mechanism -
163 and were annoying to support anyways :)
164 - removed SemaphoreSet's waiter method - use sem method instead.
165 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
166 - async_pool did free a scalar value twice
167 ("Attempt to unreference...").
168 - fix a longstanding bug where calling terminate on a coro that
169 was waiting for a semaphore that was just becoming available
170 would cause a deadlock (semaphore would get into a state where
171 it was available but waiters were still blocked).
172 - calling throw on a coroutine that is waiting for a semaphore will
173 no longer make it acquire the semaphore (and thus leak a count).
174 - perl's process emulation is now not even theoretically supported
175 anymore.
176 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
177 conversion of callback-style to blocking-style.
178 - new methods $coro->schedule_to and ->cede_to, to specifically
179 schedule or cede to a specific coroutine.
180 - new function Coro::Semaphore::wait.
181 - use named constants in Coro::Channel (Richard Hundt).
182 - directly patch the entersub opcode calling SLF functions (cede,
183 transfer and so on). this does speed up context switching, but
184 more importanly, it frees us from the hardcoded behaviour of
185 entersub, so we might actually be able to return something from
186 those functions and atcually create new ones.
187 - take advantage of __builtin_frame_address on gcc.
188 - expose THX in coroapi (not sure whether this was a wise decision,
189 as "threaded" perls are running at half speed anyways).
190 - implement execute_slf (schedule-like-function) interface that makes
191 it possible to implement schedule-like-functions in XS.
192 - use new SLF interface to massively speed up Coro::EV by roughly a
193 factor of two.
194 - used new SLF interface to massively speed up Coro::Semaphore by a
195 factor of three.
196 - used new SLF interface to speed up Coro::AIO by roughly a factor of
197 four and reduce its memory usage considerably.
198 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
199 for a nice speedup and vastly more correct behaviour. Also implement
200 a new method "sem" to get at the underlying semaphore object.
201 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
202 (in comparison) 20-40% speedup.
203 - used new SLF interface to reimplement Coro::Signal gaining
204 some unknown (because I was too lazy), but certain, speedup, and also
205 making signals reliable for the first time.
206 - used new SLF interface and other optimisations to speed up async_pool
207 by a factor of two. It also doesn't rely on perl's exception mechanism
208 to exit anymore. The overhead for terminating an async_pool, coro over
209 a normal async is now very small.
210 - sped up coroutine creation/destruction by 40%.
211 - forgot to include Coro/libcoro/README in the dist for all these years.
212 - work around a freebsd pthreads bug (manual testcancel is required as
213 pthread_cond_wait isn't a cancellation point on freebsd).
214 - use new rouse functions to speed up and simplify Coro::BDB.
215 - make "prefer perl native functions" work with threaded perls.
216 - condense Coro::Debug ps output, hint at v and w flags.
217 - (libcoro) lots of minor cleanups and portability improvements.
218
219 4.914 Wed Nov 19 12:54:18 CET 2008
220 - fix a disastrous bug in the readline optimisation
221 introduced in 4.801.
222
223 4.913 Sat Nov 15 07:58:28 CET 2008
224 - async_pool did free a scalar value twice
225 ("Attempt to unreference...").
226
227 4.912 Thu Nov 13 18:31:23 CET 2008
228 - minor cleanups.
229 - use much larger stacks on linux and perl < 5.8.8.
230 - Coro::Debug::new_unix_server did not unlink the socket
231 when destroyed.
232
233 4.911 Tue Nov 11 04:26:17 CET 2008
234 - "port" to threaded perls.
235
236 4.91 Mon Nov 10 05:36:38 CET 2008
237 - the ->throw exception object no longer leaks.
238 - creating a new cctx leaked a scopestack entry (memleak).
239 - new coroutines didn't get created with a zero flags field
240 (unknown impact).
241 - calling ->throw on a not-yet-started coroutine should now work
242 instead of being ignored.
243 - ->throw is now supported on Coro::State objects.
244 - clean up cctx creation code a bit.
245 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
246
247 4.9 Sat Nov 8 17:45:27 CET 2008
248 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
249 - (libcoro) no longer leak threads in the experimental pthread backend,
250 also speed it up considerably.
251 - (libcoro) do not rely on makecontext passing void *'s unscathed.
252 - fix compiletime dependencies on libcoro in the Makefile.
253 - cctx_count wasn't always updated properly.
254 - Coro::State::cctx_stacksize wasn't applied correctly.
255 - new function Coro::State::cctx_max_idle.
256 - the default max number of idle C contexts is now 4.
257 - (libcoro) try harder to get _setjmp/_longjmp.
258 - (libcoro) cleanup and extend the libcoro API to officially
259 allow the creation of empty source contexts.
260 - very experimental workaround for the totally broken netbsd platform.
261 - tried to hack around openbsd bugs.
262
263 4.804 Wed Nov 5 16:36:00 CET 2008
264 - Coro::Debug::new_unix_server would not create a non-blocking listening
265 socket, sometimes causing freezes.
266 - (libcoro) fix misaligned stack points for setjmp and assembly
267 methods, which can cause crashes on x86/x86_64 with a sufficiently
268 aggressive compiler.
269 - new function: Coro::Debug::new_tcp_server.
270 - move ->throw into the Coro class because it only works on coro objects.
271
272 4.803 Mon Nov 3 17:16:12 CET 2008
273 - (libcoro) use a global asm statement to become independent of gcc
274 otpimisations for CORO_ASM (thanks to pippijn for the idea).
275 - try to workaround yet another broken bsd, this time dragonfly.
276
277 4.802 Thu Oct 30 10:56:12 CET 2008
278 - support -fno-omit-frame-pointer on x86 with the assembly method.
279 - tune 01_unblock.t tests a bit.
280
281 4.801 Wed Oct 22 18:33:37 CEST 2008
282 - improve readline speed for very long "lines".
283 - backport to 5.8.8.
284
285 4.8 Thu Oct 2 13:34:40 CEST 2008
286 - new function Coro::AIO::aio_wait.
287 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
288 - greatly speed up and reduce memory usage of Coro::AIO requests.
289 - implement some other µ-optimisations.
290
291 4.749 Mon Sep 29 14:40:12 CEST 2008
292 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
293
294 4.748 Sat Sep 27 14:03:19 CEST 2008
295 - implement, but do not document, PerlIO::cede(granularity).
296 - Coro::Storable forgot to wrap Storable::pstore.
297 - work around the multitude of leaks and memory corruption
298 bugs in PerlIO::via by using our own C-level perliol. As a side
299 effect, Coro::Storable is now much, much, much faster.
300
301 4.747 Tue Sep 23 01:59:41 CEST 2008
302 - fix a per-cv memleak (one empty array was leaked per
303 code reference).
304 - avoid a crash in coro->call|rss when the coroutine was already
305 destroyed (most noticably when using Coro::Debug::ps :)
306 - also protect *Storable::FILE.
307 - push up default storable granularity to 20ms.
308
309 4.746 Sun Sep 21 03:22:20 CEST 2008
310 - be more insistent on locking Storable against reentrancy
311 in Coro::Storable.
312 - move swap_def?v and throw to Coro::State, as documented.
313
314 4.745 Thu Jul 24 00:14:38 CEST 2008
315 - remove debugging code related to MgPV_nolen_const, also try to
316 make it compile with perl 5.8.6 (yes, apple apparently loves
317 outdated software). Reported by John S.
318
319 4.744 Tue Jul 8 22:06:35 CEST 2008
320 - correctly provide default DIE/WARN handlers as documented.
321 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
322 implementations inside perl work fine for us.
323
324 4.743 Mon Jun 16 00:21:57 CEST 2008
325 - when using Coro::EV without running EV::loop it could
326 result in busy-waiting for events, this has been fixed.
327 - reduce codesize and improve performance by using EV_DEFAULT_UC.
328
329 4.742 Sat May 31 14:10:21 CEST 2008
330 - implement a workaround for (some) perl <5.8.8 versions.
331 - require EV 3.3+.
332
333 4.741 Fri May 30 23:33:09 CEST 2008
334 - tell netbsd how utterly broken their imitation of an OS is
335 and refuse to build by default if pthreads are in use.
336 - switch to "s" method on all bsd's by default, as their ucontext
337 stuff seems just too broken.
338 - fix a bug in Coro::Select.
339
340 4.74 Thu May 29 20:05:31 CEST 2008
341 - do not test Coro::LWP for lack of dependencies.
342 - do not test Coro::Select for lack of working perls.
343
344 4.73 Thu May 29 2008
345 - fix a bug in Coro::EV which would cause it to block despite
346 there being runnable coroutines.
347 - sprinkle "no warnings" freely over everything, also suppress
348 warnings for some other modules.
349 - fix typo in WSAEWOULDBLOCK.
350
351 4.72 Sun May 25 05:14:36 CEST 2008
352 - tweak META.yaml a bit, unfortunately, there is no documented way
353 to have optional dependencies with CPAN. doh :(
354 - avoid running some tests on windows because they would fail due to
355 perl bug (broken fork, broken pipes...).
356 - work around perl on windows bugs where perl returns undocumented
357 error codes for sysread, syswrite etc. by taking advantage
358 of AnyEvent's workaround for that problem.
359 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
360 a common perl implementation bug on windows.
361 - use unix domain sockets in testsuite to work around
362 common perl implementation bugs on widows (they are emulated by
363 tcp sockets on windows. ugh.)
364
365 4.71 Sat May 24 20:01:27 CEST 2008
366 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
367 - take advantage of async name resolution of AnyEvent::Util.
368 - work around brutal inet_aton override in Coro::LWP.
369 - take advantage of the readyhook in Coro::EV, for smoother
370 scheduling.
371
372 4.7 Sun May 11 00:32:19 CEST 2008
373 - completely reworked the Coro manpage.
374 - added Coro::AnyEvent, generic event loop integration.
375 - implement cancel, ready and kill commands in Coro::Debug.
376 - document find_coro in Coro::Debug.
377 - incompatible change: rename has_stack to has_cctx.
378 - Coro::AIO and Coro::BDB no longer force event model detection,
379 use AnyEvent::AIO and AnyEvent::BDB.
380
381 4.6 Sat Apr 26 10:05:14 CEST 2008
382 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
383 by the Coro module. It was a mistake to have it in the first place.
384 - (experimental) support for activestate perl 5.10 (method "w").
385 - (experimental) support for strawberry perl (method "a").
386 - coro_sigelem_set did not return a value although it had to,
387 actual impact unknown.
388
389 4.51 Mon Apr 14 13:28:27 CEST 2008
390 - make it compile again on 5.8.
391
392 4.50 Thu Apr 10 09:43:17 CEST 2008
393 - I did it twice! (see 4.49).
394
395 4.49 Sun Apr 6 21:23:31 CEST 2008
396 - grr, instead of compiling the recent changes on 5.10 only they
397 were compiled on 5.8 only.
398
399 4.48 Sun Apr 6 20:36:46 CEST 2008
400 - allow coroutine switches during eval's under 5.10.x, as it seems
401 the parser is a per-interpreter option now, so this is safe
402 (this might fix the odd crashing bug).
403 - drop support for 5.9.x versions: they are dead, jim.
404
405 4.47 Sun Apr 6 00:37:52 CEST 2008
406 - force cctx allocation on API calls: we know we need to force one
407 and gcc actually manages to confuse our heuristic nowadays,
408 leading to crashes and worse.
409 - document force_cctx.
410
411 4.46 Fri Apr 4 22:05:43 CEST 2008
412 - upgrade libcoro, resulting in pthread-backend (which was only created
413 to fulfill the rules of the programming languages shootout).
414
415 4.45 Thu Mar 13 11:55:36 CET 2008
416 - fix a file leaking bug in eg/mhyttpd that would allow
417 downloading of any file (reported by oesi).
418 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
419 (also add testcase).
420 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
421 apparently having loads of fun with that).
422 - implement $state->swap_defsv, ->swap_defav and document ->throw.
423
424 4.4 Wed Feb 13 16:44:29 CET 2008
425 - only assume .cfi pseudo isns exist on GNU/Linux.
426 - add get_timed method to Coro::Channels.
427 - fixed Coro::Debug manpage.
428 - perl 5.11 compatbility improvement based on patch by
429 Andreas König.
430
431 4.37 Sun Jan 20 11:25:23 CET 2008
432 - soften the check for an active parser for perl 5.10 (avoiding
433 "Coro::State::transfer called while parsing" in many cases).
434
435 4.36 Sun Jan 13 10:53:56 CET 2008
436 - reset diehook when terminating from an async_pool as to not
437 trigger any __DIE__ handlers.
438
439 4.35 Sun Jan 13 04:14:13 CET 2008
440 - "bt" debug command now displays any exceptions
441 from longmess and also skips the innermost
442 stackframes, giving more useufl output.
443 - allow backtraces in coroutines blocked in Coro::EV,
444 at a <1% speed hit.
445 - handle localising of $SIG{__WARN__} and __DIE__
446 properly (with a proper amount of dirty hacking).
447
448 4.34 Sat Dec 22 17:49:53 CET 2007
449 - upgrade to EV version 2.0 API.
450
451 4.33 Mon Dec 17 08:36:12 CET 2007
452 - make Coro::AIO etc. loadable in the absence of EV.
453
454 4.32 Mon Dec 17 07:46:02 CET 2007
455 - majorly improved Coro::Handle's performance with Coro::EV.
456 - implemented the until now mythical Coro::BDB module.
457 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
458
459 4.31 Wed Dec 5 12:32:39 CET 2007
460 - remove warn statement form Coro::Util.
461
462 4.3 Tue Dec 4 20:33:14 CET 2007
463 - calls to the idle function could cause stack corruption
464 when the stack changed.
465 - do no longer rely on the presence of EV::DNS if EV is used
466 (because it is gone), but instead take avdantage of EV::ADNS
467 if available.
468 - add ($) prototypes to all functions in Coro::Storable.
469 - use a conventional (safer) idle callback in Coro::EV.
470 - do accept quickly in Coro::Debug to avoid endless loops.
471
472 4.22 Fri Nov 30 16:04:04 CET 2007
473 - really use optimised versions for Event and EV in Coro::Util
474 and Coro::Handle.
475
476 4.21 Sun Nov 25 10:48:59 CET 2007
477 - fix a spurious memory read.
478 - Coro::EV no longer keeps the eventloop "alive".
479
480 4.2 Fri Nov 9 20:47:05 CET 2007
481 - enable/disable tracing from a new coroutine, not a pooled one.
482 - fix a memleak in Coro::Event.
483 - removed killall call from fork_eval.
484 - made sure store_fd is already loaded so that fork_eval
485 does not have to parse autoload in each subprocess.
486 - only use assembly method if -O switch is in $Config{optimize}.
487 - add (optional) Coro::EV module, so far the best event loop module
488 directly supported by Coro.
489 - if the event model is EV, use EV::DNS to resolve
490 stuff in Coro::Util.
491 - don't get confused by multiple event notifications in Coro::Handle.
492 - initial support for EV (libevent interface).
493 - require Event and EV using configure_requires, to force their existance.
494
495 4.13 Wed Oct 24 07:26:45 CEST 2007
496 - add Coro::Storable::blocking_thaw.
497 - use a vastly more complicated technique to localise
498 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
499 - use a coroutine for the idle callback Coro::Event,
500 instead of running Event in the current coroutine context.
501 This also catches recursive invocations.
502 - actually report fork errors in gethostbyname and inet_aton.
503
504 4.11 Thu Oct 11 02:40:24 CEST 2007
505 - port to threaded perls.
506
507 4.1 Thu Oct 11 02:38:16 CEST 2007
508 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
509 be local to each coro (see Coro::State).
510 - incompatible change: for very deep reasons, cede and cede_notself
511 cannot return anything, so nothing will be returned.
512 - possibly bring back 5.10 compatibility (untested).
513 - work around stupid (and wrong) warning on 5.10 :(.
514 - overlay the saved state over the context stack. This saves
515 a few hundred bytes per coroutine on average and also
516 speeds up context switching a bit.
517 - further tune default stack sizes.
518 - (more) correctly calculate stack usage in coro_rss.
519 - Coro::Storable::blocking_* did not properly lock
520 resulting in races between coroutines.
521 - added Coro::Storable::guard.
522 - stopping to trace a coroutine could destroy the cctx of
523 an unrelated coroutine.
524 - explain the relationship between Perl and C coroutines in
525 more detail in Coro::State.
526 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
527 causing a fork per resolve. This also affected Coro::Socket.
528 - switch to a separate stack in $coro->call/eval to avoid
529 invalidating pointers.
530
531 4.03 Sat Oct 6 21:24:00 CEST 2007
532 - added Coro::throw method.
533 - minor code cleanups.
534
535 4.02 Sat Oct 6 02:36:47 CEST 2007
536 - fix a very minor per-coroutine memleak (a single codereference).
537 - fixed a bug where the currently in-use c context would be freed
538 prematurely (can happen only when programs change the stacksize
539 or use tracing).
540 - tracing can no longer keep a coro alive after it terminated.
541 - do static branch prediction in the common path for gcc. gives
542 about 2-5% speed improvement here.
543
544 4.01 Fri Oct 5 22:10:49 CEST 2007
545 - instead of recreating *a* standard output handle we simply
546 use STDOUT, which is faster and hopefully more robust.
547
548 4.0 Fri Oct 5 12:56:00 CEST 2007
549 - incompatibly changed Coro::Storable::freeze.
550 - major new feature: added Coro::Debug, for interactive coroutine
551 debugging, tracing and much more.
552 - major bug fix: unbelievable, but true: $_, $/ and many other
553 "saved" variables actually weren't being saved. This has been fixed,
554 of course, while increasing performance while losing all the save
555 flags.
556 - save flags are gone, and all the api functions dealing with them.
557 - added Coro::Semaphore::adjust.
558 - added Coro::Util::fork_eval.
559 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
560 - added Coro::killall.
561 - reduce initial stack sizes to allow for "micro-coroutines".
562 - better async_pool resource management, moved parts of async_pool
563 handling to XS (major speed improvement).
564 - actually croak before modifying important data structures.
565 - refuse to transfer while compiling.
566 - possibly support eval EXPR better now.
567 - enable assembly per default on linux+bsd x86+amd64.
568 - all internal members were renamed _something for easier subclassing.
569 - many minor tweaks.
570
571 3.63 Wed May 16 14:10:06 CEST 2007
572 - implement handcoded assembly for x86/amd64 SVR ABI.
573
574 3.62 Fri Apr 27 21:36:06 CEST 2007
575 - upgrade libcoro (which might set unwind info correctly).
576 - change default on linux to setjmp/longjmp.
577
578 3.61 Thu Apr 19 12:36:18 CEST 2007
579 - Coro::Storable caused an endless loop when thawing invalid
580 pst's sometimes.
581 - use a Semaphore in Coro::Storable, as Storable doesn't
582 seem to be reentrant (although it is documented to
583 be threadsafe...).
584 - fix Coro::Signal to bring back the original unreliable
585 but stateful semantics.
586 - fixed a lot of typos in Coro.pm (patch submitted by David
587 Steinbrunner, which applied flawlessly).
588
589 3.6 Sat Apr 14 17:13:31 CEST 2007
590 - added some bugfixes to get eg/myhttpd working again.
591 - added Coro::Storable for often-cede'ing freeze/thaw.
592 - try to do a clean exit when a coroutine calls exit
593 (EXPERIMENTAL).
594 - got rid of indirect call through _coro_init.
595 - updated the partly antique examples in eg/ to
596 work again and be a bit less magic, too.
597 - fixed Coro::Signal semantics to work as documented again.
598
599 3.55 Sun Mar 25 01:20:47 CET 2007
600 - add SAVE_DEFFH to save the default filehandle and enable
601 it by default.
602 - finally move socket-operations from Coro::Socket to Coro::Handle
603 to be able to unblock foreign sockets.
604 - add Coro::State::save_also and guarded_save.
605 - add count accessor to Coro::Semaphore.
606 - add Coro::State::cctx_stacksize.
607 - just for the fun of it, do not rely on implicit context,
608 which can dramatically improve performance, but people
609 using windows-process-emulation perls do not care much about
610 performance.
611
612 3.51 Sun Mar 4 14:18:04 CET 2007
613 - fixed a problem when you weakened references to Coro::State's
614 (patch by Yuval Kogman).
615
616 3.501 Wed Feb 28 12:44:07 CET 2007
617 - rename some global symbols as macosx from hell redefines
618 them without asking.
619
620 3.5 Tue Feb 13 20:22:53 CET 2007
621 - do AnyEvent model detection earlier, avoiding problems
622 caused by first using AnyEvent and later Coro::Event.
623 - implement and document Coro::Event event objects.
624 - fix a potential problem in Coro::Event causing crashes.
625 - initialise PL_comppad when creating a new coroutine,
626 avoids crashes on early coro destruction.
627
628 3.41 Mon Jan 22 19:19:49 CET 2007
629 - readline on Coro::Handle did not support undefined $/,
630 nor did it deliver partial lines on EOF or error.
631 - implement malloc fallback for stack allocation because
632 stupid broken idiotic OSX has a stupid broken
633 idiotic fits-the-whole-os mmap "implementation" and
634 my dick feels longer if Coro is portable even to
635 obsolete platforms.
636
637 3.4 Fri Jan 19 21:52:54 CET 2007
638 - remove t/09_timer.t, as it isn't really testing much
639 but was rather flaky in practise.
640 - async_pool coro would keep arguments and callback alive until
641 it was reused.
642 - cancellation of a coroutine could cause spurious idle calls
643 in cede_notself.
644
645 3.3 Sat Jan 6 03:45:00 CET 2007
646 - implement $coro->on_destroy.
647 - Coro::Event blocking semantics have been changed,
648 documented and - hopefully - improved.
649 - fix nice adding, not subtracting, from priority.
650 - fix ->prio and api_is_ready (patch by Mark Hinds).
651 - fixed an assert ("... == prev__cctx->idle_te")
652 that could errronously trigger.
653 - fix various large and small memleaks.
654 - use a (hopefully) more stable cancel implementation
655 that immediately frees the coroutine data.
656 - cede/cede_notself return a status now.
657 - added Coro::guard function.
658 - added a global coroutine pool for jobs (on my machine,
659 I can create and execute 48k simple coros/s with async,
660 and 128k coros with async_pool).
661 - Coro::AIO now uses the coroutine priority as io priority.
662
663 3.2 Fri Dec 22 05:07:09 CET 2006
664 - improve portability to slightly older perls.
665 - use cleaner coroutine destruction.
666 - simplify configuration for users.
667 - optionally (unrecommended) prefer perl functions over
668 their coro replacements.
669
670 3.11 Tue Dec 5 13:11:24 CET 2006
671 - fixed some bogus assert's, but as perl.h disables assert even
672 without NDEBUG (thank you very much), not too many people should
673 notice (that did include myself). Andreas König noticed, though :)
674 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
675 it. Also noticed by Andreas König :)
676 - save/restore tainted status.
677 - verified to pass the testsuite on my 5.9.5.
678
679 3.1 Mon Dec 4 23:03:40 CET 2006
680 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
681 - incompatible change: transfer flags are now per-state.
682 - give a better error message on deadlock.
683 - document Coro::nready.
684 - enhanced testsuite.
685
686 3.01 Sun Dec 3 23:47:42 CET 2006
687 - forgot to include Coro::Timer.
688
689 3.0 Sun Dec 3 22:57:25 CET 2006
690 - the "FINALLY COMPLETELY STABLE" release (coming soon:
691 the "FAMOUS LAST WORDS" release).
692 - implement a new stack sharing algorithm, which uses a stack
693 pool (size currently hardcoded).
694 - make stack sharing mandatory (it no longer uses a heuristic).
695 - eval/die no longer cause weird problems under heavy use.
696 - Coro::Event could cause livelocks if it was used but
697 no Coro::Event watchers were used.
698 - Coro::Event now uses asynccheck as prepare does not
699 check for changed watchers.
700 - Coro::Event allows multiple waiting coros and will wake up one
701 per event.
702 - Coro::Event should be cleaner and more efficient now.
703 - new utility function Coro::unblock_sub.
704 - document the sad fact that Event is no longer reentrant.
705 - putting a coroutine into the ready queue twice could under
706 some circumstances lead to stack corruption.
707 - minor incompatibility: subclassing Coro::State is supported
708 directly now without going through a _coro_state member.
709 - state/coro switching is much faster now.
710 - very minor optimisations and code/documentation cleanup.
711 - avoid problems due to compiler inlining.
712 - removed timers from Coro::Timer -> use AnyEvent instead.
713 - replaced Coro::idle coroutine by (cleaner) idle handler.
714 - updated to newest libcoro.
715 - implement enhanced support for valgrind.
716 - implement is_ready and return value for ready.
717 - removed Coro::Cont, it was a misnomer (it's generators),
718 and it was somewhat annoying to get it right. Will come back
719 if somebody asks for it :)
720 - many ->wait methods and Coro::Event could return
721 spuriously without the event having happened.
722
723 2.5 Tue Nov 7 12:22:33 CET 2006
724 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
725 use AnyEvent, moved them to Coro/.
726 - added Coro::LWP which contains all the uglyness required to
727 make LWP non-blocking.
728 - should work with perl 5.9.x now (Andreas König made me do it).
729 - fixed another bug in Coro::Select when the timeout was undef.
730 - reuse PL_start_env for all coros, saves some memory per coroutine.
731 - manage PL_top_env differently, hopefully to avoid panic: top_env.
732 - timeout argument was not properly used in Coro::Socket.
733 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
734 - emulate undocumented(!) functionality of IO::Socket required
735 by LWP(!!).
736 - updated eg/lwp to work with newer lwp's.
737 - remove "FATAL: uncaught exception" prefix. Coroutines that die
738 kill the whole process, just as exceptions in the main "coroutine"
739 did already.
740
741 2.1 Wed Nov 1 23:01:13 CET 2006
742 - fix a long-standing bug in Coro::Select where select with
743 zero timeout would instead change the current default filehandle.
744 - use a simpler and hopefully more robust way to clone padlists
745 (uses less memory and a perl function instead of our own).
746 - coro can now create a stack guard area on many architectures.
747 - Coro::AIO properly reexports additional functions from IO::AIO.
748 - updated libcoro with a workaround for OS X,
749 pach and testing by Michael Schwern.
750
751 2.0 Tue Oct 24 05:47:17 CEST 2006
752 - support additional aio requests in Coro::AIO.
753
754 1.9 (never properly released due to a glitch)
755
756 1.8 Thu Feb 2 00:59:06 CET 2006
757 - applied suggested patch by SAMV to avoid problems during stupid
758 mark & sweep gc run.
759 - applied patch by Scott Walters for 5.9.3 compatibility.
760
761 1.7 Tue Dec 27 01:41:58 CET 2005
762 - added Coro::AIO, a thin wrapper around IO::AIO.
763 - improved Makefile.PL explanations.
764
765 1.6 Mon Dec 12 21:30:05 CET 2005
766 - additionally save PM_curpm might fix as-of-yet
767 unseen problems with regex matches being attributed to the
768 wrong package.
769 - add t/10_bugs.t, which currently checks against imho broken
770 perls that use 0.26MB of stack space per Perl_magic_get
771 invocation (newer linuxes) as opposed to the 0.0002MB perl
772 normally uses.
773 - make stacksize configurable for the ultimate debian experience.
774
775 1.51 Mon Dec 12 18:48:36 CET 2005
776 - remove debugging warn() accidentally left in Coro::Select.
777
778 1.5 Tue Nov 29 13:32:44 CET 2005
779 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
780 - fix Coro::Select 'select'.
781 - strict'ify some modules.
782 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
783 - use XSLoader in selected modules.
784 - remove some 5.6 compatibility cruft.
785
786 1.4 Tue Sep 6 00:11:05 CEST 2005
787 - libcoro did not take into account the trampoline on amd64,
788 when 'l'inux method was used.
789
790 1.31 Tue Aug 30 23:31:33 CEST 2005
791 - some portability fixes/workarounds.
792
793 1.3 Sat Aug 20 03:08:56 CEST 2005
794 - no code changes. module seems to work fine.
795 - improve Coro::State docs, remove reference to nonexisting
796 Coro::State::flush method.
797 - no longer autodetect windows, present it as an option instead.
798
799 1.2 Mon May 16 02:00:55 CEST 2005
800 - included libcoro.c earlier in State.xs, might improve portability.
801 - use a faster and possibly more stable Coro::Cont implementation.
802 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
803 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
804 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
805
806 1.11 Thu Mar 3 18:00:52 CET 2005
807 - change of contact address.
808
809 1.1 Tue Feb 22 20:51:16 CET 2005
810 - support [l]inux method on amd64.
811 - allow some context switches while compiling/eval'ing.
812
813 1.0 Tue Aug 10 03:47:27 CEST 2004
814 - Event 0.89 fixes the bug again.
815 - fix the remaining(??) memleak. I didn't free the cache-AV for
816 padlists, so one AV per created anon-sub leaked.
817
818 0.97 Fri May 14 15:23:32 CEST 2004
819 - remove stupid gcc extension that was an experiment only anyway.
820
821 0.96 Thu May 13 18:09:29 CEST 2004
822 - tell people if they have yet another broken Event variant.
823 - made Coro to work _better_ with threads (it seems that, as long
824 as you use Coros in one thread only everything will work fine.
825 Looking at the perl source the problem might be missing locking
826 between Coro and some perl internal routines. That's not easily
827 fixable).
828 - (hopefully) got rid of the static xnext inside transfer.
829
830 0.95 Sun Feb 15 03:20:28 CET 2004
831 - removed Coro::State::flush, implemented a new and much
832 faster caching mechanism by attaching magic to the CV.
833 - a memleak with (real) closures remains.
834 - dramatically reduced size of stacks to 65536 (linux/x86), after
835 observing that even myhttpd never had more than a kilobyte
836 of stack in use (stack is still only physically allocated in
837 pagesize increments for systems suporting mmap).
838 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
839 - fix for threaded perl.
840
841 0.9 Sun Nov 23 04:32:39 CET 2003
842 - fix a few very hard to track down but (of course) stupid
843 memory leaks.
844 - stringify version number to avoid locale problems :(.
845 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
846 hopefully ignoring it "just works".
847
848 0.8 Wed Nov 5 19:38:40 CET 2003
849 - port to perl5.005_03, but only 5.8.x+ is supported!
850 - honor LocalAddr even without LocalPort in Coro::Socket.
851
852 0.7 Tue May 27 03:12:38 CEST 2003
853 - the version jump indicates some level of testing, not gobs
854 of new features.
855 - uh... I found the prompt function in ExtUtils::MakeMaker.
856 Highly correct stuff, that is...
857 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
858 the fix is not well understood by the author, so beware :(.
859
860 0.652 Thu May 8 02:54:46 CEST 2003
861 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
862 cause version 4 doesn't have ucontext.
863 - Benjamin Reed reported that setjmp works fine on darwin,
864 so preselect it.
865
866 0.651 Sat Mar 29 15:00:23 CET 2003
867 - fix a bug in Coro::Handle where some bytes could get lost
868 on reading, reported by jason@nichego.net.
869
870 0.65 Sun Mar 23 00:08:26 CET 2003
871 - added a README.linux-glibc.
872 - new module Coro::Select.
873 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
874 (I never claimed that I know what I am doing ;).
875 - set default to setjmp/longjmp on non-x86-linux.
876
877 0.6 Thu Nov 21 11:09:06 CET 2002
878 - made lazy context switching the default.
879 - fixed the bug where SemaphoreSet::guard
880 was the same as timed_guard and timed_guard was missing.
881 - fix a memleak in Coro::Timer.
882 - reclassified context sharing as not experimental.
883
884 0.534 Sun Apr 14 03:05:12 CEST 2002
885 - fixed backspace => tab in header parsing (myhttpd).
886 - added eg/readline, for elmex.
887 - Coro::Event now calls ->start on first call instead of
888 ->again, which created wrong timeouts for the first ->next.
889 - fixed a bug where Coro::Socket returned a closed socket instead
890 of nothing on ECONNREFUSED and similar error conditions.
891
892 0.533 Mon Feb 18 18:49:40 CET 2002
893 - adapt to finally working Event-0.85.
894 - fixed buggy event initialization (trapped under windows, but how
895 could it possibly have worked anywhere?)
896
897 0.532 Wed Jan 16 02:45:32 CET 2002
898 - removed the reference to pp_entersub, might work on
899 (some) win32 perls now (testsuite works under cygwin,
900 without Event, which freezes).
901
902 0.531 Mon Dec 10 22:18:44 CET 2001
903 - Fixed a bug in SemaphoreSet::timed_down.
904
905 0.53 Tue Nov 27 21:11:13 CET 2001
906 - some tests for Event.
907 - slightly different internal architecture to get refcnt into the object
908 and not the reference. might fix a few bugs, certainly did introduce
909 new ones.
910 - Coro::Timer (independent of Event).
911 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
912
913 0.52 Tue Nov 6 21:36:18 CET 2001
914 - ported to cygwin (trivial).
915 patch by Gerrit P. Haase <gp@familiehaase.de>.
916 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
917
918 0.51 Thu Nov 1 20:39:01 CET 2001
919 - terminate/cancel now work properly (otherwise termination
920 could cause a "next coroutine is not and contains not..."
921 error.
922 - added Coro::Socket::shutdown.
923 - Coro::Event::loop no is the same as Event::loop.
924 - implemented terminate with args + join.
925
926 0.5 Fri Sep 28 16:15:35 CEST 2001
927 - fixed "print" on a Coro::Handle. The print method worked.
928 - small tweaks (seem to reduce memory consumption a lot)
929 in various modules.
930 - splendid use of "no warnings" scattered throughout.
931 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
932
933 0.49 Sun Sep 16 02:42:45 CEST 2001
934 - changed some method calls to function calls for speed
935 inside Coro::Handle.
936 - make Coro::Handle use an array instead of a hash for
937 speed reasons.
938 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
939 sjlj and ucontext should now work.
940 - IRIX-specific port for libcoro.
941 - swapped order of accept results to match IO::Socket.
942 - changed getsock/peername to sock/peername to match IO::Socket.
943 - fixed a bug that caused segfault when returning to main task
944 under some circumstances.
945 - other bugfixes.
946
947 0.45 Sun Sep 2 02:54:01 CEST 2001
948 - new method Coro::Handle::timeout.
949 - corrected speling of set...name to get...name in Coro::Socket.
950 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
951 - Coro::Handle now supports fileno and FILENO.
952 - added eg/myhttpd, a web server.
953 - Coro::Socket now accepts numeric ports as well as "port(number)"
954 syntax.
955 - moved some scheduling primitives into xs code. more to come.
956 - new simple priority system.
957 - implemented Coro::Event using XS, almost four times faster!
958 - small memory corruption problem fixed (boy that was difficult).
959
960 0.13 Wed Aug 8 16:53:07 CEST 2001
961 - new method Coro::Handle::readline.
962 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
963 (see eg/lwp on how to make LWP non-blocking).
964 - renamed Coro::Socket::new_inet to new.
965 - added Coro::Util, some utility functions.
966 - die/eval now works better (still get annoying (but true ;)
967 "Callback called exit" messages).
968 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
969 - much better Coro::Socket support.
970
971 0.12 Fri Jul 27 04:19:01 CEST 2001
972 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
973 - very experimental stack sharing algorithm. not 100% safe but
974 should work well in practise ;)
975 - added Coro::RWLock.
976 - Coro::Specific now works.
977
978 0.11 Tue Jul 24 22:49:21 CEST 2001
979 - added specialized hack for newer and older linux versions (fast).
980 - renamed Coro::Event::IO to Coro::Handle.
981 - new module Coro::Socket.
982
983 0.1 Tue Jul 24 01:47:53 CEST 2001
984 - release candidate 3. A rather new internal structure :(
985 - the great renaming: Damian Conway gave me a suitable replacement
986 for yield.
987 - added Coro/libcoro, a portable coroutine implementation for C
988 (not even perl-dependent!!!), which can be used by Coro::State.
989 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
990 perl-5.6.
991 - new function Coro::Event::idle.
992 - the idle process is now overriden
993 by default in Coro::Event.
994 - Coro::Channel now enforces the size.
995 - canceling events now works in all cases.
996 - Coro state now include $_ and $@.
997 - yet another bug workaround that I do not really understand :(
998 - new module Coro::Event::IO (very undocumented) to do
999 non-blocking i/o.
1000 - performance tuning.
1001
1002 0.08 Thu Jul 19 06:13:25 CEST 2001
1003 - release candidate 2 (new functionality).
1004 - add Coro::State::flush function.
1005 - transfer now supports a flags value (mostly for speed ATM).
1006 - might compile and work in the presence of threads now.
1007 - continuations have a saner syntax.
1008 - no more memleaks.
1009
1010 0.07 Tue Jul 17 17:40:18 CEST 2001
1011 - release candidate 1 ;)
1012 - slightly nicer code.
1013 - fixed a scheduling bug in Coro::Event.
1014
1015 0.06 Tue Jul 17 04:23:24 CEST 2001
1016 - ok, I found the showstopper - the same sub must not be
1017 re-entered in two different coroutines, otherwise => crash. I
1018 see no easy solution to this problem, except by walking the call
1019 chaing and saving/restoring the cv's, which is what I do now.
1020 - memory leaks still latent, especially at thread termination.
1021 - Coro::Event now works (haha).
1022
1023 0.05 Sun Jul 15 17:32:20 CEST 2001
1024 - fixed a few issues in Coro::Event.
1025 - I forgot to include Coro::Event in 0.04 :(:(:(
1026
1027 0.04 Sun Jul 15 05:24:59 CEST 2001
1028 - @_ is now properly localized.
1029 - Coro::State is now easier subclassable.
1030 - Coro::Cont now coroutine-aware.
1031 - Coro::Specific is a low-overhead module to create
1032 coroutine-specific vars.
1033 - Coro::Event provides a simple interface to Event.
1034
1035 0.03 Fri Jul 13 14:51:52 CEST 2001
1036 - transfer() now implemented in XS (beware).
1037 - new module Coro::Cont for really faked continuations.
1038 - big internal architecture changes: Coro::State is now
1039 really low-level and can thus be used to implement other
1040 interesting things, While "Coro::" implements a process-like
1041 model. Still crude and subject to change.
1042 - $_ and $@ are no longer being localized.
1043
1044 0.02 Tue Jul 10 01:38:17 CEST 2001
1045 - implemented "async" attribute.
1046 - $_ and $@ are now localized.
1047 - added Coro::Channel.
1048 - more testcases, still no docs.
1049
1050 0.01 Tue Jul 3 02:18:41 CEST 2001
1051 - original version; copied from Convert::Scalar.
1052