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

Comparing Coro/Changes (file contents):
Revision 1.161 by root, Mon Nov 27 02:08:55 2006 UTC vs.
Revision 1.383 by root, Fri Oct 24 07:52:52 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
33.0 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
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
8 - tune 01_unblock.t tests a bit.
9
104.801 Wed Oct 22 18:33:37 CEST 2008
11 - improve readline speed for very long "lines".
12 - backport to 5.8.8.
13
144.8 Thu Oct 2 13:34:40 CEST 2008
15 - new function Coro::AIO::aio_wait.
16 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
17 - greatly speed up and reduce memory usage of Coro::AIO requests.
18 - implement some other µ-optimisations.
19
204.749 Mon Sep 29 14:40:12 CEST 2008
21 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
22
234.748 Sat Sep 27 14:03:19 CEST 2008
24 - implement, but do not document, PerlIO::cede(granularity).
25 - Coro::Storable forgot to wrap Storable::pstore.
26 - work around the multitude of leaks and memory corruption
27 bugs in PerlIO::via by using our own C-level perliol. As a side
28 effect, Coro::Storable is now much, much, much faster.
29
304.747 Tue Sep 23 01:59:41 CEST 2008
31 - fix a per-cv memleak (one empty array was leaked per
32 code reference).
33 - avoid a crash in coro->call|rss when the coroutine was already
34 destroyed (most noticably when using Coro::Debug::ps :)
35 - also protect *Storable::FILE.
36 - push up default storable granularity to 20ms.
37
384.746 Sun Sep 21 03:22:20 CEST 2008
39 - be more insistent on locking Storable against reentrancy
40 in Coro::Storable.
41 - move swap_def?v and throw to Coro::State, as documented.
42
434.745 Thu Jul 24 00:14:38 CEST 2008
44 - remove debugging code related to MgPV_nolen_const, also try to
45 make it compile with perl 5.8.6 (yes, apple apparently loves
46 outdated software). Reported by John S.
47
484.744 Tue Jul 8 22:06:35 CEST 2008
49 - correctly provide default DIE/WARN handlers as documented.
50 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
51 implementations inside perl work fine for us.
52
534.743 Mon Jun 16 00:21:57 CEST 2008
54 - when using Coro::EV without running EV::loop it could
55 result in busy-waiting for events, this has been fixed.
56 - reduce codesize and improve performance by using EV_DEFAULT_UC.
57
584.742 Sat May 31 14:10:21 CEST 2008
59 - implement a workaround for (some) perl <5.8.8 versions.
60 - require EV 3.3+.
61
624.741 Fri May 30 23:33:09 CEST 2008
63 - tell netbsd how utterly broken their imitation of an OS is
64 and refuse to build by default if pthreads are in use.
65 - switch to "s" method on all bsd's by default, as their ucontext
66 stuff seems just too broken.
67 - fix a bug in Coro::Select.
68
694.74 Thu May 29 20:05:31 CEST 2008
70 - do not test Coro::LWP for lack of dependencies.
71 - do not test Coro::Select for lack of working perls.
72
734.73 Thu May 29 2008
74 - fix a bug in Coro::EV which would cause it to block despite
75 there being runnable coroutines.
76 - sprinkle "no warnings" freely over everything, also suppress
77 warnings for some other modules.
78 - fix typo in WSAEWOULDBLOCK.
79
804.72 Sun May 25 05:14:36 CEST 2008
81 - tweak META.yaml a bit, unfortunately, there is no documented way
82 to have optional dependencies with CPAN. doh :(
83 - avoid running some tests on windows because they would fail due to
84 perl bug (broken fork, broken pipes...).
85 - work around perl on windows bugs where perl returns undocumented
86 error codes for sysread, syswrite etc. by taking advantage
87 of AnyEvent's workaround for that problem.
88 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
89 a common perl implementation bug on windows.
90 - use unix domain sockets in testsuite to work around
91 common perl implementation bugs on widows (they are emulated by
92 tcp sockets on windows. ugh.)
93
944.71 Sat May 24 20:01:27 CEST 2008
95 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
96 - take advantage of async name resolution of AnyEvent::Util.
97 - work around brutal inet_aton override in Coro::LWP.
98 - take advantage of the readyhook in Coro::EV, for smoother
99 scheduling.
100
1014.7 Sun May 11 00:32:19 CEST 2008
102 - completely reworked the Coro manpage.
103 - added Coro::AnyEvent, generic event loop integration.
104 - implement cancel, ready and kill commands in Coro::Debug.
105 - document find_coro in Coro::Debug.
106 - incompatible change: rename has_stack to has_cctx.
107 - Coro::AIO and Coro::BDB no longer force event model detection,
108 use AnyEvent::AIO and AnyEvent::BDB.
109
1104.6 Sat Apr 26 10:05:14 CEST 2008
111 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
112 by the Coro module. It was a mistake to have it in the first place.
113 - (experimental) support for activestate perl 5.10 (method "w").
114 - (experimental) support for strawberry perl (method "a").
115 - coro_sigelem_set did not return a value although it had to,
116 actual impact unknown.
117
1184.51 Mon Apr 14 13:28:27 CEST 2008
119 - make it compile again on 5.8.
120
1214.50 Thu Apr 10 09:43:17 CEST 2008
122 - I did it twice! (see 4.49).
123
1244.49 Sun Apr 6 21:23:31 CEST 2008
125 - grr, instead of compiling the recent changes on 5.10 only they
126 were compiled on 5.8 only.
127
1284.48 Sun Apr 6 20:36:46 CEST 2008
129 - allow coroutine switches during eval's under 5.10.x, as it seems
130 the parser is a per-interpreter option now, so this is safe
131 (this might fix the odd crashing bug).
132 - drop support for 5.9.x versions: they are dead, jim.
133
1344.47 Sun Apr 6 00:37:52 CEST 2008
135 - force cctx allocation on API calls: we know we need to force one
136 and gcc actually manages to confuse our heuristic nowadays,
137 leading to crashes and worse.
138 - document force_cctx.
139
1404.46 Fri Apr 4 22:05:43 CEST 2008
141 - upgrade libcoro, resulting in pthread-backend (which was only created
142 to fulfill the rules of the programming languages shootout).
143
1444.45 Thu Mar 13 11:55:36 CET 2008
145 - fix a file leaking bug in eg/mhyttpd that would allow
146 downloading of any file (reported by oesi).
147 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
148 (also add testcase).
149 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
150 apparently having loads of fun with that).
151 - implement $state->swap_defsv, ->swap_defav and document ->throw.
152
1534.4 Wed Feb 13 16:44:29 CET 2008
154 - only assume .cfi pseudo isns exist on GNU/Linux.
155 - add get_timed method to Coro::Channels.
156 - fixed Coro::Debug manpage.
157 - perl 5.11 compatbility improvement based on patch by
158 Andreas König.
159
1604.37 Sun Jan 20 11:25:23 CET 2008
161 - soften the check for an active parser for perl 5.10 (avoiding
162 "Coro::State::transfer called while parsing" in many cases).
163
1644.36 Sun Jan 13 10:53:56 CET 2008
165 - reset diehook when terminating from an async_pool as to not
166 trigger any __DIE__ handlers.
167
1684.35 Sun Jan 13 04:14:13 CET 2008
169 - "bt" debug command now displays any exceptions
170 from longmess and also skips the innermost
171 stackframes, giving more useufl output.
172 - allow backtraces in coroutines blocked in Coro::EV,
173 at a <1% speed hit.
174 - handle localising of $SIG{__WARN__} and __DIE__
175 properly (with a proper amount of dirty hacking).
176
1774.34 Sat Dec 22 17:49:53 CET 2007
178 - upgrade to EV version 2.0 API.
179
1804.33 Mon Dec 17 08:36:12 CET 2007
181 - make Coro::AIO etc. loadable in the absence of EV.
182
1834.32 Mon Dec 17 07:46:02 CET 2007
184 - majorly improved Coro::Handle's performance with Coro::EV.
185 - implemented the until now mythical Coro::BDB module.
186 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
187
1884.31 Wed Dec 5 12:32:39 CET 2007
189 - remove warn statement form Coro::Util.
190
1914.3 Tue Dec 4 20:33:14 CET 2007
192 - calls to the idle function could cause stack corruption
193 when the stack changed.
194 - do no longer rely on the presence of EV::DNS if EV is used
195 (because it is gone), but instead take avdantage of EV::ADNS
196 if available.
197 - add ($) prototypes to all functions in Coro::Storable.
198 - use a conventional (safer) idle callback in Coro::EV.
199 - do accept quickly in Coro::Debug to avoid endless loops.
200
2014.22 Fri Nov 30 16:04:04 CET 2007
202 - really use optimised versions for Event and EV in Coro::Util
203 and Coro::Handle.
204
2054.21 Sun Nov 25 10:48:59 CET 2007
206 - fix a spurious memory read.
207 - Coro::EV no longer keeps the eventloop "alive".
208
2094.2 Fri Nov 9 20:47:05 CET 2007
210 - enable/disable tracing from a new coroutine, not a pooled one.
211 - fix a memleak in Coro::Event.
212 - removed killall call from fork_eval.
213 - made sure store_fd is already loaded so that fork_eval
214 does not have to parse autoload in each subprocess.
215 - only use assembly method if -O switch is in $Config{optimize}.
216 - add (optional) Coro::EV module, so far the best event loop module
217 directly supported by Coro.
218 - if the event model is EV, use EV::DNS to resolve
219 stuff in Coro::Util.
220 - don't get confused by multiple event notifications in Coro::Handle.
221 - initial support for EV (libevent interface).
222 - require Event and EV using configure_requires, to force their existance.
223
2244.13 Wed Oct 24 07:26:45 CEST 2007
225 - add Coro::Storable::blocking_thaw.
226 - use a vastly more complicated technique to localise
227 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
228 - use a coroutine for the idle callback Coro::Event,
229 instead of running Event in the current coroutine context.
230 This also catches recursive invocations.
231 - actually report fork errors in gethostbyname and inet_aton.
232
2334.11 Thu Oct 11 02:40:24 CEST 2007
234 - port to threaded perls.
235
2364.1 Thu Oct 11 02:38:16 CEST 2007
237 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
238 be local to each coro (see Coro::State).
239 - incompatible change: for very deep reasons, cede and cede_notself
240 cannot return anything, so nothing will be returned.
241 - possibly bring back 5.10 compatibility (untested).
242 - work around stupid (and wrong) warning on 5.10 :(.
243 - overlay the saved state over the context stack. This saves
244 a few hundred bytes per coroutine on average and also
245 speeds up context switching a bit.
246 - further tune default stack sizes.
247 - (more) correctly calculate stack usage in coro_rss.
248 - Coro::Storable::blocking_* did not properly lock
249 resulting in races between coroutines.
250 - added Coro::Storable::guard.
251 - stopping to trace a coroutine could destroy the cctx of
252 an unrelated coroutine.
253 - explain the relationship between Perl and C coroutines in
254 more detail in Coro::State.
255 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
256 causing a fork per resolve. This also affected Coro::Socket.
257 - switch to a separate stack in $coro->call/eval to avoid
258 invalidating pointers.
259
2604.03 Sat Oct 6 21:24:00 CEST 2007
261 - added Coro::throw method.
262 - minor code cleanups.
263
2644.02 Sat Oct 6 02:36:47 CEST 2007
265 - fix a very minor per-coroutine memleak (a single codereference).
266 - fixed a bug where the currently in-use c context would be freed
267 prematurely (can happen only when programs change the stacksize
268 or use tracing).
269 - tracing can no longer keep a coro alive after it terminated.
270 - do static branch prediction in the common path for gcc. gives
271 about 2-5% speed improvement here.
272
2734.01 Fri Oct 5 22:10:49 CEST 2007
274 - instead of recreating *a* standard output handle we simply
275 use STDOUT, which is faster and hopefully more robust.
276
2774.0 Fri Oct 5 12:56:00 CEST 2007
278 - incompatibly changed Coro::Storable::freeze.
279 - major new feature: added Coro::Debug, for interactive coroutine
280 debugging, tracing and much more.
281 - major bug fix: unbelievable, but true: $_, $/ and many other
282 "saved" variables actually weren't being saved. This has been fixed,
283 of course, while increasing performance while losing all the save
284 flags.
285 - save flags are gone, and all the api functions dealing with them.
286 - added Coro::Semaphore::adjust.
287 - added Coro::Util::fork_eval.
288 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
289 - added Coro::killall.
290 - reduce initial stack sizes to allow for "micro-coroutines".
291 - better async_pool resource management, moved parts of async_pool
292 handling to XS (major speed improvement).
293 - actually croak before modifying important data structures.
294 - refuse to transfer while compiling.
295 - possibly support eval EXPR better now.
296 - enable assembly per default on linux+bsd x86+amd64.
297 - all internal members were renamed _something for easier subclassing.
298 - many minor tweaks.
299
3003.63 Wed May 16 14:10:06 CEST 2007
301 - implement handcoded assembly for x86/amd64 SVR ABI.
302
3033.62 Fri Apr 27 21:36:06 CEST 2007
304 - upgrade libcoro (which might set unwind info correctly).
305 - change default on linux to setjmp/longjmp.
306
3073.61 Thu Apr 19 12:36:18 CEST 2007
308 - Coro::Storable caused an endless loop when thawing invalid
309 pst's sometimes.
310 - use a Semaphore in Coro::Storable, as Storable doesn't
311 seem to be reentrant (although it is documented to
312 be threadsafe...).
313 - fix Coro::Signal to bring back the original unreliable
314 but stateful semantics.
315 - fixed a lot of typos in Coro.pm (patch submitted by David
316 Steinbrunner, which applied flawlessly).
317
3183.6 Sat Apr 14 17:13:31 CEST 2007
319 - added some bugfixes to get eg/myhttpd working again.
320 - added Coro::Storable for often-cede'ing freeze/thaw.
321 - try to do a clean exit when a coroutine calls exit
322 (EXPERIMENTAL).
323 - got rid of indirect call through _coro_init.
324 - updated the partly antique examples in eg/ to
325 work again and be a bit less magic, too.
326 - fixed Coro::Signal semantics to work as documented again.
327
3283.55 Sun Mar 25 01:20:47 CET 2007
329 - add SAVE_DEFFH to save the default filehandle and enable
330 it by default.
331 - finally move socket-operations from Coro::Socket to Coro::Handle
332 to be able to unblock foreign sockets.
333 - add Coro::State::save_also and guarded_save.
334 - add count accessor to Coro::Semaphore.
335 - add Coro::State::cctx_stacksize.
336 - just for the fun of it, do not rely on implicit context,
337 which can dramatically improve performance, but people
338 using windows-process-emulation perls do not care much about
339 performance.
340
3413.51 Sun Mar 4 14:18:04 CET 2007
342 - fixed a problem when you weakened references to Coro::State's
343 (patch by Yuval Kogman).
344
3453.501 Wed Feb 28 12:44:07 CET 2007
346 - rename some global symbols as macosx from hell redefines
347 them without asking.
348
3493.5 Tue Feb 13 20:22:53 CET 2007
350 - do AnyEvent model detection earlier, avoiding problems
351 caused by first using AnyEvent and later Coro::Event.
352 - implement and document Coro::Event event objects.
353 - fix a potential problem in Coro::Event causing crashes.
354 - initialise PL_comppad when creating a new coroutine,
355 avoids crashes on early coro destruction.
356
3573.41 Mon Jan 22 19:19:49 CET 2007
358 - readline on Coro::Handle did not support undefined $/,
359 nor did it deliver partial lines on EOF or error.
360 - implement malloc fallback for stack allocation because
361 stupid broken idiotic OSX has a stupid broken
362 idiotic fits-the-whole-os mmap "implementation" and
363 my dick feels longer if Coro is portable even to
364 obsolete platforms.
365
3663.4 Fri Jan 19 21:52:54 CET 2007
367 - remove t/09_timer.t, as it isn't really testing much
368 but was rather flaky in practise.
369 - async_pool coro would keep arguments and callback alive until
370 it was reused.
371 - cancellation of a coroutine could cause spurious idle calls
372 in cede_notself.
373
3743.3 Sat Jan 6 03:45:00 CET 2007
375 - implement $coro->on_destroy.
376 - Coro::Event blocking semantics have been changed,
377 documented and - hopefully - improved.
378 - fix nice adding, not subtracting, from priority.
379 - fix ->prio and api_is_ready (patch by Mark Hinds).
380 - fixed an assert ("... == prev__cctx->idle_te")
381 that could errronously trigger.
382 - fix various large and small memleaks.
383 - use a (hopefully) more stable cancel implementation
384 that immediately frees the coroutine data.
385 - cede/cede_notself return a status now.
386 - added Coro::guard function.
387 - added a global coroutine pool for jobs (on my machine,
388 I can create and execute 48k simple coros/s with async,
389 and 128k coros with async_pool).
390 - Coro::AIO now uses the coroutine priority as io priority.
391
3923.2 Fri Dec 22 05:07:09 CET 2006
393 - improve portability to slightly older perls.
394 - use cleaner coroutine destruction.
395 - simplify configuration for users.
396 - optionally (unrecommended) prefer perl functions over
397 their coro replacements.
398
3993.11 Tue Dec 5 13:11:24 CET 2006
400 - fixed some bogus assert's, but as perl.h disables assert even
401 without NDEBUG (thank you very much), not too many people should
402 notice (that did include myself). Andreas König noticed, though :)
403 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
404 it. Also noticed by Andreas König :)
405 - save/restore tainted status.
406 - verified to pass the testsuite on my 5.9.5.
407
4083.1 Mon Dec 4 23:03:40 CET 2006
409 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
410 - incompatible change: transfer flags are now per-state.
411 - give a better error message on deadlock.
412 - document Coro::nready.
413 - enhanced testsuite.
414
4153.01 Sun Dec 3 23:47:42 CET 2006
416 - forgot to include Coro::Timer.
417
4183.0 Sun Dec 3 22:57:25 CET 2006
419 - the "FINALLY COMPLETELY STABLE" release (coming soon:
420 the "FAMOUS LAST WORDS" release).
4 - reimplement a new stack sharing algorithm. 421 - implement a new stack sharing algorithm, which uses a stack
422 pool (size currently hardcoded).
5 - make stack sharing mandatory (it no longer uses a heuristic). 423 - make stack sharing mandatory (it no longer uses a heuristic).
6 - eval/die no longer cause weird problems under heavy use. 424 - eval/die no longer cause weird problems under heavy use.
7 - Coro::Event could cause livelocks if it was used but 425 - Coro::Event could cause livelocks if it was used but
8 no Coro::Event watchers were used. 426 no Coro::Event watchers were used.
427 - Coro::Event now uses asynccheck as prepare does not
428 check for changed watchers.
429 - Coro::Event allows multiple waiting coros and will wake up one
430 per event.
9 - Coro::Event should be cleaner and more efficient now. 431 - Coro::Event should be cleaner and more efficient now.
432 - new utility function Coro::unblock_sub.
433 - document the sad fact that Event is no longer reentrant.
434 - putting a coroutine into the ready queue twice could under
435 some circumstances lead to stack corruption.
10 - minor incompatibility: subclassing Coro::State is supported 436 - minor incompatibility: subclassing Coro::State is supported
11 directly now without going through a _coro_state member. 437 directly now without going through a _coro_state member.
12 - state/coro switching is much faster now. 438 - state/coro switching is much faster now.
13 - very minor optimisations and code/documentation cleanup. 439 - very minor optimisations and code/documentation cleanup.
14 - avoid problems due to compiler inlining. 440 - avoid problems due to compiler inlining.
15 - removed timers from Coro::Timer -> use AnyEvent instead. 441 - removed timers from Coro::Timer -> use AnyEvent instead.
16 - replaced Coro::idle coroutine by (cleaner) idle handler. 442 - replaced Coro::idle coroutine by (cleaner) idle handler.
17 - updated to newest libcoro. 443 - updated to newest libcoro.
18 - implement enhanced support for valgrind. 444 - implement enhanced support for valgrind.
445 - implement is_ready and return value for ready.
446 - removed Coro::Cont, it was a misnomer (it's generators),
447 and it was somewhat annoying to get it right. Will come back
448 if somebody asks for it :)
449 - many ->wait methods and Coro::Event could return
450 spuriously without the event having happened.
19 451
202.5 Tue Nov 7 12:22:33 CET 2006 4522.5 Tue Nov 7 12:22:33 CET 2006
21 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket 453 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
22 use AnyEvent, moved them to Coro/. 454 use AnyEvent, moved them to Coro/.
23 - added Coro::LWP which contains all the uglyness required to 455 - added Coro::LWP which contains all the uglyness required to

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines