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

Comparing Coro/Changes (file contents):
Revision 1.136 by root, Fri Nov 3 00:56:02 2006 UTC vs.
Revision 1.398 by root, Sun Nov 9 23:08:49 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: possibly convert some of the Event modules into AnyEvent users. 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)).
4 7
5 - should compile with perl 5.9.x now. 84.91
9 - typo: new coroutines didn't get created with a zero flags field.
10 - clean up cctx creation code a bit.
11
124.9 Sat Nov 8 17:45:27 CET 2008
13 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
14 - (libcoro) no longer leak threads in the experimental pthread backend,
15 also speed it up considerably.
16 - (libcoro) do not rely on makecontext passing void *'s unscathed.
17 - fix compiletime dependencies on libcoro in the Makefile.
18 - cctx_count wasn't always updated properly.
19 - Coro::State::cctx_stacksize wasn't applied correctly.
20 - new function Coro::State::cctx_max_idle.
21 - the default max number of idle C contexts is now 4.
22 - (libcoro) try harder to get _setjmp/_longjmp.
23 - (libcoro) cleanup and extend the libcoro API to officially
24 allow the creation of empty source contexts.
25 - very experimental workaround for the totally broken netbsd platform.
26 - tried to hack around openbsd bugs.
27
284.804 Wed Nov 5 16:36:00 CET 2008
29 - Coro::Debug::new_unix_server would not create a non-blocking listening
30 socket, sometimes causing freezes.
31 - (libcoro) fix misaligned stack points for setjmp and assembly
32 methods, which can cause crashes on x86/x86_64 with a sufficiently
33 aggressive compiler.
34 - new function: Coro::Debug::new_tcp_server.
35 - move ->throw into the Coro class because it only works on coro objects.
36
374.803 Mon Nov 3 17:16:12 CET 2008
38 - (libcoro) use a global asm statement to become independent of gcc
39 otpimisations for CORO_ASM (thanks to pippijn for the idea).
40 - try to workaround yet another broken bsd, this time dragonfly.
41
424.802 Thu Oct 30 10:56:12 CET 2008
43 - support -fno-omit-frame-pointer on x86 with the assembly method.
44 - tune 01_unblock.t tests a bit.
45
464.801 Wed Oct 22 18:33:37 CEST 2008
47 - improve readline speed for very long "lines".
48 - backport to 5.8.8.
49
504.8 Thu Oct 2 13:34:40 CEST 2008
51 - new function Coro::AIO::aio_wait.
52 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
53 - greatly speed up and reduce memory usage of Coro::AIO requests.
54 - implement some other µ-optimisations.
55
564.749 Mon Sep 29 14:40:12 CEST 2008
57 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
58
594.748 Sat Sep 27 14:03:19 CEST 2008
60 - implement, but do not document, PerlIO::cede(granularity).
61 - Coro::Storable forgot to wrap Storable::pstore.
62 - work around the multitude of leaks and memory corruption
63 bugs in PerlIO::via by using our own C-level perliol. As a side
64 effect, Coro::Storable is now much, much, much faster.
65
664.747 Tue Sep 23 01:59:41 CEST 2008
67 - fix a per-cv memleak (one empty array was leaked per
68 code reference).
69 - avoid a crash in coro->call|rss when the coroutine was already
70 destroyed (most noticably when using Coro::Debug::ps :)
71 - also protect *Storable::FILE.
72 - push up default storable granularity to 20ms.
73
744.746 Sun Sep 21 03:22:20 CEST 2008
75 - be more insistent on locking Storable against reentrancy
76 in Coro::Storable.
77 - move swap_def?v and throw to Coro::State, as documented.
78
794.745 Thu Jul 24 00:14:38 CEST 2008
80 - remove debugging code related to MgPV_nolen_const, also try to
81 make it compile with perl 5.8.6 (yes, apple apparently loves
82 outdated software). Reported by John S.
83
844.744 Tue Jul 8 22:06:35 CEST 2008
85 - correctly provide default DIE/WARN handlers as documented.
86 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
87 implementations inside perl work fine for us.
88
894.743 Mon Jun 16 00:21:57 CEST 2008
90 - when using Coro::EV without running EV::loop it could
91 result in busy-waiting for events, this has been fixed.
92 - reduce codesize and improve performance by using EV_DEFAULT_UC.
93
944.742 Sat May 31 14:10:21 CEST 2008
95 - implement a workaround for (some) perl <5.8.8 versions.
96 - require EV 3.3+.
97
984.741 Fri May 30 23:33:09 CEST 2008
99 - tell netbsd how utterly broken their imitation of an OS is
100 and refuse to build by default if pthreads are in use.
101 - switch to "s" method on all bsd's by default, as their ucontext
102 stuff seems just too broken.
103 - fix a bug in Coro::Select.
104
1054.74 Thu May 29 20:05:31 CEST 2008
106 - do not test Coro::LWP for lack of dependencies.
107 - do not test Coro::Select for lack of working perls.
108
1094.73 Thu May 29 2008
110 - fix a bug in Coro::EV which would cause it to block despite
111 there being runnable coroutines.
112 - sprinkle "no warnings" freely over everything, also suppress
113 warnings for some other modules.
114 - fix typo in WSAEWOULDBLOCK.
115
1164.72 Sun May 25 05:14:36 CEST 2008
117 - tweak META.yaml a bit, unfortunately, there is no documented way
118 to have optional dependencies with CPAN. doh :(
119 - avoid running some tests on windows because they would fail due to
120 perl bug (broken fork, broken pipes...).
121 - work around perl on windows bugs where perl returns undocumented
122 error codes for sysread, syswrite etc. by taking advantage
123 of AnyEvent's workaround for that problem.
124 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
125 a common perl implementation bug on windows.
126 - use unix domain sockets in testsuite to work around
127 common perl implementation bugs on widows (they are emulated by
128 tcp sockets on windows. ugh.)
129
1304.71 Sat May 24 20:01:27 CEST 2008
131 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
132 - take advantage of async name resolution of AnyEvent::Util.
133 - work around brutal inet_aton override in Coro::LWP.
134 - take advantage of the readyhook in Coro::EV, for smoother
135 scheduling.
136
1374.7 Sun May 11 00:32:19 CEST 2008
138 - completely reworked the Coro manpage.
139 - added Coro::AnyEvent, generic event loop integration.
140 - implement cancel, ready and kill commands in Coro::Debug.
141 - document find_coro in Coro::Debug.
142 - incompatible change: rename has_stack to has_cctx.
143 - Coro::AIO and Coro::BDB no longer force event model detection,
144 use AnyEvent::AIO and AnyEvent::BDB.
145
1464.6 Sat Apr 26 10:05:14 CEST 2008
147 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
148 by the Coro module. It was a mistake to have it in the first place.
149 - (experimental) support for activestate perl 5.10 (method "w").
150 - (experimental) support for strawberry perl (method "a").
151 - coro_sigelem_set did not return a value although it had to,
152 actual impact unknown.
153
1544.51 Mon Apr 14 13:28:27 CEST 2008
155 - make it compile again on 5.8.
156
1574.50 Thu Apr 10 09:43:17 CEST 2008
158 - I did it twice! (see 4.49).
159
1604.49 Sun Apr 6 21:23:31 CEST 2008
161 - grr, instead of compiling the recent changes on 5.10 only they
162 were compiled on 5.8 only.
163
1644.48 Sun Apr 6 20:36:46 CEST 2008
165 - allow coroutine switches during eval's under 5.10.x, as it seems
166 the parser is a per-interpreter option now, so this is safe
167 (this might fix the odd crashing bug).
168 - drop support for 5.9.x versions: they are dead, jim.
169
1704.47 Sun Apr 6 00:37:52 CEST 2008
171 - force cctx allocation on API calls: we know we need to force one
172 and gcc actually manages to confuse our heuristic nowadays,
173 leading to crashes and worse.
174 - document force_cctx.
175
1764.46 Fri Apr 4 22:05:43 CEST 2008
177 - upgrade libcoro, resulting in pthread-backend (which was only created
178 to fulfill the rules of the programming languages shootout).
179
1804.45 Thu Mar 13 11:55:36 CET 2008
181 - fix a file leaking bug in eg/mhyttpd that would allow
182 downloading of any file (reported by oesi).
183 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
184 (also add testcase).
185 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
186 apparently having loads of fun with that).
187 - implement $state->swap_defsv, ->swap_defav and document ->throw.
188
1894.4 Wed Feb 13 16:44:29 CET 2008
190 - only assume .cfi pseudo isns exist on GNU/Linux.
191 - add get_timed method to Coro::Channels.
192 - fixed Coro::Debug manpage.
193 - perl 5.11 compatbility improvement based on patch by
194 Andreas König.
195
1964.37 Sun Jan 20 11:25:23 CET 2008
197 - soften the check for an active parser for perl 5.10 (avoiding
198 "Coro::State::transfer called while parsing" in many cases).
199
2004.36 Sun Jan 13 10:53:56 CET 2008
201 - reset diehook when terminating from an async_pool as to not
202 trigger any __DIE__ handlers.
203
2044.35 Sun Jan 13 04:14:13 CET 2008
205 - "bt" debug command now displays any exceptions
206 from longmess and also skips the innermost
207 stackframes, giving more useufl output.
208 - allow backtraces in coroutines blocked in Coro::EV,
209 at a <1% speed hit.
210 - handle localising of $SIG{__WARN__} and __DIE__
211 properly (with a proper amount of dirty hacking).
212
2134.34 Sat Dec 22 17:49:53 CET 2007
214 - upgrade to EV version 2.0 API.
215
2164.33 Mon Dec 17 08:36:12 CET 2007
217 - make Coro::AIO etc. loadable in the absence of EV.
218
2194.32 Mon Dec 17 07:46:02 CET 2007
220 - majorly improved Coro::Handle's performance with Coro::EV.
221 - implemented the until now mythical Coro::BDB module.
222 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
223
2244.31 Wed Dec 5 12:32:39 CET 2007
225 - remove warn statement form Coro::Util.
226
2274.3 Tue Dec 4 20:33:14 CET 2007
228 - calls to the idle function could cause stack corruption
229 when the stack changed.
230 - do no longer rely on the presence of EV::DNS if EV is used
231 (because it is gone), but instead take avdantage of EV::ADNS
232 if available.
233 - add ($) prototypes to all functions in Coro::Storable.
234 - use a conventional (safer) idle callback in Coro::EV.
235 - do accept quickly in Coro::Debug to avoid endless loops.
236
2374.22 Fri Nov 30 16:04:04 CET 2007
238 - really use optimised versions for Event and EV in Coro::Util
239 and Coro::Handle.
240
2414.21 Sun Nov 25 10:48:59 CET 2007
242 - fix a spurious memory read.
243 - Coro::EV no longer keeps the eventloop "alive".
244
2454.2 Fri Nov 9 20:47:05 CET 2007
246 - enable/disable tracing from a new coroutine, not a pooled one.
247 - fix a memleak in Coro::Event.
248 - removed killall call from fork_eval.
249 - made sure store_fd is already loaded so that fork_eval
250 does not have to parse autoload in each subprocess.
251 - only use assembly method if -O switch is in $Config{optimize}.
252 - add (optional) Coro::EV module, so far the best event loop module
253 directly supported by Coro.
254 - if the event model is EV, use EV::DNS to resolve
255 stuff in Coro::Util.
256 - don't get confused by multiple event notifications in Coro::Handle.
257 - initial support for EV (libevent interface).
258 - require Event and EV using configure_requires, to force their existance.
259
2604.13 Wed Oct 24 07:26:45 CEST 2007
261 - add Coro::Storable::blocking_thaw.
262 - use a vastly more complicated technique to localise
263 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
264 - use a coroutine for the idle callback Coro::Event,
265 instead of running Event in the current coroutine context.
266 This also catches recursive invocations.
267 - actually report fork errors in gethostbyname and inet_aton.
268
2694.11 Thu Oct 11 02:40:24 CEST 2007
270 - port to threaded perls.
271
2724.1 Thu Oct 11 02:38:16 CEST 2007
273 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
274 be local to each coro (see Coro::State).
275 - incompatible change: for very deep reasons, cede and cede_notself
276 cannot return anything, so nothing will be returned.
277 - possibly bring back 5.10 compatibility (untested).
278 - work around stupid (and wrong) warning on 5.10 :(.
279 - overlay the saved state over the context stack. This saves
280 a few hundred bytes per coroutine on average and also
281 speeds up context switching a bit.
282 - further tune default stack sizes.
283 - (more) correctly calculate stack usage in coro_rss.
284 - Coro::Storable::blocking_* did not properly lock
285 resulting in races between coroutines.
286 - added Coro::Storable::guard.
287 - stopping to trace a coroutine could destroy the cctx of
288 an unrelated coroutine.
289 - explain the relationship between Perl and C coroutines in
290 more detail in Coro::State.
291 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
292 causing a fork per resolve. This also affected Coro::Socket.
293 - switch to a separate stack in $coro->call/eval to avoid
294 invalidating pointers.
295
2964.03 Sat Oct 6 21:24:00 CEST 2007
297 - added Coro::throw method.
298 - minor code cleanups.
299
3004.02 Sat Oct 6 02:36:47 CEST 2007
301 - fix a very minor per-coroutine memleak (a single codereference).
302 - fixed a bug where the currently in-use c context would be freed
303 prematurely (can happen only when programs change the stacksize
304 or use tracing).
305 - tracing can no longer keep a coro alive after it terminated.
306 - do static branch prediction in the common path for gcc. gives
307 about 2-5% speed improvement here.
308
3094.01 Fri Oct 5 22:10:49 CEST 2007
310 - instead of recreating *a* standard output handle we simply
311 use STDOUT, which is faster and hopefully more robust.
312
3134.0 Fri Oct 5 12:56:00 CEST 2007
314 - incompatibly changed Coro::Storable::freeze.
315 - major new feature: added Coro::Debug, for interactive coroutine
316 debugging, tracing and much more.
317 - major bug fix: unbelievable, but true: $_, $/ and many other
318 "saved" variables actually weren't being saved. This has been fixed,
319 of course, while increasing performance while losing all the save
320 flags.
321 - save flags are gone, and all the api functions dealing with them.
322 - added Coro::Semaphore::adjust.
323 - added Coro::Util::fork_eval.
324 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
325 - added Coro::killall.
326 - reduce initial stack sizes to allow for "micro-coroutines".
327 - better async_pool resource management, moved parts of async_pool
328 handling to XS (major speed improvement).
329 - actually croak before modifying important data structures.
330 - refuse to transfer while compiling.
331 - possibly support eval EXPR better now.
332 - enable assembly per default on linux+bsd x86+amd64.
333 - all internal members were renamed _something for easier subclassing.
334 - many minor tweaks.
335
3363.63 Wed May 16 14:10:06 CEST 2007
337 - implement handcoded assembly for x86/amd64 SVR ABI.
338
3393.62 Fri Apr 27 21:36:06 CEST 2007
340 - upgrade libcoro (which might set unwind info correctly).
341 - change default on linux to setjmp/longjmp.
342
3433.61 Thu Apr 19 12:36:18 CEST 2007
344 - Coro::Storable caused an endless loop when thawing invalid
345 pst's sometimes.
346 - use a Semaphore in Coro::Storable, as Storable doesn't
347 seem to be reentrant (although it is documented to
348 be threadsafe...).
349 - fix Coro::Signal to bring back the original unreliable
350 but stateful semantics.
351 - fixed a lot of typos in Coro.pm (patch submitted by David
352 Steinbrunner, which applied flawlessly).
353
3543.6 Sat Apr 14 17:13:31 CEST 2007
355 - added some bugfixes to get eg/myhttpd working again.
356 - added Coro::Storable for often-cede'ing freeze/thaw.
357 - try to do a clean exit when a coroutine calls exit
358 (EXPERIMENTAL).
359 - got rid of indirect call through _coro_init.
360 - updated the partly antique examples in eg/ to
361 work again and be a bit less magic, too.
362 - fixed Coro::Signal semantics to work as documented again.
363
3643.55 Sun Mar 25 01:20:47 CET 2007
365 - add SAVE_DEFFH to save the default filehandle and enable
366 it by default.
367 - finally move socket-operations from Coro::Socket to Coro::Handle
368 to be able to unblock foreign sockets.
369 - add Coro::State::save_also and guarded_save.
370 - add count accessor to Coro::Semaphore.
371 - add Coro::State::cctx_stacksize.
372 - just for the fun of it, do not rely on implicit context,
373 which can dramatically improve performance, but people
374 using windows-process-emulation perls do not care much about
375 performance.
376
3773.51 Sun Mar 4 14:18:04 CET 2007
378 - fixed a problem when you weakened references to Coro::State's
379 (patch by Yuval Kogman).
380
3813.501 Wed Feb 28 12:44:07 CET 2007
382 - rename some global symbols as macosx from hell redefines
383 them without asking.
384
3853.5 Tue Feb 13 20:22:53 CET 2007
386 - do AnyEvent model detection earlier, avoiding problems
387 caused by first using AnyEvent and later Coro::Event.
388 - implement and document Coro::Event event objects.
389 - fix a potential problem in Coro::Event causing crashes.
390 - initialise PL_comppad when creating a new coroutine,
391 avoids crashes on early coro destruction.
392
3933.41 Mon Jan 22 19:19:49 CET 2007
394 - readline on Coro::Handle did not support undefined $/,
395 nor did it deliver partial lines on EOF or error.
396 - implement malloc fallback for stack allocation because
397 stupid broken idiotic OSX has a stupid broken
398 idiotic fits-the-whole-os mmap "implementation" and
399 my dick feels longer if Coro is portable even to
400 obsolete platforms.
401
4023.4 Fri Jan 19 21:52:54 CET 2007
403 - remove t/09_timer.t, as it isn't really testing much
404 but was rather flaky in practise.
405 - async_pool coro would keep arguments and callback alive until
406 it was reused.
407 - cancellation of a coroutine could cause spurious idle calls
408 in cede_notself.
409
4103.3 Sat Jan 6 03:45:00 CET 2007
411 - implement $coro->on_destroy.
412 - Coro::Event blocking semantics have been changed,
413 documented and - hopefully - improved.
414 - fix nice adding, not subtracting, from priority.
415 - fix ->prio and api_is_ready (patch by Mark Hinds).
416 - fixed an assert ("... == prev__cctx->idle_te")
417 that could errronously trigger.
418 - fix various large and small memleaks.
419 - use a (hopefully) more stable cancel implementation
420 that immediately frees the coroutine data.
421 - cede/cede_notself return a status now.
422 - added Coro::guard function.
423 - added a global coroutine pool for jobs (on my machine,
424 I can create and execute 48k simple coros/s with async,
425 and 128k coros with async_pool).
426 - Coro::AIO now uses the coroutine priority as io priority.
427
4283.2 Fri Dec 22 05:07:09 CET 2006
429 - improve portability to slightly older perls.
430 - use cleaner coroutine destruction.
431 - simplify configuration for users.
432 - optionally (unrecommended) prefer perl functions over
433 their coro replacements.
434
4353.11 Tue Dec 5 13:11:24 CET 2006
436 - fixed some bogus assert's, but as perl.h disables assert even
437 without NDEBUG (thank you very much), not too many people should
438 notice (that did include myself). Andreas König noticed, though :)
439 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
440 it. Also noticed by Andreas König :)
441 - save/restore tainted status.
442 - verified to pass the testsuite on my 5.9.5.
443
4443.1 Mon Dec 4 23:03:40 CET 2006
445 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
446 - incompatible change: transfer flags are now per-state.
447 - give a better error message on deadlock.
448 - document Coro::nready.
449 - enhanced testsuite.
450
4513.01 Sun Dec 3 23:47:42 CET 2006
452 - forgot to include Coro::Timer.
453
4543.0 Sun Dec 3 22:57:25 CET 2006
455 - the "FINALLY COMPLETELY STABLE" release (coming soon:
456 the "FAMOUS LAST WORDS" release).
457 - implement a new stack sharing algorithm, which uses a stack
458 pool (size currently hardcoded).
459 - make stack sharing mandatory (it no longer uses a heuristic).
460 - eval/die no longer cause weird problems under heavy use.
461 - Coro::Event could cause livelocks if it was used but
462 no Coro::Event watchers were used.
463 - Coro::Event now uses asynccheck as prepare does not
464 check for changed watchers.
465 - Coro::Event allows multiple waiting coros and will wake up one
466 per event.
467 - Coro::Event should be cleaner and more efficient now.
468 - new utility function Coro::unblock_sub.
469 - document the sad fact that Event is no longer reentrant.
470 - putting a coroutine into the ready queue twice could under
471 some circumstances lead to stack corruption.
472 - minor incompatibility: subclassing Coro::State is supported
473 directly now without going through a _coro_state member.
474 - state/coro switching is much faster now.
475 - very minor optimisations and code/documentation cleanup.
476 - avoid problems due to compiler inlining.
477 - removed timers from Coro::Timer -> use AnyEvent instead.
478 - replaced Coro::idle coroutine by (cleaner) idle handler.
479 - updated to newest libcoro.
480 - implement enhanced support for valgrind.
481 - implement is_ready and return value for ready.
482 - removed Coro::Cont, it was a misnomer (it's generators),
483 and it was somewhat annoying to get it right. Will come back
484 if somebody asks for it :)
485 - many ->wait methods and Coro::Event could return
486 spuriously without the event having happened.
487
4882.5 Tue Nov 7 12:22:33 CET 2006
489 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
490 use AnyEvent, moved them to Coro/.
491 - added Coro::LWP which contains all the uglyness required to
492 make LWP non-blocking.
493 - should work with perl 5.9.x now (Andreas König made me do it).
494 - fixed another bug in Coro::Select when the timeout was undef.
6 - reuse PL_start_env for all coros, saves some memory per coroutine. 495 - reuse PL_start_env for all coros, saves some memory per coroutine.
7 - manage PL_top_env differently, hopefully to avoid panic: top_env. 496 - manage PL_top_env differently, hopefully to avoid panic: top_env.
497 - timeout argument was not properly used in Coro::Socket.
498 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
499 - emulate undocumented(!) functionality of IO::Socket required
500 by LWP(!!).
501 - updated eg/lwp to work with newer lwp's.
8 - remove "FATAL: uncaught exception" prefix: coroutines that die 502 - remove "FATAL: uncaught exception" prefix. Coroutines that die
9 kill the whole process, just as exceptions in the main coroutine 503 kill the whole process, just as exceptions in the main "coroutine"
10 do. 504 did already.
11 505
122.1 Wed Nov 1 23:01:13 CET 2006 5062.1 Wed Nov 1 23:01:13 CET 2006
13 - fix a long-standing bug in Coro::Select where select with 507 - fix a long-standing bug in Coro::Select where select with
14 zero timeout would instead change the current default filehandle. 508 zero timeout would instead change the current default filehandle.
15 - use a simpler and hopefully more robust way to clone padlists 509 - use a simpler and hopefully more robust way to clone padlists

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines