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

Comparing Coro/Changes (file contents):
Revision 1.122 by root, Fri Dec 16 01:44:11 2005 UTC vs.
Revision 1.388 by root, Tue Nov 4 12:08:44 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
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 - move ->throw into the Coro class because it only works on coro objects.
9
104.803 Mon Nov 3 17:16:12 CET 2008
11 - use a global asm statement to become independent of gcc otpimisations
12 for CORO_ASM (thanks to pippijn for the idea).
13 - try to workaround yet another broken bsd, this time dragonfly.
14
154.802 Thu Oct 30 10:56:12 CET 2008
16 - support -fno-omit-frame-pointer on x86 with the assembly method.
17 - tune 01_unblock.t tests a bit.
18
194.801 Wed Oct 22 18:33:37 CEST 2008
20 - improve readline speed for very long "lines".
21 - backport to 5.8.8.
22
234.8 Thu Oct 2 13:34:40 CEST 2008
24 - new function Coro::AIO::aio_wait.
25 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
26 - greatly speed up and reduce memory usage of Coro::AIO requests.
27 - implement some other µ-optimisations.
28
294.749 Mon Sep 29 14:40:12 CEST 2008
30 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
31
324.748 Sat Sep 27 14:03:19 CEST 2008
33 - implement, but do not document, PerlIO::cede(granularity).
34 - Coro::Storable forgot to wrap Storable::pstore.
35 - work around the multitude of leaks and memory corruption
36 bugs in PerlIO::via by using our own C-level perliol. As a side
37 effect, Coro::Storable is now much, much, much faster.
38
394.747 Tue Sep 23 01:59:41 CEST 2008
40 - fix a per-cv memleak (one empty array was leaked per
41 code reference).
42 - avoid a crash in coro->call|rss when the coroutine was already
43 destroyed (most noticably when using Coro::Debug::ps :)
44 - also protect *Storable::FILE.
45 - push up default storable granularity to 20ms.
46
474.746 Sun Sep 21 03:22:20 CEST 2008
48 - be more insistent on locking Storable against reentrancy
49 in Coro::Storable.
50 - move swap_def?v and throw to Coro::State, as documented.
51
524.745 Thu Jul 24 00:14:38 CEST 2008
53 - remove debugging code related to MgPV_nolen_const, also try to
54 make it compile with perl 5.8.6 (yes, apple apparently loves
55 outdated software). Reported by John S.
56
574.744 Tue Jul 8 22:06:35 CEST 2008
58 - correctly provide default DIE/WARN handlers as documented.
59 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
60 implementations inside perl work fine for us.
61
624.743 Mon Jun 16 00:21:57 CEST 2008
63 - when using Coro::EV without running EV::loop it could
64 result in busy-waiting for events, this has been fixed.
65 - reduce codesize and improve performance by using EV_DEFAULT_UC.
66
674.742 Sat May 31 14:10:21 CEST 2008
68 - implement a workaround for (some) perl <5.8.8 versions.
69 - require EV 3.3+.
70
714.741 Fri May 30 23:33:09 CEST 2008
72 - tell netbsd how utterly broken their imitation of an OS is
73 and refuse to build by default if pthreads are in use.
74 - switch to "s" method on all bsd's by default, as their ucontext
75 stuff seems just too broken.
76 - fix a bug in Coro::Select.
77
784.74 Thu May 29 20:05:31 CEST 2008
79 - do not test Coro::LWP for lack of dependencies.
80 - do not test Coro::Select for lack of working perls.
81
824.73 Thu May 29 2008
83 - fix a bug in Coro::EV which would cause it to block despite
84 there being runnable coroutines.
85 - sprinkle "no warnings" freely over everything, also suppress
86 warnings for some other modules.
87 - fix typo in WSAEWOULDBLOCK.
88
894.72 Sun May 25 05:14:36 CEST 2008
90 - tweak META.yaml a bit, unfortunately, there is no documented way
91 to have optional dependencies with CPAN. doh :(
92 - avoid running some tests on windows because they would fail due to
93 perl bug (broken fork, broken pipes...).
94 - work around perl on windows bugs where perl returns undocumented
95 error codes for sysread, syswrite etc. by taking advantage
96 of AnyEvent's workaround for that problem.
97 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
98 a common perl implementation bug on windows.
99 - use unix domain sockets in testsuite to work around
100 common perl implementation bugs on widows (they are emulated by
101 tcp sockets on windows. ugh.)
102
1034.71 Sat May 24 20:01:27 CEST 2008
104 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
105 - take advantage of async name resolution of AnyEvent::Util.
106 - work around brutal inet_aton override in Coro::LWP.
107 - take advantage of the readyhook in Coro::EV, for smoother
108 scheduling.
109
1104.7 Sun May 11 00:32:19 CEST 2008
111 - completely reworked the Coro manpage.
112 - added Coro::AnyEvent, generic event loop integration.
113 - implement cancel, ready and kill commands in Coro::Debug.
114 - document find_coro in Coro::Debug.
115 - incompatible change: rename has_stack to has_cctx.
116 - Coro::AIO and Coro::BDB no longer force event model detection,
117 use AnyEvent::AIO and AnyEvent::BDB.
118
1194.6 Sat Apr 26 10:05:14 CEST 2008
120 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
121 by the Coro module. It was a mistake to have it in the first place.
122 - (experimental) support for activestate perl 5.10 (method "w").
123 - (experimental) support for strawberry perl (method "a").
124 - coro_sigelem_set did not return a value although it had to,
125 actual impact unknown.
126
1274.51 Mon Apr 14 13:28:27 CEST 2008
128 - make it compile again on 5.8.
129
1304.50 Thu Apr 10 09:43:17 CEST 2008
131 - I did it twice! (see 4.49).
132
1334.49 Sun Apr 6 21:23:31 CEST 2008
134 - grr, instead of compiling the recent changes on 5.10 only they
135 were compiled on 5.8 only.
136
1374.48 Sun Apr 6 20:36:46 CEST 2008
138 - allow coroutine switches during eval's under 5.10.x, as it seems
139 the parser is a per-interpreter option now, so this is safe
140 (this might fix the odd crashing bug).
141 - drop support for 5.9.x versions: they are dead, jim.
142
1434.47 Sun Apr 6 00:37:52 CEST 2008
144 - force cctx allocation on API calls: we know we need to force one
145 and gcc actually manages to confuse our heuristic nowadays,
146 leading to crashes and worse.
147 - document force_cctx.
148
1494.46 Fri Apr 4 22:05:43 CEST 2008
150 - upgrade libcoro, resulting in pthread-backend (which was only created
151 to fulfill the rules of the programming languages shootout).
152
1534.45 Thu Mar 13 11:55:36 CET 2008
154 - fix a file leaking bug in eg/mhyttpd that would allow
155 downloading of any file (reported by oesi).
156 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
157 (also add testcase).
158 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
159 apparently having loads of fun with that).
160 - implement $state->swap_defsv, ->swap_defav and document ->throw.
161
1624.4 Wed Feb 13 16:44:29 CET 2008
163 - only assume .cfi pseudo isns exist on GNU/Linux.
164 - add get_timed method to Coro::Channels.
165 - fixed Coro::Debug manpage.
166 - perl 5.11 compatbility improvement based on patch by
167 Andreas König.
168
1694.37 Sun Jan 20 11:25:23 CET 2008
170 - soften the check for an active parser for perl 5.10 (avoiding
171 "Coro::State::transfer called while parsing" in many cases).
172
1734.36 Sun Jan 13 10:53:56 CET 2008
174 - reset diehook when terminating from an async_pool as to not
175 trigger any __DIE__ handlers.
176
1774.35 Sun Jan 13 04:14:13 CET 2008
178 - "bt" debug command now displays any exceptions
179 from longmess and also skips the innermost
180 stackframes, giving more useufl output.
181 - allow backtraces in coroutines blocked in Coro::EV,
182 at a <1% speed hit.
183 - handle localising of $SIG{__WARN__} and __DIE__
184 properly (with a proper amount of dirty hacking).
185
1864.34 Sat Dec 22 17:49:53 CET 2007
187 - upgrade to EV version 2.0 API.
188
1894.33 Mon Dec 17 08:36:12 CET 2007
190 - make Coro::AIO etc. loadable in the absence of EV.
191
1924.32 Mon Dec 17 07:46:02 CET 2007
193 - majorly improved Coro::Handle's performance with Coro::EV.
194 - implemented the until now mythical Coro::BDB module.
195 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
196
1974.31 Wed Dec 5 12:32:39 CET 2007
198 - remove warn statement form Coro::Util.
199
2004.3 Tue Dec 4 20:33:14 CET 2007
201 - calls to the idle function could cause stack corruption
202 when the stack changed.
203 - do no longer rely on the presence of EV::DNS if EV is used
204 (because it is gone), but instead take avdantage of EV::ADNS
205 if available.
206 - add ($) prototypes to all functions in Coro::Storable.
207 - use a conventional (safer) idle callback in Coro::EV.
208 - do accept quickly in Coro::Debug to avoid endless loops.
209
2104.22 Fri Nov 30 16:04:04 CET 2007
211 - really use optimised versions for Event and EV in Coro::Util
212 and Coro::Handle.
213
2144.21 Sun Nov 25 10:48:59 CET 2007
215 - fix a spurious memory read.
216 - Coro::EV no longer keeps the eventloop "alive".
217
2184.2 Fri Nov 9 20:47:05 CET 2007
219 - enable/disable tracing from a new coroutine, not a pooled one.
220 - fix a memleak in Coro::Event.
221 - removed killall call from fork_eval.
222 - made sure store_fd is already loaded so that fork_eval
223 does not have to parse autoload in each subprocess.
224 - only use assembly method if -O switch is in $Config{optimize}.
225 - add (optional) Coro::EV module, so far the best event loop module
226 directly supported by Coro.
227 - if the event model is EV, use EV::DNS to resolve
228 stuff in Coro::Util.
229 - don't get confused by multiple event notifications in Coro::Handle.
230 - initial support for EV (libevent interface).
231 - require Event and EV using configure_requires, to force their existance.
232
2334.13 Wed Oct 24 07:26:45 CEST 2007
234 - add Coro::Storable::blocking_thaw.
235 - use a vastly more complicated technique to localise
236 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
237 - use a coroutine for the idle callback Coro::Event,
238 instead of running Event in the current coroutine context.
239 This also catches recursive invocations.
240 - actually report fork errors in gethostbyname and inet_aton.
241
2424.11 Thu Oct 11 02:40:24 CEST 2007
243 - port to threaded perls.
244
2454.1 Thu Oct 11 02:38:16 CEST 2007
246 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
247 be local to each coro (see Coro::State).
248 - incompatible change: for very deep reasons, cede and cede_notself
249 cannot return anything, so nothing will be returned.
250 - possibly bring back 5.10 compatibility (untested).
251 - work around stupid (and wrong) warning on 5.10 :(.
252 - overlay the saved state over the context stack. This saves
253 a few hundred bytes per coroutine on average and also
254 speeds up context switching a bit.
255 - further tune default stack sizes.
256 - (more) correctly calculate stack usage in coro_rss.
257 - Coro::Storable::blocking_* did not properly lock
258 resulting in races between coroutines.
259 - added Coro::Storable::guard.
260 - stopping to trace a coroutine could destroy the cctx of
261 an unrelated coroutine.
262 - explain the relationship between Perl and C coroutines in
263 more detail in Coro::State.
264 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
265 causing a fork per resolve. This also affected Coro::Socket.
266 - switch to a separate stack in $coro->call/eval to avoid
267 invalidating pointers.
268
2694.03 Sat Oct 6 21:24:00 CEST 2007
270 - added Coro::throw method.
271 - minor code cleanups.
272
2734.02 Sat Oct 6 02:36:47 CEST 2007
274 - fix a very minor per-coroutine memleak (a single codereference).
275 - fixed a bug where the currently in-use c context would be freed
276 prematurely (can happen only when programs change the stacksize
277 or use tracing).
278 - tracing can no longer keep a coro alive after it terminated.
279 - do static branch prediction in the common path for gcc. gives
280 about 2-5% speed improvement here.
281
2824.01 Fri Oct 5 22:10:49 CEST 2007
283 - instead of recreating *a* standard output handle we simply
284 use STDOUT, which is faster and hopefully more robust.
285
2864.0 Fri Oct 5 12:56:00 CEST 2007
287 - incompatibly changed Coro::Storable::freeze.
288 - major new feature: added Coro::Debug, for interactive coroutine
289 debugging, tracing and much more.
290 - major bug fix: unbelievable, but true: $_, $/ and many other
291 "saved" variables actually weren't being saved. This has been fixed,
292 of course, while increasing performance while losing all the save
293 flags.
294 - save flags are gone, and all the api functions dealing with them.
295 - added Coro::Semaphore::adjust.
296 - added Coro::Util::fork_eval.
297 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
298 - added Coro::killall.
299 - reduce initial stack sizes to allow for "micro-coroutines".
300 - better async_pool resource management, moved parts of async_pool
301 handling to XS (major speed improvement).
302 - actually croak before modifying important data structures.
303 - refuse to transfer while compiling.
304 - possibly support eval EXPR better now.
305 - enable assembly per default on linux+bsd x86+amd64.
306 - all internal members were renamed _something for easier subclassing.
307 - many minor tweaks.
308
3093.63 Wed May 16 14:10:06 CEST 2007
310 - implement handcoded assembly for x86/amd64 SVR ABI.
311
3123.62 Fri Apr 27 21:36:06 CEST 2007
313 - upgrade libcoro (which might set unwind info correctly).
314 - change default on linux to setjmp/longjmp.
315
3163.61 Thu Apr 19 12:36:18 CEST 2007
317 - Coro::Storable caused an endless loop when thawing invalid
318 pst's sometimes.
319 - use a Semaphore in Coro::Storable, as Storable doesn't
320 seem to be reentrant (although it is documented to
321 be threadsafe...).
322 - fix Coro::Signal to bring back the original unreliable
323 but stateful semantics.
324 - fixed a lot of typos in Coro.pm (patch submitted by David
325 Steinbrunner, which applied flawlessly).
326
3273.6 Sat Apr 14 17:13:31 CEST 2007
328 - added some bugfixes to get eg/myhttpd working again.
329 - added Coro::Storable for often-cede'ing freeze/thaw.
330 - try to do a clean exit when a coroutine calls exit
331 (EXPERIMENTAL).
332 - got rid of indirect call through _coro_init.
333 - updated the partly antique examples in eg/ to
334 work again and be a bit less magic, too.
335 - fixed Coro::Signal semantics to work as documented again.
336
3373.55 Sun Mar 25 01:20:47 CET 2007
338 - add SAVE_DEFFH to save the default filehandle and enable
339 it by default.
340 - finally move socket-operations from Coro::Socket to Coro::Handle
341 to be able to unblock foreign sockets.
342 - add Coro::State::save_also and guarded_save.
343 - add count accessor to Coro::Semaphore.
344 - add Coro::State::cctx_stacksize.
345 - just for the fun of it, do not rely on implicit context,
346 which can dramatically improve performance, but people
347 using windows-process-emulation perls do not care much about
348 performance.
349
3503.51 Sun Mar 4 14:18:04 CET 2007
351 - fixed a problem when you weakened references to Coro::State's
352 (patch by Yuval Kogman).
353
3543.501 Wed Feb 28 12:44:07 CET 2007
355 - rename some global symbols as macosx from hell redefines
356 them without asking.
357
3583.5 Tue Feb 13 20:22:53 CET 2007
359 - do AnyEvent model detection earlier, avoiding problems
360 caused by first using AnyEvent and later Coro::Event.
361 - implement and document Coro::Event event objects.
362 - fix a potential problem in Coro::Event causing crashes.
363 - initialise PL_comppad when creating a new coroutine,
364 avoids crashes on early coro destruction.
365
3663.41 Mon Jan 22 19:19:49 CET 2007
367 - readline on Coro::Handle did not support undefined $/,
368 nor did it deliver partial lines on EOF or error.
369 - implement malloc fallback for stack allocation because
370 stupid broken idiotic OSX has a stupid broken
371 idiotic fits-the-whole-os mmap "implementation" and
372 my dick feels longer if Coro is portable even to
373 obsolete platforms.
374
3753.4 Fri Jan 19 21:52:54 CET 2007
376 - remove t/09_timer.t, as it isn't really testing much
377 but was rather flaky in practise.
378 - async_pool coro would keep arguments and callback alive until
379 it was reused.
380 - cancellation of a coroutine could cause spurious idle calls
381 in cede_notself.
382
3833.3 Sat Jan 6 03:45:00 CET 2007
384 - implement $coro->on_destroy.
385 - Coro::Event blocking semantics have been changed,
386 documented and - hopefully - improved.
387 - fix nice adding, not subtracting, from priority.
388 - fix ->prio and api_is_ready (patch by Mark Hinds).
389 - fixed an assert ("... == prev__cctx->idle_te")
390 that could errronously trigger.
391 - fix various large and small memleaks.
392 - use a (hopefully) more stable cancel implementation
393 that immediately frees the coroutine data.
394 - cede/cede_notself return a status now.
395 - added Coro::guard function.
396 - added a global coroutine pool for jobs (on my machine,
397 I can create and execute 48k simple coros/s with async,
398 and 128k coros with async_pool).
399 - Coro::AIO now uses the coroutine priority as io priority.
400
4013.2 Fri Dec 22 05:07:09 CET 2006
402 - improve portability to slightly older perls.
403 - use cleaner coroutine destruction.
404 - simplify configuration for users.
405 - optionally (unrecommended) prefer perl functions over
406 their coro replacements.
407
4083.11 Tue Dec 5 13:11:24 CET 2006
409 - fixed some bogus assert's, but as perl.h disables assert even
410 without NDEBUG (thank you very much), not too many people should
411 notice (that did include myself). Andreas König noticed, though :)
412 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
413 it. Also noticed by Andreas König :)
414 - save/restore tainted status.
415 - verified to pass the testsuite on my 5.9.5.
416
4173.1 Mon Dec 4 23:03:40 CET 2006
418 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
419 - incompatible change: transfer flags are now per-state.
420 - give a better error message on deadlock.
421 - document Coro::nready.
422 - enhanced testsuite.
423
4243.01 Sun Dec 3 23:47:42 CET 2006
425 - forgot to include Coro::Timer.
426
4273.0 Sun Dec 3 22:57:25 CET 2006
428 - the "FINALLY COMPLETELY STABLE" release (coming soon:
429 the "FAMOUS LAST WORDS" release).
430 - implement a new stack sharing algorithm, which uses a stack
431 pool (size currently hardcoded).
432 - make stack sharing mandatory (it no longer uses a heuristic).
433 - eval/die no longer cause weird problems under heavy use.
434 - Coro::Event could cause livelocks if it was used but
435 no Coro::Event watchers were used.
436 - Coro::Event now uses asynccheck as prepare does not
437 check for changed watchers.
438 - Coro::Event allows multiple waiting coros and will wake up one
439 per event.
440 - Coro::Event should be cleaner and more efficient now.
441 - new utility function Coro::unblock_sub.
442 - document the sad fact that Event is no longer reentrant.
443 - putting a coroutine into the ready queue twice could under
444 some circumstances lead to stack corruption.
445 - minor incompatibility: subclassing Coro::State is supported
446 directly now without going through a _coro_state member.
447 - state/coro switching is much faster now.
448 - very minor optimisations and code/documentation cleanup.
449 - avoid problems due to compiler inlining.
450 - removed timers from Coro::Timer -> use AnyEvent instead.
451 - replaced Coro::idle coroutine by (cleaner) idle handler.
452 - updated to newest libcoro.
453 - implement enhanced support for valgrind.
454 - implement is_ready and return value for ready.
455 - removed Coro::Cont, it was a misnomer (it's generators),
456 and it was somewhat annoying to get it right. Will come back
457 if somebody asks for it :)
458 - many ->wait methods and Coro::Event could return
459 spuriously without the event having happened.
460
4612.5 Tue Nov 7 12:22:33 CET 2006
462 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
463 use AnyEvent, moved them to Coro/.
464 - added Coro::LWP which contains all the uglyness required to
465 make LWP non-blocking.
466 - should work with perl 5.9.x now (Andreas König made me do it).
467 - fixed another bug in Coro::Select when the timeout was undef.
468 - reuse PL_start_env for all coros, saves some memory per coroutine.
469 - manage PL_top_env differently, hopefully to avoid panic: top_env.
470 - timeout argument was not properly used in Coro::Socket.
471 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
472 - emulate undocumented(!) functionality of IO::Socket required
473 by LWP(!!).
474 - updated eg/lwp to work with newer lwp's.
475 - remove "FATAL: uncaught exception" prefix. Coroutines that die
476 kill the whole process, just as exceptions in the main "coroutine"
477 did already.
478
4792.1 Wed Nov 1 23:01:13 CET 2006
480 - fix a long-standing bug in Coro::Select where select with
481 zero timeout would instead change the current default filehandle.
482 - use a simpler and hopefully more robust way to clone padlists
483 (uses less memory and a perl function instead of our own).
484 - coro can now create a stack guard area on many architectures.
485 - Coro::AIO properly reexports additional functions from IO::AIO.
486 - updated libcoro with a workaround for OS X,
487 pach and testing by Michael Schwern.
488
4892.0 Tue Oct 24 05:47:17 CEST 2006
490 - support additional aio requests in Coro::AIO.
491
4921.9 (never properly released due to a glitch)
493
4941.8 Thu Feb 2 00:59:06 CET 2006
495 - applied suggested patch by SAMV to avoid problems during stupid
496 mark & sweep gc run.
497 - applied patch by Scott Walters for 5.9.3 compatibility.
498
4991.7 Tue Dec 27 01:41:58 CET 2005
500 - added Coro::AIO, a thin wrapper around IO::AIO.
3 - improved Makefile.PL explanations. 501 - improved Makefile.PL explanations.
4 502
51.6 Mon Dec 12 21:30:05 CET 2005 5031.6 Mon Dec 12 21:30:05 CET 2005
6 - additionally save PM_curpm might fix as-of-yet 504 - additionally save PM_curpm might fix as-of-yet
7 unseen problems with regex matches being attributed to the 505 unseen problems with regex matches being attributed to the

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines