ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
(Generate patch)

Comparing Coro/Changes (file contents):
Revision 1.86 by pcg, Sun Nov 30 16:59:14 2003 UTC vs.
Revision 1.357 by root, Thu May 29 18:05:53 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO.
4TODO: 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
84.74 Thu May 29 20:05:31 CEST 2008
9 - do not test Coro::LWP for lack of dependencies.
10 - do not test Coro::Select for lack of working perls.
11
124.73 Thu May 29 2008
13 - fix a bug in Coro::EV which would cause it to block despite
14 there being runnable coroutines.
15 - sprinkle "no warnings" freely over everything, also suppress
16 warnings for some other modules.
17 - fix typo in WSAEWOULDBLOCK.
18
194.72 Sun May 25 05:14:36 CEST 2008
20 - tweak META.yaml a bit, unfortunately, there is no documented way
21 to have optional dependencies with CPAN. doh :(
22 - avoid running some tests on windows because they would fail due to
23 perl bug (broken fork, broken pipes...).
24 - work around perl on windows bugs where perl returns undocumented
25 error codes for sysread, syswrite etc. by taking advantage
26 of AnyEvent's workaround for that problem.
27 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
28 a common perl implementation bug on windows.
29 - use unix domain sockets in testsuite to work around
30 common perl implementation bugs on widows (they are emulated by
31 tcp sockets on windows. ugh.)
32
334.71 Sat May 24 20:01:27 CEST 2008
34 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
35 - take advantage of async name resolution of AnyEvent::Util.
36 - work around brutal inet_aton override in Coro::LWP.
37 - take advantage of the readyhook in Coro::EV, for smoother
38 scheduling.
39
404.7 Sun May 11 00:32:19 CEST 2008
41 - completely reworked the Coro manpage.
42 - added Coro::AnyEvent, generic event loop integration.
43 - implement cancel, ready and kill commands in Coro::Debug.
44 - document find_coro in Coro::Debug.
45 - incompatible change: rename has_stack to has_cctx.
46 - Coro::AIO and Coro::BDB no longer force event model detection,
47 use AnyEvent::AIO and AnyEvent::BDB.
48
494.6 Sat Apr 26 10:05:14 CEST 2008
50 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
51 by the Coro module. It was a mistake to have it in the first place.
52 - (experimental) support for activestate perl 5.10 (method "w").
53 - (experimental) support for strawberry perl (method "a").
54 - coro_sigelem_set did not return a value although it had to,
55 actual impact unknown.
56
574.51 Mon Apr 14 13:28:27 CEST 2008
58 - make it compile again on 5.8.
59
604.50 Thu Apr 10 09:43:17 CEST 2008
61 - I did it twice! (see 4.49).
62
634.49 Sun Apr 6 21:23:31 CEST 2008
64 - grr, instead of compiling the recent changes on 5.10 only they
65 were compiled on 5.8 only.
66
674.48 Sun Apr 6 20:36:46 CEST 2008
68 - allow coroutine switches during eval's under 5.10.x, as it seems
69 the parser is a per-interpreter option now, so this is safe
70 (this might fix the odd crashing bug).
71 - drop support for 5.9.x versions: they are dead, jim.
72
734.47 Sun Apr 6 00:37:52 CEST 2008
74 - force cctx allocation on API calls: we know we need to force one
75 and gcc actually manages to confuse our heuristic nowadays,
76 leading to crashes and worse.
77 - document force_cctx.
78
794.46 Fri Apr 4 22:05:43 CEST 2008
80 - upgrade libcoro, resulting in pthread-backend (which was only created
81 to fulfill the rules of the programming languages shootout).
82
834.45 Thu Mar 13 11:55:36 CET 2008
84 - fix a file leaking bug in eg/mhyttpd that would allow
85 downloading of any file (reported by oesi).
86 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
87 (also add testcase).
88 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
89 apparently having loads of fun with that).
90 - implement $state->swap_defsv, ->swap_defav and document ->throw.
91
924.4 Wed Feb 13 16:44:29 CET 2008
93 - only assume .cfi pseudo isns exist on GNU/Linux.
94 - add get_timed method to Coro::Channels.
95 - fixed Coro::Debug manpage.
96 - perl 5.11 compatbility improvement based on patch by
97 Andreas König.
98
994.37 Sun Jan 20 11:25:23 CET 2008
100 - soften the check for an active parser for perl 5.10 (avoiding
101 "Coro::State::transfer called while parsing" in many cases).
102
1034.36 Sun Jan 13 10:53:56 CET 2008
104 - reset diehook when terminating from an async_pool as to not
105 trigger any __DIE__ handlers.
106
1074.35 Sun Jan 13 04:14:13 CET 2008
108 - "bt" debug command now displays any exceptions
109 from longmess and also skips the innermost
110 stackframes, giving more useufl output.
111 - allow backtraces in coroutines blocked in Coro::EV,
112 at a <1% speed hit.
113 - handle localising of $SIG{__WARN__} and __DIE__
114 properly (with a proper amount of dirty hacking).
115
1164.34 Sat Dec 22 17:49:53 CET 2007
117 - upgrade to EV version 2.0 API.
118
1194.33 Mon Dec 17 08:36:12 CET 2007
120 - make Coro::AIO etc. loadable in the absence of EV.
121
1224.32 Mon Dec 17 07:46:02 CET 2007
123 - majorly improved Coro::Handle's performance with Coro::EV.
124 - implemented the until now mythical Coro::BDB module.
125 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
126
1274.31 Wed Dec 5 12:32:39 CET 2007
128 - remove warn statement form Coro::Util.
129
1304.3 Tue Dec 4 20:33:14 CET 2007
131 - calls to the idle function could cause stack corruption
132 when the stack changed.
133 - do no longer rely on the presence of EV::DNS if EV is used
134 (because it is gone), but instead take avdantage of EV::ADNS
135 if available.
136 - add ($) prototypes to all functions in Coro::Storable.
137 - use a conventional (safer) idle callback in Coro::EV.
138 - do accept quickly in Coro::Debug to avoid endless loops.
139
1404.22 Fri Nov 30 16:04:04 CET 2007
141 - really use optimised versions for Event and EV in Coro::Util
142 and Coro::Handle.
143
1444.21 Sun Nov 25 10:48:59 CET 2007
145 - fix a spurious memory read.
146 - Coro::EV no longer keeps the eventloop "alive".
147
1484.2 Fri Nov 9 20:47:05 CET 2007
149 - enable/disable tracing from a new coroutine, not a pooled one.
150 - fix a memleak in Coro::Event.
151 - removed killall call from fork_eval.
152 - made sure store_fd is already loaded so that fork_eval
153 does not have to parse autoload in each subprocess.
154 - only use assembly method if -O switch is in $Config{optimize}.
155 - add (optional) Coro::EV module, so far the best event loop module
156 directly supported by Coro.
157 - if the event model is EV, use EV::DNS to resolve
158 stuff in Coro::Util.
159 - don't get confused by multiple event notifications in Coro::Handle.
160 - initial support for EV (libevent interface).
161 - require Event and EV using configure_requires, to force their existance.
162
1634.13 Wed Oct 24 07:26:45 CEST 2007
164 - add Coro::Storable::blocking_thaw.
165 - use a vastly more complicated technique to localise
166 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
167 - use a coroutine for the idle callback Coro::Event,
168 instead of running Event in the current coroutine context.
169 This also catches recursive invocations.
170 - actually report fork errors in gethostbyname and inet_aton.
171
1724.11 Thu Oct 11 02:40:24 CEST 2007
173 - port to threaded perls.
174
1754.1 Thu Oct 11 02:38:16 CEST 2007
176 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
177 be local to each coro (see Coro::State).
178 - incompatible change: for very deep reasons, cede and cede_notself
179 cannot return anything, so nothing will be returned.
180 - possibly bring back 5.10 compatibility (untested).
181 - work around stupid (and wrong) warning on 5.10 :(.
182 - overlay the saved state over the context stack. This saves
183 a few hundred bytes per coroutine on average and also
184 speeds up context switching a bit.
185 - further tune default stack sizes.
186 - (more) correctly calculate stack usage in coro_rss.
187 - Coro::Storable::blocking_* did not properly lock
188 resulting in races between coroutines.
189 - added Coro::Storable::guard.
190 - stopping to trace a coroutine could destroy the cctx of
191 an unrelated coroutine.
192 - explain the relationship between Perl and C coroutines in
193 more detail in Coro::State.
194 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
195 causing a fork per resolve. This also affected Coro::Socket.
196 - switch to a separate stack in $coro->call/eval to avoid
197 invalidating pointers.
198
1994.03 Sat Oct 6 21:24:00 CEST 2007
200 - added Coro::throw method.
201 - minor code cleanups.
202
2034.02 Sat Oct 6 02:36:47 CEST 2007
204 - fix a very minor per-coroutine memleak (a single codereference).
205 - fixed a bug where the currently in-use c context would be freed
206 prematurely (can happen only when programs change the stacksize
207 or use tracing).
208 - tracing can no longer keep a coro alive after it terminated.
209 - do static branch prediction in the common path for gcc. gives
210 about 2-5% speed improvement here.
211
2124.01 Fri Oct 5 22:10:49 CEST 2007
213 - instead of recreating *a* standard output handle we simply
214 use STDOUT, which is faster and hopefully more robust.
215
2164.0 Fri Oct 5 12:56:00 CEST 2007
217 - incompatibly changed Coro::Storable::freeze.
218 - major new feature: added Coro::Debug, for interactive coroutine
219 debugging, tracing and much more.
220 - major bug fix: unbelievable, but true: $_, $/ and many other
221 "saved" variables actually weren't being saved. This has been fixed,
222 of course, while increasing performance while losing all the save
223 flags.
224 - save flags are gone, and all the api functions dealing with them.
225 - added Coro::Semaphore::adjust.
226 - added Coro::Util::fork_eval.
227 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
228 - added Coro::killall.
229 - reduce initial stack sizes to allow for "micro-coroutines".
230 - better async_pool resource management, moved parts of async_pool
231 handling to XS (major speed improvement).
232 - actually croak before modifying important data structures.
233 - refuse to transfer while compiling.
234 - possibly support eval EXPR better now.
235 - enable assembly per default on linux+bsd x86+amd64.
236 - all internal members were renamed _something for easier subclassing.
237 - many minor tweaks.
238
2393.63 Wed May 16 14:10:06 CEST 2007
240 - implement handcoded assembly for x86/amd64 SVR ABI.
241
2423.62 Fri Apr 27 21:36:06 CEST 2007
243 - upgrade libcoro (which might set unwind info correctly).
244 - change default on linux to setjmp/longjmp.
245
2463.61 Thu Apr 19 12:36:18 CEST 2007
247 - Coro::Storable caused an endless loop when thawing invalid
248 pst's sometimes.
249 - use a Semaphore in Coro::Storable, as Storable doesn't
250 seem to be reentrant (although it is documented to
251 be threadsafe...).
252 - fix Coro::Signal to bring back the original unreliable
253 but stateful semantics.
254 - fixed a lot of typos in Coro.pm (patch submitted by David
255 Steinbrunner, which applied flawlessly).
256
2573.6 Sat Apr 14 17:13:31 CEST 2007
258 - added some bugfixes to get eg/myhttpd working again.
259 - added Coro::Storable for often-cede'ing freeze/thaw.
260 - try to do a clean exit when a coroutine calls exit
261 (EXPERIMENTAL).
262 - got rid of indirect call through _coro_init.
263 - updated the partly antique examples in eg/ to
264 work again and be a bit less magic, too.
265 - fixed Coro::Signal semantics to work as documented again.
266
2673.55 Sun Mar 25 01:20:47 CET 2007
268 - add SAVE_DEFFH to save the default filehandle and enable
269 it by default.
270 - finally move socket-operations from Coro::Socket to Coro::Handle
271 to be able to unblock foreign sockets.
272 - add Coro::State::save_also and guarded_save.
273 - add count accessor to Coro::Semaphore.
274 - add Coro::State::cctx_stacksize.
275 - just for the fun of it, do not rely on implicit context,
276 which can dramatically improve performance, but people
277 using windows-process-emulation perls do not care much about
278 performance.
279
2803.51 Sun Mar 4 14:18:04 CET 2007
281 - fixed a problem when you weakened references to Coro::State's
282 (patch by Yuval Kogman).
283
2843.501 Wed Feb 28 12:44:07 CET 2007
285 - rename some global symbols as macosx from hell redefines
286 them without asking.
287
2883.5 Tue Feb 13 20:22:53 CET 2007
289 - do AnyEvent model detection earlier, avoiding problems
290 caused by first using AnyEvent and later Coro::Event.
291 - implement and document Coro::Event event objects.
292 - fix a potential problem in Coro::Event causing crashes.
293 - initialise PL_comppad when creating a new coroutine,
294 avoids crashes on early coro destruction.
295
2963.41 Mon Jan 22 19:19:49 CET 2007
297 - readline on Coro::Handle did not support undefined $/,
298 nor did it deliver partial lines on EOF or error.
299 - implement malloc fallback for stack allocation because
300 stupid broken idiotic OSX has a stupid broken
301 idiotic fits-the-whole-os mmap "implementation" and
302 my dick feels longer if Coro is portable even to
303 obsolete platforms.
304
3053.4 Fri Jan 19 21:52:54 CET 2007
306 - remove t/09_timer.t, as it isn't really testing much
307 but was rather flaky in practise.
308 - async_pool coro would keep arguments and callback alive until
309 it was reused.
310 - cancellation of a coroutine could cause spurious idle calls
311 in cede_notself.
312
3133.3 Sat Jan 6 03:45:00 CET 2007
314 - implement $coro->on_destroy.
315 - Coro::Event blocking semantics have been changed,
316 documented and - hopefully - improved.
317 - fix nice adding, not subtracting, from priority.
318 - fix ->prio and api_is_ready (patch by Mark Hinds).
319 - fixed an assert ("... == prev__cctx->idle_te")
320 that could errronously trigger.
321 - fix various large and small memleaks.
322 - use a (hopefully) more stable cancel implementation
323 that immediately frees the coroutine data.
324 - cede/cede_notself return a status now.
325 - added Coro::guard function.
326 - added a global coroutine pool for jobs (on my machine,
327 I can create and execute 48k simple coros/s with async,
328 and 128k coros with async_pool).
329 - Coro::AIO now uses the coroutine priority as io priority.
330
3313.2 Fri Dec 22 05:07:09 CET 2006
332 - improve portability to slightly older perls.
333 - use cleaner coroutine destruction.
334 - simplify configuration for users.
335 - optionally (unrecommended) prefer perl functions over
336 their coro replacements.
337
3383.11 Tue Dec 5 13:11:24 CET 2006
339 - fixed some bogus assert's, but as perl.h disables assert even
340 without NDEBUG (thank you very much), not too many people should
341 notice (that did include myself). Andreas König noticed, though :)
342 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
343 it. Also noticed by Andreas König :)
344 - save/restore tainted status.
345 - verified to pass the testsuite on my 5.9.5.
346
3473.1 Mon Dec 4 23:03:40 CET 2006
348 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
349 - incompatible change: transfer flags are now per-state.
350 - give a better error message on deadlock.
351 - document Coro::nready.
352 - enhanced testsuite.
353
3543.01 Sun Dec 3 23:47:42 CET 2006
355 - forgot to include Coro::Timer.
356
3573.0 Sun Dec 3 22:57:25 CET 2006
358 - the "FINALLY COMPLETELY STABLE" release (coming soon:
359 the "FAMOUS LAST WORDS" release).
360 - implement a new stack sharing algorithm, which uses a stack
361 pool (size currently hardcoded).
362 - make stack sharing mandatory (it no longer uses a heuristic).
363 - eval/die no longer cause weird problems under heavy use.
364 - Coro::Event could cause livelocks if it was used but
365 no Coro::Event watchers were used.
366 - Coro::Event now uses asynccheck as prepare does not
367 check for changed watchers.
368 - Coro::Event allows multiple waiting coros and will wake up one
369 per event.
370 - Coro::Event should be cleaner and more efficient now.
371 - new utility function Coro::unblock_sub.
372 - document the sad fact that Event is no longer reentrant.
373 - putting a coroutine into the ready queue twice could under
374 some circumstances lead to stack corruption.
375 - minor incompatibility: subclassing Coro::State is supported
376 directly now without going through a _coro_state member.
377 - state/coro switching is much faster now.
378 - very minor optimisations and code/documentation cleanup.
379 - avoid problems due to compiler inlining.
380 - removed timers from Coro::Timer -> use AnyEvent instead.
381 - replaced Coro::idle coroutine by (cleaner) idle handler.
382 - updated to newest libcoro.
383 - implement enhanced support for valgrind.
384 - implement is_ready and return value for ready.
385 - removed Coro::Cont, it was a misnomer (it's generators),
386 and it was somewhat annoying to get it right. Will come back
387 if somebody asks for it :)
388 - many ->wait methods and Coro::Event could return
389 spuriously without the event having happened.
390
3912.5 Tue Nov 7 12:22:33 CET 2006
392 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
393 use AnyEvent, moved them to Coro/.
394 - added Coro::LWP which contains all the uglyness required to
395 make LWP non-blocking.
396 - should work with perl 5.9.x now (Andreas König made me do it).
397 - fixed another bug in Coro::Select when the timeout was undef.
398 - reuse PL_start_env for all coros, saves some memory per coroutine.
399 - manage PL_top_env differently, hopefully to avoid panic: top_env.
400 - timeout argument was not properly used in Coro::Socket.
401 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
402 - emulate undocumented(!) functionality of IO::Socket required
403 by LWP(!!).
404 - updated eg/lwp to work with newer lwp's.
405 - remove "FATAL: uncaught exception" prefix. Coroutines that die
406 kill the whole process, just as exceptions in the main "coroutine"
407 did already.
408
4092.1 Wed Nov 1 23:01:13 CET 2006
410 - fix a long-standing bug in Coro::Select where select with
411 zero timeout would instead change the current default filehandle.
412 - use a simpler and hopefully more robust way to clone padlists
413 (uses less memory and a perl function instead of our own).
414 - coro can now create a stack guard area on many architectures.
415 - Coro::AIO properly reexports additional functions from IO::AIO.
416 - updated libcoro with a workaround for OS X,
417 pach and testing by Michael Schwern.
418
4192.0 Tue Oct 24 05:47:17 CEST 2006
420 - support additional aio requests in Coro::AIO.
421
4221.9 (never properly released due to a glitch)
423
4241.8 Thu Feb 2 00:59:06 CET 2006
425 - applied suggested patch by SAMV to avoid problems during stupid
426 mark & sweep gc run.
427 - applied patch by Scott Walters for 5.9.3 compatibility.
428
4291.7 Tue Dec 27 01:41:58 CET 2005
430 - added Coro::AIO, a thin wrapper around IO::AIO.
431 - improved Makefile.PL explanations.
432
4331.6 Mon Dec 12 21:30:05 CET 2005
434 - additionally save PM_curpm might fix as-of-yet
435 unseen problems with regex matches being attributed to the
436 wrong package.
437 - add t/10_bugs.t, which currently checks against imho broken
438 perls that use 0.26MB of stack space per Perl_magic_get
439 invocation (newer linuxes) as opposed to the 0.0002MB perl
440 normally uses.
441 - make stacksize configurable for the ultimate debian experience.
442
4431.51 Mon Dec 12 18:48:36 CET 2005
444 - remove debugging warn() accidentally left in Coro::Select.
445
4461.5 Tue Nov 29 13:32:44 CET 2005
447 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
448 - fix Coro::Select 'select'.
449 - strict'ify some modules.
450 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
451 - use XSLoader in selected modules.
452 - remove some 5.6 compatibility cruft.
453
4541.4 Tue Sep 6 00:11:05 CEST 2005
455 - libcoro did not take into account the trampoline on amd64,
456 when 'l'inux method was used.
457
4581.31 Tue Aug 30 23:31:33 CEST 2005
459 - some portability fixes/workarounds.
460
4611.3 Sat Aug 20 03:08:56 CEST 2005
462 - no code changes. module seems to work fine.
463 - improve Coro::State docs, remove reference to nonexisting
464 Coro::State::flush method.
465 - no longer autodetect windows, present it as an option instead.
466
4671.2 Mon May 16 02:00:55 CEST 2005
468 - included libcoro.c earlier in State.xs, might improve portability.
469 - use a faster and possibly more stable Coro::Cont implementation.
470 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
471 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
472 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
473
4741.11 Thu Mar 3 18:00:52 CET 2005
475 - change of contact address.
476
4771.1 Tue Feb 22 20:51:16 CET 2005
478 - support [l]inux method on amd64.
479 - allow some context switches while compiling/eval'ing.
480
4811.0 Tue Aug 10 03:47:27 CEST 2004
482 - Event 0.89 fixes the bug again.
483 - fix the remaining(??) memleak. I didn't free the cache-AV for
484 padlists, so one AV per created anon-sub leaked.
485
4860.97 Fri May 14 15:23:32 CEST 2004
487 - remove stupid gcc extension that was an experiment only anyway.
488
4890.96 Thu May 13 18:09:29 CEST 2004
490 - tell people if they have yet another broken Event variant.
491 - made Coro to work _better_ with threads (it seems that, as long
492 as you use Coros in one thread only everything will work fine.
493 Looking at the perl source the problem might be missing locking
494 between Coro and some perl internal routines. That's not easily
495 fixable).
496 - (hopefully) got rid of the static xnext inside transfer.
497
4980.95 Sun Feb 15 03:20:28 CET 2004
3 - removed Coro::State::flush, implemented a new and much 499 - removed Coro::State::flush, implemented a new and much
4 faster caching mechanism by attaching magic to the CV. 500 faster caching mechanism by attaching magic to the CV.
501 - a memleak with (real) closures remains.
502 - dramatically reduced size of stacks to 65536 (linux/x86), after
503 observing that even myhttpd never had more than a kilobyte
504 of stack in use (stack is still only physically allocated in
505 pagesize increments for systems suporting mmap).
506 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
507 - fix for threaded perl.
5 508
60.9 Sun Nov 23 04:32:39 CET 2003 5090.9 Sun Nov 23 04:32:39 CET 2003
7 - fix a few very hard to track down but (of course) stupid 510 - fix a few very hard to track down but (of course) stupid
8 memory leaks. 511 memory leaks.
9 - stringify version number to avoid locale problems :(. 512 - stringify version number to avoid locale problems :(.

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines