ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.443
Committed: Wed Nov 19 15:29:57 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
Changes since 1.442: +1 -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
8 TODO: async_pool to SLF
9 5.0
10 - NEW ARCHITECTURE: use the latest 4.x version if you experience
11 stability issues.
12 - bump API version to 7 - all dependents must be recompiled.
13 - removed timed_* functions - they were not being used anyways
14 and should be replaced by a more generic mechanism -
15 and were annoying to support anyways :)
16 - removed SemaphoreSet's waiter method - use sem instead.
17 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
18 - async_pool did free a scalar value twice
19 ("Attempt to unreference...").
20 - fix a disastrous bug in the readline optimisation
21 introduced in 4.801.
22 - fix a longstanding bug where calling terminate on a coro that
23 was waiting for a semaphore that was just becoming available
24 would cause a deadlock (semaphore would get into a state where
25 it was available but waiters were still blocked).
26 - calling throw on a coroutine that is waiting for a semaphore will
27 no longer make it acquire the semaphore (and thus leak a count).
28 - perl's process emulation is now not even theoretically supported
29 anymore.
30 - new functions Coro::rouse_cb and Coro::rouse_wait.
31 - new function Coro::Semaphore::wait.
32 - use named constants in Coro::Channel (Richard Hundt).
33 - directly patch the entersub opcode calling SLF functions (cede,
34 transfer and so on). this does speed up context switching, but
35 more importanly, it frees us from the hardcoded behaviour of
36 entersub, so we might actually be able to return something from
37 those functions and atcually create new ones.
38 - take advantage of __builtin_frame_address on gcc.
39 - expose THX in coroapi (not sure whether this was a wise decision,
40 as "threaded" perls are running at half speed anyways).
41 - implement execute_slf (schedule-like-function) interface that makes
42 it possible to implement schedule-like-functions in XS.
43 - use new SLF interface to massively speed up Coro::EV by roughly a
44 factor of two.
45 - used new SLF interface to massively speed up Coro::Semaphore by a
46 factor of three.
47 - used new SLF interface to speed up Coro::AIO by roughly a factor of
48 four and reduce its memory usage considerably.
49 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
50 for a nice speedup and vastly more correct behaviour. Also implement
51 a new method "sem" to get at the underlying semaphore object.
52 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
53 (in comparison) 20-40% speedup.
54 - used new SLF interface to reimplement Coro::Signal gaining
55 some unknown (because I was too lazy), but certain, speedup, and also
56 making signals reliable for the first time.
57 - sped up coroutine creation/destruction by 20%.
58 - forgot to include Coro/libcoro/README in the dist for all these years.
59 - work around a freebsd pthreads bug (manual testcancel is required as
60 pthread_cond_wait isn't a cancellation point on freebsd).
61 - use new rouse functions to speed up and simplify Coro::BDB.
62 - make "prefer perl native functions" work with threaded perls.
63 - condense Coro::Debug ps output, hint at v and w flags.
64 - (libcoro) lots of minor cleanups and portability improvements.
65
66 4.913 Sat Nov 15 07:58:28 CET 2008
67 - async_pool did free a scalar value twice
68 ("Attempt to unreference...").
69
70 4.912 Thu Nov 13 18:31:23 CET 2008
71 - minor cleanups.
72 - use much larger stacks on linux and perl < 5.8.8.
73 - Coro::Debug::new_unix_server did not unlink the socket
74 when destroyed.
75
76 4.911 Tue Nov 11 04:26:17 CET 2008
77 - "port" to threaded perls.
78
79 4.91 Mon Nov 10 05:36:38 CET 2008
80 - the ->throw exception object no longer leaks.
81 - creating a new cctx leaked a scopestack entry (memleak).
82 - new coroutines didn't get created with a zero flags field
83 (unknown impact).
84 - calling ->throw on a not-yet-started coroutine should now work
85 instead of being ignored.
86 - ->throw is now supported on Coro::State objects.
87 - clean up cctx creation code a bit.
88 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
89
90 4.9 Sat Nov 8 17:45:27 CET 2008
91 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
92 - (libcoro) no longer leak threads in the experimental pthread backend,
93 also speed it up considerably.
94 - (libcoro) do not rely on makecontext passing void *'s unscathed.
95 - fix compiletime dependencies on libcoro in the Makefile.
96 - cctx_count wasn't always updated properly.
97 - Coro::State::cctx_stacksize wasn't applied correctly.
98 - new function Coro::State::cctx_max_idle.
99 - the default max number of idle C contexts is now 4.
100 - (libcoro) try harder to get _setjmp/_longjmp.
101 - (libcoro) cleanup and extend the libcoro API to officially
102 allow the creation of empty source contexts.
103 - very experimental workaround for the totally broken netbsd platform.
104 - tried to hack around openbsd bugs.
105
106 4.804 Wed Nov 5 16:36:00 CET 2008
107 - Coro::Debug::new_unix_server would not create a non-blocking listening
108 socket, sometimes causing freezes.
109 - (libcoro) fix misaligned stack points for setjmp and assembly
110 methods, which can cause crashes on x86/x86_64 with a sufficiently
111 aggressive compiler.
112 - new function: Coro::Debug::new_tcp_server.
113 - move ->throw into the Coro class because it only works on coro objects.
114
115 4.803 Mon Nov 3 17:16:12 CET 2008
116 - (libcoro) use a global asm statement to become independent of gcc
117 otpimisations for CORO_ASM (thanks to pippijn for the idea).
118 - try to workaround yet another broken bsd, this time dragonfly.
119
120 4.802 Thu Oct 30 10:56:12 CET 2008
121 - support -fno-omit-frame-pointer on x86 with the assembly method.
122 - tune 01_unblock.t tests a bit.
123
124 4.801 Wed Oct 22 18:33:37 CEST 2008
125 - improve readline speed for very long "lines".
126 - backport to 5.8.8.
127
128 4.8 Thu Oct 2 13:34:40 CEST 2008
129 - new function Coro::AIO::aio_wait.
130 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
131 - greatly speed up and reduce memory usage of Coro::AIO requests.
132 - implement some other µ-optimisations.
133
134 4.749 Mon Sep 29 14:40:12 CEST 2008
135 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
136
137 4.748 Sat Sep 27 14:03:19 CEST 2008
138 - implement, but do not document, PerlIO::cede(granularity).
139 - Coro::Storable forgot to wrap Storable::pstore.
140 - work around the multitude of leaks and memory corruption
141 bugs in PerlIO::via by using our own C-level perliol. As a side
142 effect, Coro::Storable is now much, much, much faster.
143
144 4.747 Tue Sep 23 01:59:41 CEST 2008
145 - fix a per-cv memleak (one empty array was leaked per
146 code reference).
147 - avoid a crash in coro->call|rss when the coroutine was already
148 destroyed (most noticably when using Coro::Debug::ps :)
149 - also protect *Storable::FILE.
150 - push up default storable granularity to 20ms.
151
152 4.746 Sun Sep 21 03:22:20 CEST 2008
153 - be more insistent on locking Storable against reentrancy
154 in Coro::Storable.
155 - move swap_def?v and throw to Coro::State, as documented.
156
157 4.745 Thu Jul 24 00:14:38 CEST 2008
158 - remove debugging code related to MgPV_nolen_const, also try to
159 make it compile with perl 5.8.6 (yes, apple apparently loves
160 outdated software). Reported by John S.
161
162 4.744 Tue Jul 8 22:06:35 CEST 2008
163 - correctly provide default DIE/WARN handlers as documented.
164 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
165 implementations inside perl work fine for us.
166
167 4.743 Mon Jun 16 00:21:57 CEST 2008
168 - when using Coro::EV without running EV::loop it could
169 result in busy-waiting for events, this has been fixed.
170 - reduce codesize and improve performance by using EV_DEFAULT_UC.
171
172 4.742 Sat May 31 14:10:21 CEST 2008
173 - implement a workaround for (some) perl <5.8.8 versions.
174 - require EV 3.3+.
175
176 4.741 Fri May 30 23:33:09 CEST 2008
177 - tell netbsd how utterly broken their imitation of an OS is
178 and refuse to build by default if pthreads are in use.
179 - switch to "s" method on all bsd's by default, as their ucontext
180 stuff seems just too broken.
181 - fix a bug in Coro::Select.
182
183 4.74 Thu May 29 20:05:31 CEST 2008
184 - do not test Coro::LWP for lack of dependencies.
185 - do not test Coro::Select for lack of working perls.
186
187 4.73 Thu May 29 2008
188 - fix a bug in Coro::EV which would cause it to block despite
189 there being runnable coroutines.
190 - sprinkle "no warnings" freely over everything, also suppress
191 warnings for some other modules.
192 - fix typo in WSAEWOULDBLOCK.
193
194 4.72 Sun May 25 05:14:36 CEST 2008
195 - tweak META.yaml a bit, unfortunately, there is no documented way
196 to have optional dependencies with CPAN. doh :(
197 - avoid running some tests on windows because they would fail due to
198 perl bug (broken fork, broken pipes...).
199 - work around perl on windows bugs where perl returns undocumented
200 error codes for sysread, syswrite etc. by taking advantage
201 of AnyEvent's workaround for that problem.
202 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
203 a common perl implementation bug on windows.
204 - use unix domain sockets in testsuite to work around
205 common perl implementation bugs on widows (they are emulated by
206 tcp sockets on windows. ugh.)
207
208 4.71 Sat May 24 20:01:27 CEST 2008
209 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
210 - take advantage of async name resolution of AnyEvent::Util.
211 - work around brutal inet_aton override in Coro::LWP.
212 - take advantage of the readyhook in Coro::EV, for smoother
213 scheduling.
214
215 4.7 Sun May 11 00:32:19 CEST 2008
216 - completely reworked the Coro manpage.
217 - added Coro::AnyEvent, generic event loop integration.
218 - implement cancel, ready and kill commands in Coro::Debug.
219 - document find_coro in Coro::Debug.
220 - incompatible change: rename has_stack to has_cctx.
221 - Coro::AIO and Coro::BDB no longer force event model detection,
222 use AnyEvent::AIO and AnyEvent::BDB.
223
224 4.6 Sat Apr 26 10:05:14 CEST 2008
225 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
226 by the Coro module. It was a mistake to have it in the first place.
227 - (experimental) support for activestate perl 5.10 (method "w").
228 - (experimental) support for strawberry perl (method "a").
229 - coro_sigelem_set did not return a value although it had to,
230 actual impact unknown.
231
232 4.51 Mon Apr 14 13:28:27 CEST 2008
233 - make it compile again on 5.8.
234
235 4.50 Thu Apr 10 09:43:17 CEST 2008
236 - I did it twice! (see 4.49).
237
238 4.49 Sun Apr 6 21:23:31 CEST 2008
239 - grr, instead of compiling the recent changes on 5.10 only they
240 were compiled on 5.8 only.
241
242 4.48 Sun Apr 6 20:36:46 CEST 2008
243 - allow coroutine switches during eval's under 5.10.x, as it seems
244 the parser is a per-interpreter option now, so this is safe
245 (this might fix the odd crashing bug).
246 - drop support for 5.9.x versions: they are dead, jim.
247
248 4.47 Sun Apr 6 00:37:52 CEST 2008
249 - force cctx allocation on API calls: we know we need to force one
250 and gcc actually manages to confuse our heuristic nowadays,
251 leading to crashes and worse.
252 - document force_cctx.
253
254 4.46 Fri Apr 4 22:05:43 CEST 2008
255 - upgrade libcoro, resulting in pthread-backend (which was only created
256 to fulfill the rules of the programming languages shootout).
257
258 4.45 Thu Mar 13 11:55:36 CET 2008
259 - fix a file leaking bug in eg/mhyttpd that would allow
260 downloading of any file (reported by oesi).
261 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
262 (also add testcase).
263 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
264 apparently having loads of fun with that).
265 - implement $state->swap_defsv, ->swap_defav and document ->throw.
266
267 4.4 Wed Feb 13 16:44:29 CET 2008
268 - only assume .cfi pseudo isns exist on GNU/Linux.
269 - add get_timed method to Coro::Channels.
270 - fixed Coro::Debug manpage.
271 - perl 5.11 compatbility improvement based on patch by
272 Andreas König.
273
274 4.37 Sun Jan 20 11:25:23 CET 2008
275 - soften the check for an active parser for perl 5.10 (avoiding
276 "Coro::State::transfer called while parsing" in many cases).
277
278 4.36 Sun Jan 13 10:53:56 CET 2008
279 - reset diehook when terminating from an async_pool as to not
280 trigger any __DIE__ handlers.
281
282 4.35 Sun Jan 13 04:14:13 CET 2008
283 - "bt" debug command now displays any exceptions
284 from longmess and also skips the innermost
285 stackframes, giving more useufl output.
286 - allow backtraces in coroutines blocked in Coro::EV,
287 at a <1% speed hit.
288 - handle localising of $SIG{__WARN__} and __DIE__
289 properly (with a proper amount of dirty hacking).
290
291 4.34 Sat Dec 22 17:49:53 CET 2007
292 - upgrade to EV version 2.0 API.
293
294 4.33 Mon Dec 17 08:36:12 CET 2007
295 - make Coro::AIO etc. loadable in the absence of EV.
296
297 4.32 Mon Dec 17 07:46:02 CET 2007
298 - majorly improved Coro::Handle's performance with Coro::EV.
299 - implemented the until now mythical Coro::BDB module.
300 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
301
302 4.31 Wed Dec 5 12:32:39 CET 2007
303 - remove warn statement form Coro::Util.
304
305 4.3 Tue Dec 4 20:33:14 CET 2007
306 - calls to the idle function could cause stack corruption
307 when the stack changed.
308 - do no longer rely on the presence of EV::DNS if EV is used
309 (because it is gone), but instead take avdantage of EV::ADNS
310 if available.
311 - add ($) prototypes to all functions in Coro::Storable.
312 - use a conventional (safer) idle callback in Coro::EV.
313 - do accept quickly in Coro::Debug to avoid endless loops.
314
315 4.22 Fri Nov 30 16:04:04 CET 2007
316 - really use optimised versions for Event and EV in Coro::Util
317 and Coro::Handle.
318
319 4.21 Sun Nov 25 10:48:59 CET 2007
320 - fix a spurious memory read.
321 - Coro::EV no longer keeps the eventloop "alive".
322
323 4.2 Fri Nov 9 20:47:05 CET 2007
324 - enable/disable tracing from a new coroutine, not a pooled one.
325 - fix a memleak in Coro::Event.
326 - removed killall call from fork_eval.
327 - made sure store_fd is already loaded so that fork_eval
328 does not have to parse autoload in each subprocess.
329 - only use assembly method if -O switch is in $Config{optimize}.
330 - add (optional) Coro::EV module, so far the best event loop module
331 directly supported by Coro.
332 - if the event model is EV, use EV::DNS to resolve
333 stuff in Coro::Util.
334 - don't get confused by multiple event notifications in Coro::Handle.
335 - initial support for EV (libevent interface).
336 - require Event and EV using configure_requires, to force their existance.
337
338 4.13 Wed Oct 24 07:26:45 CEST 2007
339 - add Coro::Storable::blocking_thaw.
340 - use a vastly more complicated technique to localise
341 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
342 - use a coroutine for the idle callback Coro::Event,
343 instead of running Event in the current coroutine context.
344 This also catches recursive invocations.
345 - actually report fork errors in gethostbyname and inet_aton.
346
347 4.11 Thu Oct 11 02:40:24 CEST 2007
348 - port to threaded perls.
349
350 4.1 Thu Oct 11 02:38:16 CEST 2007
351 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
352 be local to each coro (see Coro::State).
353 - incompatible change: for very deep reasons, cede and cede_notself
354 cannot return anything, so nothing will be returned.
355 - possibly bring back 5.10 compatibility (untested).
356 - work around stupid (and wrong) warning on 5.10 :(.
357 - overlay the saved state over the context stack. This saves
358 a few hundred bytes per coroutine on average and also
359 speeds up context switching a bit.
360 - further tune default stack sizes.
361 - (more) correctly calculate stack usage in coro_rss.
362 - Coro::Storable::blocking_* did not properly lock
363 resulting in races between coroutines.
364 - added Coro::Storable::guard.
365 - stopping to trace a coroutine could destroy the cctx of
366 an unrelated coroutine.
367 - explain the relationship between Perl and C coroutines in
368 more detail in Coro::State.
369 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
370 causing a fork per resolve. This also affected Coro::Socket.
371 - switch to a separate stack in $coro->call/eval to avoid
372 invalidating pointers.
373
374 4.03 Sat Oct 6 21:24:00 CEST 2007
375 - added Coro::throw method.
376 - minor code cleanups.
377
378 4.02 Sat Oct 6 02:36:47 CEST 2007
379 - fix a very minor per-coroutine memleak (a single codereference).
380 - fixed a bug where the currently in-use c context would be freed
381 prematurely (can happen only when programs change the stacksize
382 or use tracing).
383 - tracing can no longer keep a coro alive after it terminated.
384 - do static branch prediction in the common path for gcc. gives
385 about 2-5% speed improvement here.
386
387 4.01 Fri Oct 5 22:10:49 CEST 2007
388 - instead of recreating *a* standard output handle we simply
389 use STDOUT, which is faster and hopefully more robust.
390
391 4.0 Fri Oct 5 12:56:00 CEST 2007
392 - incompatibly changed Coro::Storable::freeze.
393 - major new feature: added Coro::Debug, for interactive coroutine
394 debugging, tracing and much more.
395 - major bug fix: unbelievable, but true: $_, $/ and many other
396 "saved" variables actually weren't being saved. This has been fixed,
397 of course, while increasing performance while losing all the save
398 flags.
399 - save flags are gone, and all the api functions dealing with them.
400 - added Coro::Semaphore::adjust.
401 - added Coro::Util::fork_eval.
402 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
403 - added Coro::killall.
404 - reduce initial stack sizes to allow for "micro-coroutines".
405 - better async_pool resource management, moved parts of async_pool
406 handling to XS (major speed improvement).
407 - actually croak before modifying important data structures.
408 - refuse to transfer while compiling.
409 - possibly support eval EXPR better now.
410 - enable assembly per default on linux+bsd x86+amd64.
411 - all internal members were renamed _something for easier subclassing.
412 - many minor tweaks.
413
414 3.63 Wed May 16 14:10:06 CEST 2007
415 - implement handcoded assembly for x86/amd64 SVR ABI.
416
417 3.62 Fri Apr 27 21:36:06 CEST 2007
418 - upgrade libcoro (which might set unwind info correctly).
419 - change default on linux to setjmp/longjmp.
420
421 3.61 Thu Apr 19 12:36:18 CEST 2007
422 - Coro::Storable caused an endless loop when thawing invalid
423 pst's sometimes.
424 - use a Semaphore in Coro::Storable, as Storable doesn't
425 seem to be reentrant (although it is documented to
426 be threadsafe...).
427 - fix Coro::Signal to bring back the original unreliable
428 but stateful semantics.
429 - fixed a lot of typos in Coro.pm (patch submitted by David
430 Steinbrunner, which applied flawlessly).
431
432 3.6 Sat Apr 14 17:13:31 CEST 2007
433 - added some bugfixes to get eg/myhttpd working again.
434 - added Coro::Storable for often-cede'ing freeze/thaw.
435 - try to do a clean exit when a coroutine calls exit
436 (EXPERIMENTAL).
437 - got rid of indirect call through _coro_init.
438 - updated the partly antique examples in eg/ to
439 work again and be a bit less magic, too.
440 - fixed Coro::Signal semantics to work as documented again.
441
442 3.55 Sun Mar 25 01:20:47 CET 2007
443 - add SAVE_DEFFH to save the default filehandle and enable
444 it by default.
445 - finally move socket-operations from Coro::Socket to Coro::Handle
446 to be able to unblock foreign sockets.
447 - add Coro::State::save_also and guarded_save.
448 - add count accessor to Coro::Semaphore.
449 - add Coro::State::cctx_stacksize.
450 - just for the fun of it, do not rely on implicit context,
451 which can dramatically improve performance, but people
452 using windows-process-emulation perls do not care much about
453 performance.
454
455 3.51 Sun Mar 4 14:18:04 CET 2007
456 - fixed a problem when you weakened references to Coro::State's
457 (patch by Yuval Kogman).
458
459 3.501 Wed Feb 28 12:44:07 CET 2007
460 - rename some global symbols as macosx from hell redefines
461 them without asking.
462
463 3.5 Tue Feb 13 20:22:53 CET 2007
464 - do AnyEvent model detection earlier, avoiding problems
465 caused by first using AnyEvent and later Coro::Event.
466 - implement and document Coro::Event event objects.
467 - fix a potential problem in Coro::Event causing crashes.
468 - initialise PL_comppad when creating a new coroutine,
469 avoids crashes on early coro destruction.
470
471 3.41 Mon Jan 22 19:19:49 CET 2007
472 - readline on Coro::Handle did not support undefined $/,
473 nor did it deliver partial lines on EOF or error.
474 - implement malloc fallback for stack allocation because
475 stupid broken idiotic OSX has a stupid broken
476 idiotic fits-the-whole-os mmap "implementation" and
477 my dick feels longer if Coro is portable even to
478 obsolete platforms.
479
480 3.4 Fri Jan 19 21:52:54 CET 2007
481 - remove t/09_timer.t, as it isn't really testing much
482 but was rather flaky in practise.
483 - async_pool coro would keep arguments and callback alive until
484 it was reused.
485 - cancellation of a coroutine could cause spurious idle calls
486 in cede_notself.
487
488 3.3 Sat Jan 6 03:45:00 CET 2007
489 - implement $coro->on_destroy.
490 - Coro::Event blocking semantics have been changed,
491 documented and - hopefully - improved.
492 - fix nice adding, not subtracting, from priority.
493 - fix ->prio and api_is_ready (patch by Mark Hinds).
494 - fixed an assert ("... == prev__cctx->idle_te")
495 that could errronously trigger.
496 - fix various large and small memleaks.
497 - use a (hopefully) more stable cancel implementation
498 that immediately frees the coroutine data.
499 - cede/cede_notself return a status now.
500 - added Coro::guard function.
501 - added a global coroutine pool for jobs (on my machine,
502 I can create and execute 48k simple coros/s with async,
503 and 128k coros with async_pool).
504 - Coro::AIO now uses the coroutine priority as io priority.
505
506 3.2 Fri Dec 22 05:07:09 CET 2006
507 - improve portability to slightly older perls.
508 - use cleaner coroutine destruction.
509 - simplify configuration for users.
510 - optionally (unrecommended) prefer perl functions over
511 their coro replacements.
512
513 3.11 Tue Dec 5 13:11:24 CET 2006
514 - fixed some bogus assert's, but as perl.h disables assert even
515 without NDEBUG (thank you very much), not too many people should
516 notice (that did include myself). Andreas König noticed, though :)
517 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
518 it. Also noticed by Andreas König :)
519 - save/restore tainted status.
520 - verified to pass the testsuite on my 5.9.5.
521
522 3.1 Mon Dec 4 23:03:40 CET 2006
523 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
524 - incompatible change: transfer flags are now per-state.
525 - give a better error message on deadlock.
526 - document Coro::nready.
527 - enhanced testsuite.
528
529 3.01 Sun Dec 3 23:47:42 CET 2006
530 - forgot to include Coro::Timer.
531
532 3.0 Sun Dec 3 22:57:25 CET 2006
533 - the "FINALLY COMPLETELY STABLE" release (coming soon:
534 the "FAMOUS LAST WORDS" release).
535 - implement a new stack sharing algorithm, which uses a stack
536 pool (size currently hardcoded).
537 - make stack sharing mandatory (it no longer uses a heuristic).
538 - eval/die no longer cause weird problems under heavy use.
539 - Coro::Event could cause livelocks if it was used but
540 no Coro::Event watchers were used.
541 - Coro::Event now uses asynccheck as prepare does not
542 check for changed watchers.
543 - Coro::Event allows multiple waiting coros and will wake up one
544 per event.
545 - Coro::Event should be cleaner and more efficient now.
546 - new utility function Coro::unblock_sub.
547 - document the sad fact that Event is no longer reentrant.
548 - putting a coroutine into the ready queue twice could under
549 some circumstances lead to stack corruption.
550 - minor incompatibility: subclassing Coro::State is supported
551 directly now without going through a _coro_state member.
552 - state/coro switching is much faster now.
553 - very minor optimisations and code/documentation cleanup.
554 - avoid problems due to compiler inlining.
555 - removed timers from Coro::Timer -> use AnyEvent instead.
556 - replaced Coro::idle coroutine by (cleaner) idle handler.
557 - updated to newest libcoro.
558 - implement enhanced support for valgrind.
559 - implement is_ready and return value for ready.
560 - removed Coro::Cont, it was a misnomer (it's generators),
561 and it was somewhat annoying to get it right. Will come back
562 if somebody asks for it :)
563 - many ->wait methods and Coro::Event could return
564 spuriously without the event having happened.
565
566 2.5 Tue Nov 7 12:22:33 CET 2006
567 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
568 use AnyEvent, moved them to Coro/.
569 - added Coro::LWP which contains all the uglyness required to
570 make LWP non-blocking.
571 - should work with perl 5.9.x now (Andreas König made me do it).
572 - fixed another bug in Coro::Select when the timeout was undef.
573 - reuse PL_start_env for all coros, saves some memory per coroutine.
574 - manage PL_top_env differently, hopefully to avoid panic: top_env.
575 - timeout argument was not properly used in Coro::Socket.
576 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
577 - emulate undocumented(!) functionality of IO::Socket required
578 by LWP(!!).
579 - updated eg/lwp to work with newer lwp's.
580 - remove "FATAL: uncaught exception" prefix. Coroutines that die
581 kill the whole process, just as exceptions in the main "coroutine"
582 did already.
583
584 2.1 Wed Nov 1 23:01:13 CET 2006
585 - fix a long-standing bug in Coro::Select where select with
586 zero timeout would instead change the current default filehandle.
587 - use a simpler and hopefully more robust way to clone padlists
588 (uses less memory and a perl function instead of our own).
589 - coro can now create a stack guard area on many architectures.
590 - Coro::AIO properly reexports additional functions from IO::AIO.
591 - updated libcoro with a workaround for OS X,
592 pach and testing by Michael Schwern.
593
594 2.0 Tue Oct 24 05:47:17 CEST 2006
595 - support additional aio requests in Coro::AIO.
596
597 1.9 (never properly released due to a glitch)
598
599 1.8 Thu Feb 2 00:59:06 CET 2006
600 - applied suggested patch by SAMV to avoid problems during stupid
601 mark & sweep gc run.
602 - applied patch by Scott Walters for 5.9.3 compatibility.
603
604 1.7 Tue Dec 27 01:41:58 CET 2005
605 - added Coro::AIO, a thin wrapper around IO::AIO.
606 - improved Makefile.PL explanations.
607
608 1.6 Mon Dec 12 21:30:05 CET 2005
609 - additionally save PM_curpm might fix as-of-yet
610 unseen problems with regex matches being attributed to the
611 wrong package.
612 - add t/10_bugs.t, which currently checks against imho broken
613 perls that use 0.26MB of stack space per Perl_magic_get
614 invocation (newer linuxes) as opposed to the 0.0002MB perl
615 normally uses.
616 - make stacksize configurable for the ultimate debian experience.
617
618 1.51 Mon Dec 12 18:48:36 CET 2005
619 - remove debugging warn() accidentally left in Coro::Select.
620
621 1.5 Tue Nov 29 13:32:44 CET 2005
622 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
623 - fix Coro::Select 'select'.
624 - strict'ify some modules.
625 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
626 - use XSLoader in selected modules.
627 - remove some 5.6 compatibility cruft.
628
629 1.4 Tue Sep 6 00:11:05 CEST 2005
630 - libcoro did not take into account the trampoline on amd64,
631 when 'l'inux method was used.
632
633 1.31 Tue Aug 30 23:31:33 CEST 2005
634 - some portability fixes/workarounds.
635
636 1.3 Sat Aug 20 03:08:56 CEST 2005
637 - no code changes. module seems to work fine.
638 - improve Coro::State docs, remove reference to nonexisting
639 Coro::State::flush method.
640 - no longer autodetect windows, present it as an option instead.
641
642 1.2 Mon May 16 02:00:55 CEST 2005
643 - included libcoro.c earlier in State.xs, might improve portability.
644 - use a faster and possibly more stable Coro::Cont implementation.
645 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
646 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
647 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
648
649 1.11 Thu Mar 3 18:00:52 CET 2005
650 - change of contact address.
651
652 1.1 Tue Feb 22 20:51:16 CET 2005
653 - support [l]inux method on amd64.
654 - allow some context switches while compiling/eval'ing.
655
656 1.0 Tue Aug 10 03:47:27 CEST 2004
657 - Event 0.89 fixes the bug again.
658 - fix the remaining(??) memleak. I didn't free the cache-AV for
659 padlists, so one AV per created anon-sub leaked.
660
661 0.97 Fri May 14 15:23:32 CEST 2004
662 - remove stupid gcc extension that was an experiment only anyway.
663
664 0.96 Thu May 13 18:09:29 CEST 2004
665 - tell people if they have yet another broken Event variant.
666 - made Coro to work _better_ with threads (it seems that, as long
667 as you use Coros in one thread only everything will work fine.
668 Looking at the perl source the problem might be missing locking
669 between Coro and some perl internal routines. That's not easily
670 fixable).
671 - (hopefully) got rid of the static xnext inside transfer.
672
673 0.95 Sun Feb 15 03:20:28 CET 2004
674 - removed Coro::State::flush, implemented a new and much
675 faster caching mechanism by attaching magic to the CV.
676 - a memleak with (real) closures remains.
677 - dramatically reduced size of stacks to 65536 (linux/x86), after
678 observing that even myhttpd never had more than a kilobyte
679 of stack in use (stack is still only physically allocated in
680 pagesize increments for systems suporting mmap).
681 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
682 - fix for threaded perl.
683
684 0.9 Sun Nov 23 04:32:39 CET 2003
685 - fix a few very hard to track down but (of course) stupid
686 memory leaks.
687 - stringify version number to avoid locale problems :(.
688 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
689 hopefully ignoring it "just works".
690
691 0.8 Wed Nov 5 19:38:40 CET 2003
692 - port to perl5.005_03, but only 5.8.x+ is supported!
693 - honor LocalAddr even without LocalPort in Coro::Socket.
694
695 0.7 Tue May 27 03:12:38 CEST 2003
696 - the version jump indicates some level of testing, not gobs
697 of new features.
698 - uh... I found the prompt function in ExtUtils::MakeMaker.
699 Highly correct stuff, that is...
700 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
701 the fix is not well understood by the author, so beware :(.
702
703 0.652 Thu May 8 02:54:46 CEST 2003
704 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
705 cause version 4 doesn't have ucontext.
706 - Benjamin Reed reported that setjmp works fine on darwin,
707 so preselect it.
708
709 0.651 Sat Mar 29 15:00:23 CET 2003
710 - fix a bug in Coro::Handle where some bytes could get lost
711 on reading, reported by jason@nichego.net.
712
713 0.65 Sun Mar 23 00:08:26 CET 2003
714 - added a README.linux-glibc.
715 - new module Coro::Select.
716 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
717 (I never claimed that I know what I am doing ;).
718 - set default to setjmp/longjmp on non-x86-linux.
719
720 0.6 Thu Nov 21 11:09:06 CET 2002
721 - made lazy context switching the default.
722 - fixed the bug where SemaphoreSet::guard
723 was the same as timed_guard and timed_guard was missing.
724 - fix a memleak in Coro::Timer.
725 - reclassified context sharing as not experimental.
726
727 0.534 Sun Apr 14 03:05:12 CEST 2002
728 - fixed backspace => tab in header parsing (myhttpd).
729 - added eg/readline, for elmex.
730 - Coro::Event now calls ->start on first call instead of
731 ->again, which created wrong timeouts for the first ->next.
732 - fixed a bug where Coro::Socket returned a closed socket instead
733 of nothing on ECONNREFUSED and similar error conditions.
734
735 0.533 Mon Feb 18 18:49:40 CET 2002
736 - adapt to finally working Event-0.85.
737 - fixed buggy event initialization (trapped under windows, but how
738 could it possibly have worked anywhere?)
739
740 0.532 Wed Jan 16 02:45:32 CET 2002
741 - removed the reference to pp_entersub, might work on
742 (some) win32 perls now (testsuite works under cygwin,
743 without Event, which freezes).
744
745 0.531 Mon Dec 10 22:18:44 CET 2001
746 - Fixed a bug in SemaphoreSet::timed_down.
747
748 0.53 Tue Nov 27 21:11:13 CET 2001
749 - some tests for Event.
750 - slightly different internal architecture to get refcnt into the object
751 and not the reference. might fix a few bugs, certainly did introduce
752 new ones.
753 - Coro::Timer (independent of Event).
754 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
755
756 0.52 Tue Nov 6 21:36:18 CET 2001
757 - ported to cygwin (trivial).
758 patch by Gerrit P. Haase <gp@familiehaase.de>.
759 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
760
761 0.51 Thu Nov 1 20:39:01 CET 2001
762 - terminate/cancel now work properly (otherwise termination
763 could cause a "next coroutine is not and contains not..."
764 error.
765 - added Coro::Socket::shutdown.
766 - Coro::Event::loop no is the same as Event::loop.
767 - implemented terminate with args + join.
768
769 0.5 Fri Sep 28 16:15:35 CEST 2001
770 - fixed "print" on a Coro::Handle. The print method worked.
771 - small tweaks (seem to reduce memory consumption a lot)
772 in various modules.
773 - splendid use of "no warnings" scattered throughout.
774 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
775
776 0.49 Sun Sep 16 02:42:45 CEST 2001
777 - changed some method calls to function calls for speed
778 inside Coro::Handle.
779 - make Coro::Handle use an array instead of a hash for
780 speed reasons.
781 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
782 sjlj and ucontext should now work.
783 - IRIX-specific port for libcoro.
784 - swapped order of accept results to match IO::Socket.
785 - changed getsock/peername to sock/peername to match IO::Socket.
786 - fixed a bug that caused segfault when returning to main task
787 under some circumstances.
788 - other bugfixes.
789
790 0.45 Sun Sep 2 02:54:01 CEST 2001
791 - new method Coro::Handle::timeout.
792 - corrected speling of set...name to get...name in Coro::Socket.
793 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
794 - Coro::Handle now supports fileno and FILENO.
795 - added eg/myhttpd, a web server.
796 - Coro::Socket now accepts numeric ports as well as "port(number)"
797 syntax.
798 - moved some scheduling primitives into xs code. more to come.
799 - new simple priority system.
800 - implemented Coro::Event using XS, almost four times faster!
801 - small memory corruption problem fixed (boy that was difficult).
802
803 0.13 Wed Aug 8 16:53:07 CEST 2001
804 - new method Coro::Handle::readline.
805 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
806 (see eg/lwp on how to make LWP non-blocking).
807 - renamed Coro::Socket::new_inet to new.
808 - added Coro::Util, some utility functions.
809 - die/eval now works better (still get annoying (but true ;)
810 "Callback called exit" messages).
811 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
812 - much better Coro::Socket support.
813
814 0.12 Fri Jul 27 04:19:01 CEST 2001
815 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
816 - very experimental stack sharing algorithm. not 100% safe but
817 should work well in practise ;)
818 - added Coro::RWLock.
819 - Coro::Specific now works.
820
821 0.11 Tue Jul 24 22:49:21 CEST 2001
822 - added specialized hack for newer and older linux versions (fast).
823 - renamed Coro::Event::IO to Coro::Handle.
824 - new module Coro::Socket.
825
826 0.1 Tue Jul 24 01:47:53 CEST 2001
827 - release candidate 3. A rather new internal structure :(
828 - the great renaming: Damian Conway gave me a suitable replacement
829 for yield.
830 - added Coro/libcoro, a portable coroutine implementation for C
831 (not even perl-dependent!!!), which can be used by Coro::State.
832 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
833 perl-5.6.
834 - new function Coro::Event::idle.
835 - the idle process is now overriden
836 by default in Coro::Event.
837 - Coro::Channel now enforces the size.
838 - canceling events now works in all cases.
839 - Coro state now include $_ and $@.
840 - yet another bug workaround that I do not really understand :(
841 - new module Coro::Event::IO (very undocumented) to do
842 non-blocking i/o.
843 - performance tuning.
844
845 0.08 Thu Jul 19 06:13:25 CEST 2001
846 - release candidate 2 (new functionality).
847 - add Coro::State::flush function.
848 - transfer now supports a flags value (mostly for speed ATM).
849 - might compile and work in the presence of threads now.
850 - continuations have a saner syntax.
851 - no more memleaks.
852
853 0.07 Tue Jul 17 17:40:18 CEST 2001
854 - release candidate 1 ;)
855 - slightly nicer code.
856 - fixed a scheduling bug in Coro::Event.
857
858 0.06 Tue Jul 17 04:23:24 CEST 2001
859 - ok, I found the showstopper - the same sub must not be
860 re-entered in two different coroutines, otherwise => crash. I
861 see no easy solution to this problem, except by walking the call
862 chaing and saving/restoring the cv's, which is what I do now.
863 - memory leaks still latent, especially at thread termination.
864 - Coro::Event now works (haha).
865
866 0.05 Sun Jul 15 17:32:20 CEST 2001
867 - fixed a few issues in Coro::Event.
868 - I forgot to include Coro::Event in 0.04 :(:(:(
869
870 0.04 Sun Jul 15 05:24:59 CEST 2001
871 - @_ is now properly localized.
872 - Coro::State is now easier subclassable.
873 - Coro::Cont now coroutine-aware.
874 - Coro::Specific is a low-overhead module to create
875 coroutine-specific vars.
876 - Coro::Event provides a simple interface to Event.
877
878 0.03 Fri Jul 13 14:51:52 CEST 2001
879 - transfer() now implemented in XS (beware).
880 - new module Coro::Cont for really faked continuations.
881 - big internal architecture changes: Coro::State is now
882 really low-level and can thus be used to implement other
883 interesting things, While "Coro::" implements a process-like
884 model. Still crude and subject to change.
885 - $_ and $@ are no longer being localized.
886
887 0.02 Tue Jul 10 01:38:17 CEST 2001
888 - implemented "async" attribute.
889 - $_ and $@ are now localized.
890 - added Coro::Channel.
891 - more testcases, still no docs.
892
893 0.01 Tue Jul 3 02:18:41 CEST 2001
894 - original version; copied from Convert::Scalar.
895