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

Comparing Coro/Changes (file contents):
Revision 1.284 by root, Sat Oct 27 14:07:31 2007 UTC vs.
Revision 1.691 by root, Thu Jun 24 19:10:06 2021 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. 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang 4TODO: channel->maxsize(newsize)?
5 (which is cool in a lot of important aspects (failures!), 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)). 6TODO: swap_sv, maybe add scope_swap_sv?
7TODO: croak when async_pool tries to run canceled thread?
8TODO: how to debug safe_cancel failure (vebrose mode? show how to make backtrace?)
9TODO: scalar context for Coro::AIO wrappers?
7 10
11TODO: State.xs could short-circuit some macros such as CORO_EXECUTE_SLF
12TODO: use frame->destroy to improve Coro::EV
13
14TODO: concurrent ->recv on same condvar
15sub Coro::AnyEvent::CondVar::_send {
16 if ($_[0]{_ae_coro}) {
17 $_->ready
18 for values %{ delete $_[0]{_ae_coro} }
19 }
20};
21
22sub Coro::AnyEvent::CondVar::_wait {
23 until ($_[0]{_ae_sent}) {
24 $AnyEvent::CondVar::Base::WAITING = 0; # avoid recursive check by AnyEvent
25 local $_[0]{_ae_coro}{$Coro::current+0} = $Coro::current;
26 Coro::schedule;
27 }
28};
29
30
316.57 Wed Jul 29 15:08:44 CEST 2020
32 - override AnyEvent condvar methods "properly", as AnyEvent
33 has support for pluggable condvars since 2011. This also
34 avoids problems with recursive waits by avoiding to mix
35 impl methods with coro code.
36 - allow rouse_wait to be called in a different thread.
37 - slightly clarify rouse_cb/rouse_wait documentation.
38 - slightly improve rouse callback performance and memory usage.
39 - try to report correct calller when throwing exception in
40 on $condvar->recv, not a useless line in Coro::AnyEvent.
41
426.56 Thu Feb 20 19:40:07 CET 2020
43 - Coro::EV::time*once functions used ev_once, which is not cancelable.
44 This could cause problems due to I/O watchers not being stopped up
45 on exceptions/cancellation.
46
476.55 Fri Jul 19 17:28:09 CEST 2019
48 - Coro::RWLock woke up only one reader instead of all after a wrlock
49 (analyzed and patch by Maxim Skorobogatov).
50 - slightly improved documentation of Coro::RWLock and RWLock testcase.
51 - quickly (re)-ported to minix 3.3 before minix crashed again.
52
536.54 Sun Dec 16 11:06:07 CET 2018
54 - reinstate Coro::Select::select prototype - it breaks more than it fixes.
55
566.53 Sun Dec 16 10:33:15 CET 2018
57 - the autogenerated prototypes of Coro::AIO now add ";" at the end
58 to avoid generating named unary operators and messing up parsing
59 (mostly affects aio_mlockall 4 | 8 which was parsed as
60 (aio_mlockall 4) | 8).
61 - do not use a prototype for select in Coro::Select, as CORE::select also
62 doesn't have one.
63
646.52 Tue Aug 14 18:49:43 CEST 2018
65 - libcoro update: allow sharing of coro threads using the pthreads
66 backend among different pthreads, as is required by Coro::Multicore.
67 - always allow ->call and ->eval when the coro is the current coroutine. this
68 makes it possible, among other things, to call these on $Coro::main.
69 - minor code cleanups.
70
716.514 Thu Aug 31 18:25:31 CEST 2017
72 - safe_cancel no longer croaks when called on already-destroyed
73 thread, but instead "just works".
74
756.513 Sat Jul 15 01:18:51 CEST 2017
76 - experimental 5.22 comaptibility changes (PadlistNAMES).
77 - removed attributes example and made sure the other examples
78 still work.
79 - fix some statevar types for newer perls, should be of no consequence.
80
816.512 Fri Jul 14 05:03:17 CEST 2017
82 - seems to pass test suite down till 5.8.9, so make this the
83 minimum requirement in Makefile.PL.
84 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
85 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
86 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
87 fortified longjmp variants in some libcs.
88 - re-try JIT allocation without PROT_EXEC, to hopefully improve
89 portability (still doesn't seem to work around broken selinux
90 mmap everywhere).
91 - speling fixes (Xavier Guimard).
92
936.511 Sun Jun 26 23:44:50 CEST 2016
94 - make it compile with threaded perls again.
95 - simplify and speed up __DIE__ and __WARN__ handling.
96
976.51 Sat Jun 25 20:57:20 CEST 2016
98 - 6.5 release didn't compile on most older perls due to
99 a logic error enabling experimental 5.24 code for all versions.
100 - since various broken or deficient patches float around
101 to work around the vtbl API breakage in perl, let's provide
102 our own workaround, which seems to be much less invasive
103 and more compatible than the approaches seen so far. YMMV.
104 - work around assertion failure in perl_destruct on 5.24
105 (analyzed by Dave Mitchell).
106
1076.5 Wed Jun 22 22:23:50 CEST 2016
108 - swap_sv swapping order was not symmetrical, causing
109 wrong swaps when swap_sv was used multiple times on the
110 same sv in the same thread.
111 - swap_sv calls can now be undone by calling it again
112 with the same variables.
113 - swap_sv calls will now be undone in async_pool threads.
114 - split Coro::Semaphore::up/adjust into separate xs functions
115 for better error reporting, at slight codesize increase.
116 - (libcoro) arm assembly support, please test and report.
117 - adjust to PL_savestack changes in perl 5.24 (adapted from
118 the debian patch, which unfortunately gets it wrong).
119
1206.49 Sat Oct 17 01:40:12 CEST 2015
121 - throwing an exception to a thread waiting in
122 Coro::Handle using Coro::EV did not stop the watchers,
123 causing the next call to fail (testcase by Martin Pritchard).
124 - bump minimum perl version to 5.10.
125
1266.48 Sun Oct 4 19:03:51 CEST 2015
127 - fix memory leak when Coro::Handle uses Coro::EV internally
128 (testcase by Sten Sten).
129 - update code to libev 4 API (internally, Coro still used the
130 version 3 API calls).
131 - remove unused hv_sig.
132
1336.47 Sat Jul 11 03:58:20 CEST 2015
134 - swap_svs - NVs can be in the SV head beginning with 5.20.
135
1366.46 Tue Jun 30 14:36:30 CEST 2015
137 - restore portability to perls with windows fork emulation
138 (patch by Petr Písař).
139
1406.45 Tue Jun 30 01:40:08 CEST 2015
141 - restore portability to perl 5.16 and below.
142
1436.44 Tue Jun 30 00:41:54 CEST 2015
144 - avoid segfaulting (or worse) when tracing a canceled thread.
145 - the Event module silently broke it's public hook API, causing
146 failures on perls compiled with -Duselongdouble. Requiring
147 the latest version of Event because I don't know which release
148 changed this (it's not mentioned in the ChangeLog).
149 - allow xs level enterleave hooks via CoroAPI.
150
1516.43 Thu Jun 4 15:38:14 CEST 2015
152 - use stability canary.
153 - port to stableperl-5.22.0-1.001.
154 - update libecb.
155
1566.42 Wed Feb 11 20:29:52 CET 2015
157 - Coro::SemaphoreSet->try did not actually work (analyzed by
158 SATO Kentaro).
159 - upgrade libecb to be C11 compliant.
160
1616.41 Sat Sep 6 22:08:46 CEST 2014
162 - restore portability to perl 5.8.x.
163 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
164 up some memory, but fixes "use feature" and similar modules
165 when used inside a Coro.
166 - allow zero as argument to Coro::Channel to mean the same thing
167 as no argument. this works with older versions as well,
168 but wasn't legal until now.
169 - slightly better c header file detection.
170
1716.39 Mon Jun 2 00:00:08 CEST 2014
172 - work around more incompatible changes in 5.20.
173
1746.38 Sun Jun 1 21:54:23 CEST 2014
175 - check that perl slots actually have enough space to hold
176 interpreter variables.
177 - untested port to perl 5.19 (64 bit tmps indices) (reported
178 by Andreas König).
179 - croak when cancel is called without a thread context.
180
1816.37 Tue Mar 4 13:27:33 CET 2014
182 - *sigh*, removed leftover debugging code from debugging a
183 perl bug, of all things.
184
1856.36 Tue Mar 4 07:11:59 CET 2014
186 - semaphores would not clear the destroy hook when interrupted
187 by ->throw, causing segfaults or worse.
188 - ->throw on a thread waiting for a semaphore did not acquire
189 the semaphore, but also didn't wake up other waiters,
190 possibly causing a deadlock.
191 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
192 to report a full stacktrace, which should help find out
193 where the actual call is.
194 - "atomically" destroy data in slf_destroy, because it is
195 the right thing to do, just in case.
196 - disable ecb.h libm dependency, to compile on ancient systems
197 or under adverse conditions.
198
1996.33 Mon Nov 18 11:26:27 CET 2013
200 - do not crash when freeing padlists with holes (in 5.18).
201 - tentative SVt_BIND 5.19 port/fix.
202
2036.32 Tue Nov 5 15:35:35 CET 2013
204 - use a new algorithm to derive padlists for perl 5.18. The old
205 one could lead to 0-pointer accesses inside perl (reported
206 by Darin McBride).
207
2086.31 Thu May 9 07:39:48 CEST 2013
209 - Coro::AIO requests would crash if the thread was ready'd
210 while the request was ongoing.
211
2126.29 Wed May 8 02:55:18 CEST 2013
213 - when an on_destroy handler destructs the coro currently being
214 destructed a perl scalar could be accessed after being freed,
215 likely causing a crash.
216
2176.28 Wed Mar 6 06:58:02 CET 2013
218 - clean remnants of existing __DIE__ and __WARN__ handlers so
219 they lose their magic and will not cause segfaults later
220 (testcase by Andrey Sagulin).
221 - improved Coro::State documentation a bit.
222 - Coro::Debug::command now flushes the output.
223 - add hack detection code for x32 abi, because the braindead slugs
224 who designed that made it look exactly like x86_64 without
225 providing proper compile time symbols to test for it. as a result,
226 this detection cannot work reliably.
227 - valgrind stack registering was broken.
228 - do not rely on Time::HiRes anymore in Coro::Debug.
229
2306.23 Fri Dec 7 23:36:37 CET 2012
231 - use experimental fiber implementation on native windows
232 perls.
233 - use sizeof (void *) as multiplication factor for stack sizes,
234 to accomodate the totally braindamaged microsoft 64 bit "os".
235 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
236 activeperl 5.16 crashes when PerlIO_define_layer is called due
237 to some bug in the perl dll, strawberry perl at least passes
238 the testsuite.
239 - implement Coro::Handle->peeraddr/host/port, for slightly
240 improved compatibility with LWP.
241 - implement 5.17 compatibility by almost blindly applying a
242 good-looking patch by Father Chrysostomos.
243 - move stack management functions into libcoro 3.
244 - libcoro version 3 "released".
245 - support magic values as timed_io_once args.
246 - recommend AnyEvent 7+ or EV 4+, also require EV
247 version 4 or newer for Coro::EV.
248
2496.10 Tue Oct 9 01:14:27 CEST 2012
250 - updated ecb.h, it had a typo that caused it to not compile on many
251 big endian systems (reported by many people).
252 - disable memory fences in ecb.h to improve portability.
253
2546.09 Sat Oct 6 23:25:02 CEST 2012
255 - Coro::EV I/O watchers were not interruptible by exceptions
256 (Coro::State::throw) (testcase by sten).
257 - ->throw now puts threads into the ready queue, as this seems to
258 be expected by existing code, and code that doesn't cope with spurious
259 wakeups needs fixing anyway.
260 - use fd -1 in mmap.
261 - cast I32 to int in error message printf.
262 - warn about broken so-called "hardened" kernels.
263
2646.08 Fri Apr 13 12:05:47 CEST 2012
265 - be more aggressive about exiting like perl does - formerly,
266 exiting from the non-main thread would not execute END blocks.
267
2686.07 Fri Nov 11 21:21:48 CET 2011
269 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
270 - update ecb.h.
271
2726.06 Mon Aug 8 23:59:48 CEST 2011
273 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
274 backend, which is an order of magnitude slower. unfortunately, cygwins
275 pthread implementation isn't very complete either, so allocate the stack
276 twice just to be sure.
277 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
278 in its ithreaded perl - best recompile cygwin and use the 'w'indows
279 backend for much better performance. also disable ithreads for
280 even better performance...).
281
2826.05 Thu Aug 4 21:36:36 CEST 2011
283 - blush, condvar values would not be propagated from send to recv anymore
284 (reported by Chip Salzenberg).
285 - use exponential increase for the readline buffer length in
286 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
287
2886.04 Wed Aug 3 17:15:45 CEST 2011
289 - use even more efficient and more compatible condvars for
290 compatibility to AnyEvent 6.x :)
291 - more inconsequential ecb.h updates.
292
2936.03 Wed Aug 3 11:41:30 CEST 2011
294 - change how Coro patches AnyEvent condvars for compatibility to
295 AnyEvent 6.x.
296 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
297 hasn't been included.
298
2996.02 Wed Jul 13 04:35:19 CEST 2011
300 - "improve portability to Gentoo" - gentoo manages to put perl variables
301 in memory areas that are farther than 2gb apart, which the jit couldn't
302 handle and barfed. now it's just a bit slower on gentoo and similar
303 systems.
304
3056.01 Sun Jul 3 12:31:14 CEST 2011
306 - workarounds are good, but the test for whether pthreads are used
307 was not good. this one should be better.
308 - check differently whether gcc generates cfi instructions itself.
309
3106.0 Wed Jun 29 19:43:35 CEST 2011
311 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
312 destroyed and cleaned up automatically (e.g. async { schedule }).
313 - implement a JIT compiler for part of the thread switch code,
314 which gives a 50% speed improvement on threaded perls, and
315 about 4% on non-threaded perls (so threaded perls now finally
316 reach about half the speed of non-threaded perls).
317 - slightly modernise Coro::Intro, add section about rouse functions.
318 - avoid DEFSV and ERRSV, giving another 10% improvement
319 in thread switching.
320 - Coro::State->is_destroyed is now called is_zombie.
321 - implement a Coro->safe_cancel method that might fail, but
322 cancels in a "safer" way if it succeeds.
323 - add preliminary support for DEBUGGING perls.
324 - get rid of two hash-accesses when initialising a new Coro - this
325 speeds up coro creation by almost a factor of two.
326 - croak when a coro that is being cancelled tries to block
327 (e.g. while executing a guard block), instead of crashing or
328 deadlocking.
329 - use a more robust and also faster method to identify Coro::State
330 objects - speeds up everything a bit.
331 - implement Coro->cancel in XS for a 20% speed improvement, and to
332 be able to implement mutual cancellation.
333 - speed up context switches by a percent or two by more efficiently
334 allocating context stack entries.
335 - implement Coro->join and Coro->on_destroy in XS for a speedup and
336 a reduction in memory use.
337 - cancelling a coro while it itself is cancelling another coro is
338 now supported and working, instead of triggering an assertion.
339 - be a bit more crash-resistant when calling (buggy) on_destroy
340 callbacks (best effort).
341 - move on_destroy into the slf_frame, to allow extension slf
342 functions to have destructors.
343 - get rid if coro refcounting - simply crash in other interpreter
344 threads by nulling the pointers on clone.
345 - simplify warn/die hook handling when loading Coro - the convoluted
346 logic seems to be no longer neccessary.
347 - use libecb instead of our own home-grown gcc hacks.
348 - document alternatives to Coro::LWP. Please use them :)
349 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
350 sigaltstack. Really. wine suffers from it, erlang suffers from it,
351 and it's known since at least 2006.
352
3535.372 Wed Feb 23 06:14:30 CET 2011
354 - apparently mingw doesn't provide a working gettimeofday, try to
355 work around that by relying on Time::HiRes (indirectly brought to
356 my attention by Max Maischein).
357 - fix some portability issues when Time::HiRes was used.
358
3595.371 Mon Feb 21 14:36:08 CET 2011
360 - backport to windows process emulation code again.
361
3625.37 Sat Feb 19 07:49:44 CET 2011
363 - add a big "Coro thread life cycle" section to "man Coro".
364 - try a tentative workaround against the breakage that 5.13 has
365 introduced without depreciation period. sigh.
366 - no longer use Time::HiRes if gettimeofday is available, which
367 saves quite a lot of memory.
368
3695.36 Sun Feb 13 05:33:41 CET 2011
370 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
371 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
372 method call.
373 - undocument Coro::Timer::sleep and obsolete whole module.
374 - optimise Coro::Timer::timeout memory and cpu usage.
375 - slightly updated Coro::Intro for recent changes.
376 - do not initialise PL_dirty anymore.
377
3785.25 Thu Nov 11 01:08:39 CET 2010
379 - try a different approach on netbsd - netbsd 5 finally has marginally
380 working pthreads, but still broken ucontext/sigaltstack.
381 - openbsd 4.8 finally got their act together, Coro works out of the box
382 with asm, setjmp and pthreads (no change, just informational).
383
3845.24 Sat Oct 23 11:27:12 CEST 2010
385 - port to the EV 4.0 API.
386 - work around bugs in mingw32, making strawberry perl work
387 out of the box.
388 - correctly modify Coro::AIO function prototypes
389 so that they reflect the "no optional parameters" rule.
390 - "ported" libcoro to C++.
391
3925.23 Mon May 17 18:50:42 CEST 2010
393 - be more resistant to ordering changes when initialising
394 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
395 Waldorf).
396 - document that perl 5.12 deliberately removed support for cloning.
397
3985.22 Wed Apr 14 03:55:35 CEST 2010
399 - correctly return udnef on errors in Coro::Handle::read/write
400 (testcase by Marc Mims).
401 - convert Coro::Util into a "perl compatibility wrapper" - the functions
402 are less useful now, but are drop-in replacements for existing
403 functions, listing better alternatives in the documentation. This also
404 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
405 with Coro::Util::inet_aton.
406 - do not override $Coro::idle unconditionally in Coro.pm, as other
407 modules could have provided their own idle coro already
408 (for exmaple, Coro::AnyEvent).
409 - fix Coro::Util::gethost* functions.
410 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
411
4125.21 Wed Dec 16 07:19:51 CET 2009
413 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
414 together.
415 - add some examples on how to combine other event loops with Coro in
416 Coro::AnyEvent, and how to run it (and not to block). Seems to be
417 the most common source of confusion.
418 - try to catch people naively blocking in an event callback.
419 - work around the perl filehandle bug issue in conjunction with
420 older common::sense (as indirectly pointed out by ZSystem).
421 - clarify the "not from signal handlers" section.
422
4235.2 Sun Oct 4 14:54:24 CEST 2009
424 - Coro::Storable destroyed the prototypes of the functions it wrapped.
425 - export rouse_cb and rouse_wait by default now.
426 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
427 - new method $state->swap_sv.
428 - added section on "windows process emulation" to the manpage,
429 after a not-so-fruitful (nor-friendly) "discussion" with chip
430 salzenberg (discussion implies arguments, but his only arguments
431 were ad-hominems, one wonders why he started it in the first
432 place). I hope this explains it well enough for him to understand,
433 and maybe well enough for others to understand.
434 - use common::sense everywhere now.
435 - idle callbacks are no longer supported, use idle coros instead.
436 - print a thread listing when a deadlock is detected.
437
4385.17 Sat Aug 22 23:09:31 CEST 2009
439 - work around a bug in the perl debugger causing crashes
440 when running under the debugger by marking _pool_handler as nodebug.
441 - speed up Coro::async considerably.
442 - try some hacks to get netbsd to work "more often" - their broken
443 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
444 - convert Coro to AE 5.0 API.
445
4465.162 Tue Jul 28 04:04:03 CEST 2009
447 - perl 5.8.2 is now minimum requirement.
448 - skip t/19_handle.t on broken windows perls.
449
4505.161 Wed Jul 22 04:47:38 CEST 2009
451 - Coro::AnyEvent::poll could have a different prototype when EV was
452 used as backend (analyzed by Tatsuhiko Miyagawa).
453 - Coro::AnyEvent errornously initialised the event loop when loaded,
454 not on demand.
455 - try to workaround rare */t/01_unblock.t failures.
456
4575.16 Tue Jul 21 01:44:37 CEST 2009
458 - Coro::AnyEvent failed to hook into the event loop
459 when no threads had been readied between detecting
460 the event loop and actually running it.
461 - considerably speed up Coro::Select by taking avdantage
462 of AnyEvent > 4.8 and some other optimisations.
463 - implement paragraph readline mode in Coro::Handle
464 (based on patches by Zsbán Ambrus).
465 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
466 and analyzed by Yasuhiro MATSUMOTO).
467 - clarified libcoro license and copyright.
468 - someone stole my EXTRA_META!!!
469 - implement Coro::Select::patch_pp_sselect and it's brother,
470 for hardcode select overriding.
471
4725.151 Mon Jul 6 05:41:57 CEST 2009
473 - backport to windows process emulation code again (patch by
474 Yasuhiro MATSUMOTO).
475 - slightly update Coro::MakeMaker.
476
4775.15 Tue Jun 30 10:28:06 CEST 2009
478 - deprecate Coro::Socket, document how to get ipv6 support via
479 AnyEvent::Socket instead.
480 - implement signal->wait ($cb) interface, similar to semaphores.
481 - work around SvOK not supporting getmagic, so we have to getmagic
482 to test for undef :/ (reported by Matthias Waldorf).
483 - load Coro::AnyEvent in all modules using AnyEvent.
484 - work around perl corrupting our internal data structures,
485 reported by Tokuhiro Matsuno.
486 - enable per-coroutine real and cpu time gathering
487 (Coro::State::enable_times).
488
4895.14 Wed Jun 24 01:37:48 CEST 2009
490 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
491 as this seems to have been a difficult concept (poll, sleep, idle,
492 idle_upto).
493 - add Coro::AnyEvent::readable/writable functions.
494 - clarify Coro::EV/Event/AnyEvent manpages.
495 - free per-thread global scalars in the thread calling ->cancel, to
496 avoid crashes when $_, $@ etc., are magical but some of those
497 had already been freed.
498 - "unexperimentalise" the callback interface for Coro::Semaphore.
499 - speed up ready queue management/context switching by using a linked
500 list instead of an array (~5-10%).
501 - implement "watch" command in Coro::Debug shells.
502 - for fun, implement time-slicing as an example in the manpage.
503 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
504 Coro::Event, use the normal AnyEvent handling instead of dieing
505 (the same is true for Coro::Handle).
506 - properly document Coro::EV::timed_io_once.
507 - avoid unneccessary ->cancel calls in Coro::Handle.
508 - maybe make it work on mingw32 with win32 backend
509 (based on patch by Yasuhiro Matsumoto).
510
5115.132 Fri May 29 09:00:39 CEST 2009
512 - do not keep a reference to the argument itself in
513 Coro::Semaphore::guard, as it could change later.
514 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
515 should support a prepare callback.
516
5175.131 Mon Mar 16 23:20:37 CET 2009
518 - implement and document Coro->suspend, Coro->resume.
519 - fix Coro::Select implementation to not (often) close
520 the passed file descriptors (testcase provided by pippijn).
521
5225.13 Mon Dec 15 21:51:42 CET 2008
523 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
524 - don't set diehook to C<undef> but instead to NULL, to avoid
525 spurious warnings.
526 - fix a lot of bugs in Coro::SemaphoreSet.
527 - Coro::SemaphoreSet will less often create a semaphore needlessly.
528 - add Coro::SemaphoreSet::count and wait methods.
529 - take advantage of the new Guard module.
530 - deprecate Coro::guard.
531 - try to fix the dreaded 01_unblock tests once more. I hate it when
532 testsuites need more fixing than the code they are supposed to test.
533 - croak in more cases when a required callback isn't resolvable.
534 - fix some minor issues in Coro::State->call/eval.
535 - use current coroutine context instead of a temporary one
536 when temporarily switching to another coroutine.
537 - do not call C-level on_destroy handlers during global destruction,
538 to avoid needless segfaults.
539
5405.12 Sun Dec 7 13:30:38 CET 2008
541 - add default config for MirOS, which seems to be a bug-to-bug
542 compatible fork of openbsd ("world domination by releasing
543 openbsd cvs before the openbsd folks do it" or so :).
544 - free_padlist did destroy the names pad, not good, but didn't
545 seem to bother perl - this could fix issues such as eval ""
546 inside a function called from multiple coroutines.
547 - use a different method to detect destruction time.
548 - be more careful when freeing padlists just before global
549 destruction.
550 - fixed and expanded the call/cc example.
551 - renamed _terminate to _coro_run.
552 - new method Coro::Channel->shutdown.
553 - try pthreads on openbsd <4.4 (broken sigaltstack, will
554 pthreads fare better?).
555 - be less picky about destroying "a" running coroutine.
556
5575.11 Tue Nov 25 21:49:05 CET 2008
558 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
559 working code with failed assertions, introducing more bugs than
560 it fixes, requiring elaborate workarounds :(
561
5625.1 Mon Nov 24 08:54:59 CET 2008
563 - wrote a small introductory tutorial - Coro::Intro.
564 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
565 - Coro::Select did errornously dup the file descriptors
566 and didn't work with all AnyEvent backends.
567 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
568 LWP data transfers.
569 - disassociate c contexts from coro objects - this is agruably more
570 correct, but mostly allows sharing of cctxs between coro and state
571 objects, for added memory savings and speed increases.
572 - bumped $Coro::POOL_RSS up to 32kb by default.
573 - no longer set the optype to OP_CUSTOM, as B::* understandably
574 doesn't like this very much (and we *are* a type of entersub).
575 - implement state cloning, just to prove that call/cc can be done.
576 - automatically load Coro::AnyEvent in Coro::Handle.
577 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
578 do not have this method (and don't need it either).
579 - speed up generic anyevent methods in Coro::Handle by using rouse
580 callbacks.
581 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
582 others. It also makes the debugger happier, as you can trace
583 through the idle threads now.
584 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
585 - eg/event was pretty broken.
586 - better 5.8.6 compatibility.
587
5885.0 Thu Nov 20 10:35:05 CET 2008
589 - NEW ARCHITECTURE: use the latest 4.x version if you experience
590 stability issues.
591 - bump API version to 7 - all dependents must be recompiled.
592 - removed timed_* functions - they were not being used anyways
593 and should be replaced by a more generic mechanism -
594 and were annoying to support anyways :)
595 - removed SemaphoreSet's waiter method - use sem method instead.
596 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
597 - async_pool did free a scalar value twice
598 ("Attempt to unreference...").
599 - fix a longstanding bug where calling terminate on a coro that
600 was waiting for a semaphore that was just becoming available
601 would cause a deadlock (semaphore would get into a state where
602 it was available but waiters were still blocked).
603 - calling throw on a coroutine that is waiting for a semaphore will
604 no longer make it acquire the semaphore (and thus leak a count).
605 - perl's process emulation is now not even theoretically supported
606 anymore.
607 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
608 conversion of callback-style to blocking-style.
609 - new methods $coro->schedule_to and ->cede_to, to specifically
610 schedule or cede to a specific coroutine.
611 - new function Coro::Semaphore::wait.
612 - use named constants in Coro::Channel (Richard Hundt).
613 - directly patch the entersub opcode calling SLF functions (cede,
614 transfer and so on). this does speed up context switching, but
615 more importanly, it frees us from the hardcoded behaviour of
616 entersub, so we might actually be able to return something from
617 those functions and atcually create new ones.
618 - take advantage of __builtin_frame_address on gcc.
619 - expose THX in coroapi (not sure whether this was a wise decision,
620 as "threaded" perls are running at half speed anyways).
621 - implement execute_slf (schedule-like-function) interface that makes
622 it possible to implement schedule-like-functions in XS.
623 - use new SLF interface to massively speed up Coro::EV by roughly a
624 factor of two.
625 - used new SLF interface to massively speed up Coro::Semaphore by a
626 factor of three.
627 - used new SLF interface to speed up Coro::AIO by roughly a factor of
628 four and reduce its memory usage considerably.
629 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
630 for a nice speedup and vastly more correct behaviour. Also implement
631 a new method "sem" to get at the underlying semaphore object.
632 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
633 (in comparison) 20-40% speedup.
634 - used new SLF interface to reimplement Coro::Signal gaining
635 some unknown (because I was too lazy), but certain, speedup, and also
636 making signals reliable for the first time.
637 - used new SLF interface and other optimisations to speed up async_pool
638 by a factor of two. It also doesn't rely on perl's exception mechanism
639 to exit anymore. The overhead for terminating an async_pool, coro over
640 a normal async is now very small.
641 - sped up coroutine creation/destruction by 40%.
642 - forgot to include Coro/libcoro/README in the dist for all these years.
643 - work around a freebsd pthreads bug (manual testcancel is required as
644 pthread_cond_wait isn't a cancellation point on freebsd).
645 - use new rouse functions to speed up and simplify Coro::BDB.
646 - make "prefer perl native functions" work with threaded perls.
647 - condense Coro::Debug ps output, hint at v and w flags.
648 - (libcoro) lots of minor cleanups and portability improvements.
649
6504.914 Wed Nov 19 12:54:18 CET 2008
651 - fix a disastrous bug in the readline optimisation
652 introduced in 4.801.
653
6544.913 Sat Nov 15 07:58:28 CET 2008
655 - async_pool did free a scalar value twice
656 ("Attempt to unreference...").
657
6584.912 Thu Nov 13 18:31:23 CET 2008
659 - minor cleanups.
660 - use much larger stacks on linux and perl < 5.8.8.
661 - Coro::Debug::new_unix_server did not unlink the socket
662 when destroyed.
663
6644.911 Tue Nov 11 04:26:17 CET 2008
665 - "port" to threaded perls.
666
6674.91 Mon Nov 10 05:36:38 CET 2008
668 - the ->throw exception object no longer leaks.
669 - creating a new cctx leaked a scopestack entry (memleak).
670 - new coroutines didn't get created with a zero flags field
671 (unknown impact).
672 - calling ->throw on a not-yet-started coroutine should now work
673 instead of being ignored.
674 - ->throw is now supported on Coro::State objects.
675 - clean up cctx creation code a bit.
676 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
677
6784.9 Sat Nov 8 17:45:27 CET 2008
679 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
680 - (libcoro) no longer leak threads in the experimental pthread backend,
681 also speed it up considerably.
682 - (libcoro) do not rely on makecontext passing void *'s unscathed.
683 - fix compiletime dependencies on libcoro in the Makefile.
684 - cctx_count wasn't always updated properly.
685 - Coro::State::cctx_stacksize wasn't applied correctly.
686 - new function Coro::State::cctx_max_idle.
687 - the default max number of idle C contexts is now 4.
688 - (libcoro) try harder to get _setjmp/_longjmp.
689 - (libcoro) cleanup and extend the libcoro API to officially
690 allow the creation of empty source contexts.
691 - very experimental workaround for the totally broken netbsd platform.
692 - tried to hack around openbsd bugs.
693
6944.804 Wed Nov 5 16:36:00 CET 2008
695 - Coro::Debug::new_unix_server would not create a non-blocking listening
696 socket, sometimes causing freezes.
697 - (libcoro) fix misaligned stack points for setjmp and assembly
698 methods, which can cause crashes on x86/x86_64 with a sufficiently
699 aggressive compiler.
700 - new function: Coro::Debug::new_tcp_server.
701 - move ->throw into the Coro class because it only works on coro objects.
702
7034.803 Mon Nov 3 17:16:12 CET 2008
704 - (libcoro) use a global asm statement to become independent of gcc
705 otpimisations for CORO_ASM (thanks to pippijn for the idea).
706 - try to workaround yet another broken bsd, this time dragonfly.
707
7084.802 Thu Oct 30 10:56:12 CET 2008
709 - support -fno-omit-frame-pointer on x86 with the assembly method.
710 - tune 01_unblock.t tests a bit.
711
7124.801 Wed Oct 22 18:33:37 CEST 2008
713 - improve readline speed for very long "lines".
714 - backport to 5.8.8.
715
7164.8 Thu Oct 2 13:34:40 CEST 2008
717 - new function Coro::AIO::aio_wait.
718 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
719 - greatly speed up and reduce memory usage of Coro::AIO requests.
720 - implement some other µ-optimisations.
721
7224.749 Mon Sep 29 14:40:12 CEST 2008
723 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
724
7254.748 Sat Sep 27 14:03:19 CEST 2008
726 - implement, but do not document, PerlIO::cede(granularity).
727 - Coro::Storable forgot to wrap Storable::pstore.
728 - work around the multitude of leaks and memory corruption
729 bugs in PerlIO::via by using our own C-level perliol. As a side
730 effect, Coro::Storable is now much, much, much faster.
731
7324.747 Tue Sep 23 01:59:41 CEST 2008
733 - fix a per-cv memleak (one empty array was leaked per
734 code reference).
735 - avoid a crash in coro->call|rss when the coroutine was already
736 destroyed (most noticably when using Coro::Debug::ps :)
737 - also protect *Storable::FILE.
738 - push up default storable granularity to 20ms.
739
7404.746 Sun Sep 21 03:22:20 CEST 2008
741 - be more insistent on locking Storable against reentrancy
742 in Coro::Storable.
743 - move swap_def?v and throw to Coro::State, as documented.
744
7454.745 Thu Jul 24 00:14:38 CEST 2008
746 - remove debugging code related to MgPV_nolen_const, also try to
747 make it compile with perl 5.8.6 (yes, apple apparently loves
748 outdated software). Reported by John S.
749
7504.744 Tue Jul 8 22:06:35 CEST 2008
751 - correctly provide default DIE/WARN handlers as documented.
752 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
753 implementations inside perl work fine for us.
754
7554.743 Mon Jun 16 00:21:57 CEST 2008
756 - when using Coro::EV without running EV::loop it could
757 result in busy-waiting for events, this has been fixed.
758 - reduce codesize and improve performance by using EV_DEFAULT_UC.
759
7604.742 Sat May 31 14:10:21 CEST 2008
761 - implement a workaround for (some) perl <5.8.8 versions.
762 - require EV 3.3+.
763
7644.741 Fri May 30 23:33:09 CEST 2008
765 - tell netbsd how utterly broken their imitation of an OS is
766 and refuse to build by default if pthreads are in use.
767 - switch to "s" method on all bsd's by default, as their ucontext
768 stuff seems just too broken.
769 - fix a bug in Coro::Select.
770
7714.74 Thu May 29 20:05:31 CEST 2008
772 - do not test Coro::LWP for lack of dependencies.
773 - do not test Coro::Select for lack of working perls.
774
7754.73 Thu May 29 2008
776 - fix a bug in Coro::EV which would cause it to block despite
777 there being runnable coroutines.
778 - sprinkle "no warnings" freely over everything, also suppress
779 warnings for some other modules.
780 - fix typo in WSAEWOULDBLOCK.
781
7824.72 Sun May 25 05:14:36 CEST 2008
783 - tweak META.yaml a bit, unfortunately, there is no documented way
784 to have optional dependencies with CPAN. doh :(
785 - avoid running some tests on windows because they would fail due to
786 perl bug (broken fork, broken pipes...).
787 - work around perl on windows bugs where perl returns undocumented
788 error codes for sysread, syswrite etc. by taking advantage
789 of AnyEvent's workaround for that problem.
790 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
791 a common perl implementation bug on windows.
792 - use unix domain sockets in testsuite to work around
793 common perl implementation bugs on widows (they are emulated by
794 tcp sockets on windows. ugh.)
795
7964.71 Sat May 24 20:01:27 CEST 2008
797 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
798 - take advantage of async name resolution of AnyEvent::Util.
799 - work around brutal inet_aton override in Coro::LWP.
800 - take advantage of the readyhook in Coro::EV, for smoother
801 scheduling.
802
8034.7 Sun May 11 00:32:19 CEST 2008
804 - completely reworked the Coro manpage.
805 - added Coro::AnyEvent, generic event loop integration.
806 - implement cancel, ready and kill commands in Coro::Debug.
807 - document find_coro in Coro::Debug.
808 - incompatible change: rename has_stack to has_cctx.
809 - Coro::AIO and Coro::BDB no longer force event model detection,
810 use AnyEvent::AIO and AnyEvent::BDB.
811
8124.6 Sat Apr 26 10:05:14 CEST 2008
813 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
814 by the Coro module. It was a mistake to have it in the first place.
815 - (experimental) support for activestate perl 5.10 (method "w").
816 - (experimental) support for strawberry perl (method "a").
817 - coro_sigelem_set did not return a value although it had to,
818 actual impact unknown.
819
8204.51 Mon Apr 14 13:28:27 CEST 2008
821 - make it compile again on 5.8.
822
8234.50 Thu Apr 10 09:43:17 CEST 2008
824 - I did it twice! (see 4.49).
825
8264.49 Sun Apr 6 21:23:31 CEST 2008
827 - grr, instead of compiling the recent changes on 5.10 only they
828 were compiled on 5.8 only.
829
8304.48 Sun Apr 6 20:36:46 CEST 2008
831 - allow coroutine switches during eval's under 5.10.x, as it seems
832 the parser is a per-interpreter option now, so this is safe
833 (this might fix the odd crashing bug).
834 - drop support for 5.9.x versions: they are dead, jim.
835
8364.47 Sun Apr 6 00:37:52 CEST 2008
837 - force cctx allocation on API calls: we know we need to force one
838 and gcc actually manages to confuse our heuristic nowadays,
839 leading to crashes and worse.
840 - document force_cctx.
841
8424.46 Fri Apr 4 22:05:43 CEST 2008
843 - upgrade libcoro, resulting in pthread-backend (which was only created
844 to fulfill the rules of the programming languages shootout).
845
8464.45 Thu Mar 13 11:55:36 CET 2008
847 - fix a file leaking bug in eg/mhyttpd that would allow
848 downloading of any file (reported by oesi).
849 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
850 (also add testcase).
851 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
852 apparently having loads of fun with that).
853 - implement $state->swap_defsv, ->swap_defav and document ->throw.
854
8554.4 Wed Feb 13 16:44:29 CET 2008
856 - only assume .cfi pseudo isns exist on GNU/Linux.
857 - add get_timed method to Coro::Channels.
858 - fixed Coro::Debug manpage.
859 - perl 5.11 compatbility improvement based on patch by
860 Andreas König.
861
8624.37 Sun Jan 20 11:25:23 CET 2008
863 - soften the check for an active parser for perl 5.10 (avoiding
864 "Coro::State::transfer called while parsing" in many cases).
865
8664.36 Sun Jan 13 10:53:56 CET 2008
867 - reset diehook when terminating from an async_pool as to not
868 trigger any __DIE__ handlers.
869
8704.35 Sun Jan 13 04:14:13 CET 2008
871 - "bt" debug command now displays any exceptions
872 from longmess and also skips the innermost
873 stackframes, giving more useufl output.
874 - allow backtraces in coroutines blocked in Coro::EV,
875 at a <1% speed hit.
876 - handle localising of $SIG{__WARN__} and __DIE__
877 properly (with a proper amount of dirty hacking).
878
8794.34 Sat Dec 22 17:49:53 CET 2007
880 - upgrade to EV version 2.0 API.
881
8824.33 Mon Dec 17 08:36:12 CET 2007
883 - make Coro::AIO etc. loadable in the absence of EV.
884
8854.32 Mon Dec 17 07:46:02 CET 2007
886 - majorly improved Coro::Handle's performance with Coro::EV.
887 - implemented the until now mythical Coro::BDB module.
888 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
889
8904.31 Wed Dec 5 12:32:39 CET 2007
891 - remove warn statement form Coro::Util.
892
8934.3 Tue Dec 4 20:33:14 CET 2007
894 - calls to the idle function could cause stack corruption
895 when the stack changed.
896 - do no longer rely on the presence of EV::DNS if EV is used
897 (because it is gone), but instead take avdantage of EV::ADNS
898 if available.
899 - add ($) prototypes to all functions in Coro::Storable.
900 - use a conventional (safer) idle callback in Coro::EV.
901 - do accept quickly in Coro::Debug to avoid endless loops.
902
9034.22 Fri Nov 30 16:04:04 CET 2007
904 - really use optimised versions for Event and EV in Coro::Util
905 and Coro::Handle.
906
9074.21 Sun Nov 25 10:48:59 CET 2007
908 - fix a spurious memory read.
909 - Coro::EV no longer keeps the eventloop "alive".
910
9114.2 Fri Nov 9 20:47:05 CET 2007
8 - enable/disable tracing from a new coroutine, not a pooled one. 912 - enable/disable tracing from a new coroutine, not a pooled one.
9 - fix a memleak in Coro::Event. 913 - fix a memleak in Coro::Event.
10 - removed killall call from fork_eval. 914 - removed killall call from fork_eval.
11 - made sure store_fd is already loaded so that fork_eval 915 - made sure store_fd is already loaded so that fork_eval
12 does not have to parse autoload in each subprocess. 916 does not have to parse autoload in each subprocess.
917 - only use assembly method if -O switch is in $Config{optimize}.
918 - add (optional) Coro::EV module, so far the best event loop module
919 directly supported by Coro.
920 - if the event model is EV, use EV::DNS to resolve
921 stuff in Coro::Util.
922 - don't get confused by multiple event notifications in Coro::Handle.
923 - initial support for EV (libevent interface).
924 - require Event and EV using configure_requires, to force their existance.
13 925
144.13 Wed Oct 24 07:26:45 CEST 2007 9264.13 Wed Oct 24 07:26:45 CEST 2007
15 - add Coro::Storable::blocking_thaw. 927 - add Coro::Storable::blocking_thaw.
16 - use a vastly more complicated technique to localise 928 - use a vastly more complicated technique to localise
17 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8. 929 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
… …
22 934
234.11 Thu Oct 11 02:40:24 CEST 2007 9354.11 Thu Oct 11 02:40:24 CEST 2007
24 - port to threaded perls. 936 - port to threaded perls.
25 937
264.1 Thu Oct 11 02:38:16 CEST 2007 9384.1 Thu Oct 11 02:38:16 CEST 2007
27 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 939 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
28 be local to each coro (see Coro::State). 940 be local to each coro (see Coro::State).
29 - incompatible change: for very deep reasons, cede and cede_notself 941 - incompatible change: for very deep reasons, cede and cede_notself
30 cannot return anything, so nothing will be returned. 942 cannot return anything, so nothing will be returned.
31 - possibly bring back 5.10 compatibility (untested). 943 - possibly bring back 5.10 compatibility (untested).
32 - work around stupid (and wrong) warning on 5.10 :(. 944 - work around stupid (and wrong) warning on 5.10 :(.
… …
104 but stateful semantics. 1016 but stateful semantics.
105 - fixed a lot of typos in Coro.pm (patch submitted by David 1017 - fixed a lot of typos in Coro.pm (patch submitted by David
106 Steinbrunner, which applied flawlessly). 1018 Steinbrunner, which applied flawlessly).
107 1019
1083.6 Sat Apr 14 17:13:31 CEST 2007 10203.6 Sat Apr 14 17:13:31 CEST 2007
109 - added some bugfixes to get eg/myhttpd working again. 1021 - added some bugfixes to get eg/myhttpd working again.
110 - added Coro::Storable for often-cede'ing freeze/thaw. 1022 - added Coro::Storable for often-cede'ing freeze/thaw.
111 - try to do a clean exit when a coroutine calls exit 1023 - try to do a clean exit when a coroutine calls exit
112 (EXPERIMENTAL). 1024 (EXPERIMENTAL).
113 - got rid of indirect call through _coro_init. 1025 - got rid of indirect call through _coro_init.
114 - updated the partly antique examples in eg/ to 1026 - updated the partly antique examples in eg/ to

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines