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

Comparing Coro/Changes (file contents):
Revision 1.246 by root, Sun Sep 30 13:57:34 2007 UTC vs.
Revision 1.678 by root, Sun Dec 16 09:33:44 2018 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: destroy gets only called on cancel: fix(?) 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. 4TODO: channel->maxsize(newsize)?
5TODO: maybe implement a default message channel, very much like Erlang 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 (which is cool in a lot of important aspects (failures!), 6TODO: swap_sv, maybe add scope_swap_sv?
7 but very lacking in others). 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?
8 10
93.8 116.53 Sun Dec 16 10:33:15 CET 2018
12 - the autogenerated prototypes of Coro::AIO now add ";" at the end
13 to avoid generating named unary operators and messing up parsing
14 (mostly affects aio_mlockall 4 | 8 which was parsed as
15 (aio_mlockall 4) | 8).
16 - do not use a prototype for select in Coro::Select, as CORE::select also
17 doesn't have one.
18
196.52 Tue Aug 14 18:49:43 CEST 2018
20 - libcoro update: allow sharing of coro threads using the pthreads
21 backend among different pthreads, as is required by Coro::Multicore.
22 - always allow ->call and ->eval when the coro is the current coroutine. this
23 makes it possible, among other things, to call these on $Coro::main.
24 - minor code cleanups.
25
266.514 Thu Aug 31 18:25:31 CEST 2017
27 - safe_cancel no longer croaks when called on already-destroyed
28 thread, but instead "just works".
29
306.513 Sat Jul 15 01:18:51 CEST 2017
31 - experimental 5.22 comaptibility changes (PadlistNAMES).
32 - removed attributes example and made sure the other examples
33 still work.
34 - fix some statevar types for newer perls, should be of no consequence.
35
366.512 Fri Jul 14 05:03:17 CEST 2017
37 - seems to pass test suite down till 5.8.9, so make this the
38 minimum requirement in Makefile.PL.
39 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
40 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
41 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
42 fortified longjmp variants in some libcs.
43 - re-try JIT allocation without PROT_EXEC, to hopefully improve
44 portability (still doesn't seem to work around broken selinux
45 mmap everywhere).
46 - speling fixes (Xavier Guimard).
47
486.511 Sun Jun 26 23:44:50 CEST 2016
49 - make it compile with threaded perls again.
50 - simplify and speed up __DIE__ and __WARN__ handling.
51
526.51 Sat Jun 25 20:57:20 CEST 2016
53 - 6.5 release didn't compile on most older perls due to
54 a logic error enabling experimental 5.24 code for all versions.
55 - since various broken or deficient patches float around
56 to work around the vtbl API breakage in perl, let's provide
57 our own workaround, which seems to be much less invasive
58 and more compatible than the approaches seen so far. YMMV.
59 - work around assertion failure in perl_destruct on 5.24
60 (analyzed by Dave Mitchell).
61
626.5 Wed Jun 22 22:23:50 CEST 2016
63 - swap_sv swapping order was not symmetrical, causing
64 wrong swaps when swap_sv was used multiple times on the
65 same sv in the same thread.
66 - swap_sv calls can now be undone by calling it again
67 with the same variables.
68 - swap_sv calls will now be undone in async_pool threads.
69 - split Coro::Semaphore::up/adjust into separate xs functions
70 for better error reporting, at slight codesize increase.
71 - (libcoro) arm assembly support, please test and report.
72 - adjust to PL_savestack changes in perl 5.24 (adapted from
73 the debian patch, which unfortunately gets it wrong).
74
756.49 Sat Oct 17 01:40:12 CEST 2015
76 - throwing an exception to a thread waiting in
77 Coro::Handle using Coro::EV did not stop the watchers,
78 causing the next call to fail (testcase by Martin Pritchard).
79 - bump minimum perl version to 5.10.
80
816.48 Sun Oct 4 19:03:51 CEST 2015
82 - fix memory leak when Coro::Handle uses Coro::EV internally
83 (testcase by Sten Sten).
84 - update code to libev 4 API (internally, Coro still used the
85 version 3 API calls).
86 - remove unused hv_sig.
87
886.47 Sat Jul 11 03:58:20 CEST 2015
89 - swap_svs - NVs can be in the SV head beginning with 5.20.
90
916.46 Tue Jun 30 14:36:30 CEST 2015
92 - restore portability to perls with windows fork emulation
93 (patch by Petr Písař).
94
956.45 Tue Jun 30 01:40:08 CEST 2015
96 - restore portability to perl 5.16 and below.
97
986.44 Tue Jun 30 00:41:54 CEST 2015
99 - avoid segfaulting (or worse) when tracing a canceled thread.
100 - the Event module silently broke it's public hook API, causing
101 failures on perls compiled with -Duselongdouble. Requiring
102 the latest version of Event because I don't know which release
103 changed this (it's not mentioned in the ChangeLog).
104 - allow xs level enterleave hooks via CoroAPI.
105
1066.43 Thu Jun 4 15:38:14 CEST 2015
107 - use stability canary.
108 - port to stableperl-5.22.0-1.001.
109 - update libecb.
110
1116.42 Wed Feb 11 20:29:52 CET 2015
112 - Coro::SemaphoreSet->try did not actually work (analyzed by
113 SATO Kentaro).
114 - upgrade libecb to be C11 compliant.
115
1166.41 Sat Sep 6 22:08:46 CEST 2014
117 - restore portability to perl 5.8.x.
118 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
119 up some memory, but fixes "use feature" and similar modules
120 when used inside a Coro.
121 - allow zero as argument to Coro::Channel to mean the same thing
122 as no argument. this works with older versions as well,
123 but wasn't legal until now.
124 - slightly better c header file detection.
125
1266.39 Mon Jun 2 00:00:08 CEST 2014
127 - work around more incompatible changes in 5.20.
128
1296.38 Sun Jun 1 21:54:23 CEST 2014
130 - check that perl slots actually have enough space to hold
131 interpreter variables.
132 - untested port to perl 5.19 (64 bit tmps indices) (reported
133 by Andreas König).
134 - croak when cancel is called without a thread context.
135
1366.37 Tue Mar 4 13:27:33 CET 2014
137 - *sigh*, removed leftover debugging code from debugging a
138 perl bug, of all things.
139
1406.36 Tue Mar 4 07:11:59 CET 2014
141 - semaphores would not clear the destroy hook when interrupted
142 by ->throw, causing segfaults or worse.
143 - ->throw on a thread waiting for a semaphore did not acquire
144 the semaphore, but also didn't wake up other waiters,
145 possibly causing a deadlock.
146 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
147 to report a full stacktrace, which should help find out
148 where the actual call is.
149 - "atomically" destroy data in slf_destroy, because it is
150 the right thing to do, just in case.
151 - disable ecb.h libm dependency, to compile on ancient systems
152 or under adverse conditions.
153
1546.33 Mon Nov 18 11:26:27 CET 2013
155 - do not crash when freeing padlists with holes (in 5.18).
156 - tentative SVt_BIND 5.19 port/fix.
157
1586.32 Tue Nov 5 15:35:35 CET 2013
159 - use a new algorithm to derive padlists for perl 5.18. The old
160 one could lead to 0-pointer accesses inside perl (reported
161 by Darin McBride).
162
1636.31 Thu May 9 07:39:48 CEST 2013
164 - Coro::AIO requests would crash if the thread was ready'd
165 while the request was ongoing.
166
1676.29 Wed May 8 02:55:18 CEST 2013
168 - when an on_destroy handler destructs the coro currently being
169 destructed a perl scalar could be accessed after being freed,
170 likely causing a crash.
171
1726.28 Wed Mar 6 06:58:02 CET 2013
173 - clean remnants of existing __DIE__ and __WARN__ handlers so
174 they lose their magic and will not cause segfaults later
175 (testcase by Andrey Sagulin).
176 - improved Coro::State documentation a bit.
177 - Coro::Debug::command now flushes the output.
178 - add hack detection code for x32 abi, because the braindead slugs
179 who designed that made it look exactly like x86_64 without
180 providing proper compile time symbols to test for it. as a result,
181 this detection cannot work reliably.
182 - valgrind stack registering was broken.
183 - do not rely on Time::HiRes anymore in Coro::Debug.
184
1856.23 Fri Dec 7 23:36:37 CET 2012
186 - use experimental fiber implementation on native windows
187 perls.
188 - use sizeof (void *) as multiplication factor for stack sizes,
189 to accomodate the totally braindamaged microsoft 64 bit "os".
190 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
191 activeperl 5.16 crashes when PerlIO_define_layer is called due
192 to some bug in the perl dll, strawberry perl at least passes
193 the testsuite.
194 - implement Coro::Handle->peeraddr/host/port, for slightly
195 improved compatibility with LWP.
196 - implement 5.17 compatibility by almost blindly applying a
197 good-looking patch by Father Chrysostomos.
198 - move stack management functions into libcoro 3.
199 - libcoro version 3 "released".
200 - support magic values as timed_io_once args.
201 - recommend AnyEvent 7+ or EV 4+, also require EV
202 version 4 or newer for Coro::EV.
203
2046.10 Tue Oct 9 01:14:27 CEST 2012
205 - updated ecb.h, it had a typo that caused it to not compile on many
206 big endian systems (reported by many people).
207 - disable memory fences in ecb.h to improve portability.
208
2096.09 Sat Oct 6 23:25:02 CEST 2012
210 - Coro::EV I/O watchers were not interruptible by exceptions
211 (Coro::State::throw) (testcase by sten).
212 - ->throw now puts threads into the ready queue, as this seems to
213 be expected by existing code, and code that doesn't cope with spurious
214 wakeups needs fixing anyway.
215 - use fd -1 in mmap.
216 - cast I32 to int in error message printf.
217 - warn about broken so-called "hardened" kernels.
218
2196.08 Fri Apr 13 12:05:47 CEST 2012
220 - be more aggressive about exiting like perl does - formerly,
221 exiting from the non-main thread would not execute END blocks.
222
2236.07 Fri Nov 11 21:21:48 CET 2011
224 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
225 - update ecb.h.
226
2276.06 Mon Aug 8 23:59:48 CEST 2011
228 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
229 backend, which is an order of magnitude slower. unfortunately, cygwins
230 pthread implementation isn't very complete either, so allocate the stack
231 twice just to be sure.
232 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
233 in its ithreaded perl - best recompile cygwin and use the 'w'indows
234 backend for much better performance. also disable ithreads for
235 even better performance...).
236
2376.05 Thu Aug 4 21:36:36 CEST 2011
238 - blush, condvar values would not be propagated from send to recv anymore
239 (reported by Chip Salzenberg).
240 - use exponential increase for the readline buffer length in
241 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
242
2436.04 Wed Aug 3 17:15:45 CEST 2011
244 - use even more efficient and more compatible condvars for
245 compatibility to AnyEvent 6.x :)
246 - more inconsequential ecb.h updates.
247
2486.03 Wed Aug 3 11:41:30 CEST 2011
249 - change how Coro patches AnyEvent condvars for compatibility to
250 AnyEvent 6.x.
251 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
252 hasn't been included.
253
2546.02 Wed Jul 13 04:35:19 CEST 2011
255 - "improve portability to Gentoo" - gentoo manages to put perl variables
256 in memory areas that are farther than 2gb apart, which the jit couldn't
257 handle and barfed. now it's just a bit slower on gentoo and similar
258 systems.
259
2606.01 Sun Jul 3 12:31:14 CEST 2011
261 - workarounds are good, but the test for whether pthreads are used
262 was not good. this one should be better.
263 - check differently whether gcc generates cfi instructions itself.
264
2656.0 Wed Jun 29 19:43:35 CEST 2011
266 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
267 destroyed and cleaned up automatically (e.g. async { schedule }).
268 - implement a JIT compiler for part of the thread switch code,
269 which gives a 50% speed improvement on threaded perls, and
270 about 4% on non-threaded perls (so threaded perls now finally
271 reach about half the speed of non-threaded perls).
272 - slightly modernise Coro::Intro, add section about rouse functions.
273 - avoid DEFSV and ERRSV, giving another 10% improvement
274 in thread switching.
275 - Coro::State->is_destroyed is now called is_zombie.
276 - implement a Coro->safe_cancel method that might fail, but
277 cancels in a "safer" way if it succeeds.
278 - add preliminary support for DEBUGGING perls.
279 - get rid of two hash-accesses when initialising a new Coro - this
280 speeds up coro creation by almost a factor of two.
281 - croak when a coro that is being cancelled tries to block
282 (e.g. while executing a guard block), instead of crashing or
283 deadlocking.
284 - use a more robust and also faster method to identify Coro::State
285 objects - speeds up everything a bit.
286 - implement Coro->cancel in XS for a 20% speed improvement, and to
287 be able to implement mutual cancellation.
288 - speed up context switches by a percent or two by more efficiently
289 allocating context stack entries.
290 - implement Coro->join and Coro->on_destroy in XS for a speedup and
291 a reduction in memory use.
292 - cancelling a coro while it itself is cancelling another coro is
293 now supported and working, instead of triggering an assertion.
294 - be a bit more crash-resistant when calling (buggy) on_destroy
295 callbacks (best effort).
296 - move on_destroy into the slf_frame, to allow extension slf
297 functions to have destructors.
298 - get rid if coro refcounting - simply crash in other interpreter
299 threads by nulling the pointers on clone.
300 - simplify warn/die hook handling when loading Coro - the convoluted
301 logic seems to be no longer neccessary.
302 - use libecb instead of our own home-grown gcc hacks.
303 - document alternatives to Coro::LWP. Please use them :)
304 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
305 sigaltstack. Really. wine suffers from it, erlang suffers from it,
306 and it's known since at least 2006.
307
3085.372 Wed Feb 23 06:14:30 CET 2011
309 - apparently mingw doesn't provide a working gettimeofday, try to
310 work around that by relying on Time::HiRes (indirectly brought to
311 my attention by Max Maischein).
312 - fix some portability issues when Time::HiRes was used.
313
3145.371 Mon Feb 21 14:36:08 CET 2011
315 - backport to windows process emulation code again.
316
3175.37 Sat Feb 19 07:49:44 CET 2011
318 - add a big "Coro thread life cycle" section to "man Coro".
319 - try a tentative workaround against the breakage that 5.13 has
320 introduced without depreciation period. sigh.
321 - no longer use Time::HiRes if gettimeofday is available, which
322 saves quite a lot of memory.
323
3245.36 Sun Feb 13 05:33:41 CET 2011
325 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
326 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
327 method call.
328 - undocument Coro::Timer::sleep and obsolete whole module.
329 - optimise Coro::Timer::timeout memory and cpu usage.
330 - slightly updated Coro::Intro for recent changes.
331 - do not initialise PL_dirty anymore.
332
3335.25 Thu Nov 11 01:08:39 CET 2010
334 - try a different approach on netbsd - netbsd 5 finally has marginally
335 working pthreads, but still broken ucontext/sigaltstack.
336 - openbsd 4.8 finally got their act together, Coro works out of the box
337 with asm, setjmp and pthreads (no change, just informational).
338
3395.24 Sat Oct 23 11:27:12 CEST 2010
340 - port to the EV 4.0 API.
341 - work around bugs in mingw32, making strawberry perl work
342 out of the box.
343 - correctly modify Coro::AIO function prototypes
344 so that they reflect the "no optional parameters" rule.
345 - "ported" libcoro to C++.
346
3475.23 Mon May 17 18:50:42 CEST 2010
348 - be more resistant to ordering changes when initialising
349 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
350 Waldorf).
351 - document that perl 5.12 deliberately removed support for cloning.
352
3535.22 Wed Apr 14 03:55:35 CEST 2010
354 - correctly return udnef on errors in Coro::Handle::read/write
355 (testcase by Marc Mims).
356 - convert Coro::Util into a "perl compatibility wrapper" - the functions
357 are less useful now, but are drop-in replacements for existing
358 functions, listing better alternatives in the documentation. This also
359 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
360 with Coro::Util::inet_aton.
361 - do not override $Coro::idle unconditionally in Coro.pm, as other
362 modules could have provided their own idle coro already
363 (for exmaple, Coro::AnyEvent).
364 - fix Coro::Util::gethost* functions.
365 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
366
3675.21 Wed Dec 16 07:19:51 CET 2009
368 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
369 together.
370 - add some examples on how to combine other event loops with Coro in
371 Coro::AnyEvent, and how to run it (and not to block). Seems to be
372 the most common source of confusion.
373 - try to catch people naively blocking in an event callback.
374 - work around the perl filehandle bug issue in conjunction with
375 older common::sense (as indirectly pointed out by ZSystem).
376 - clarify the "not from signal handlers" section.
377
3785.2 Sun Oct 4 14:54:24 CEST 2009
379 - Coro::Storable destroyed the prototypes of the functions it wrapped.
380 - export rouse_cb and rouse_wait by default now.
381 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
382 - new method $state->swap_sv.
383 - added section on "windows process emulation" to the manpage,
384 after a not-so-fruitful (nor-friendly) "discussion" with chip
385 salzenberg (discussion implies arguments, but his only arguments
386 were ad-hominems, one wonders why he started it in the first
387 place). I hope this explains it well enough for him to understand,
388 and maybe well enough for others to understand.
389 - use common::sense everywhere now.
390 - idle callbacks are no longer supported, use idle coros instead.
391 - print a thread listing when a deadlock is detected.
392
3935.17 Sat Aug 22 23:09:31 CEST 2009
394 - work around a bug in the perl debugger causing crashes
395 when running under the debugger by marking _pool_handler as nodebug.
396 - speed up Coro::async considerably.
397 - try some hacks to get netbsd to work "more often" - their broken
398 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
399 - convert Coro to AE 5.0 API.
400
4015.162 Tue Jul 28 04:04:03 CEST 2009
402 - perl 5.8.2 is now minimum requirement.
403 - skip t/19_handle.t on broken windows perls.
404
4055.161 Wed Jul 22 04:47:38 CEST 2009
406 - Coro::AnyEvent::poll could have a different prototype when EV was
407 used as backend (analyzed by Tatsuhiko Miyagawa).
408 - Coro::AnyEvent errornously initialised the event loop when loaded,
409 not on demand.
410 - try to workaround rare */t/01_unblock.t failures.
411
4125.16 Tue Jul 21 01:44:37 CEST 2009
413 - Coro::AnyEvent failed to hook into the event loop
414 when no threads had been readied between detecting
415 the event loop and actually running it.
416 - considerably speed up Coro::Select by taking avdantage
417 of AnyEvent > 4.8 and some other optimisations.
418 - implement paragraph readline mode in Coro::Handle
419 (based on patches by Zsbán Ambrus).
420 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
421 and analyzed by Yasuhiro MATSUMOTO).
422 - clarified libcoro license and copyright.
423 - someone stole my EXTRA_META!!!
424 - implement Coro::Select::patch_pp_sselect and it's brother,
425 for hardcode select overriding.
426
4275.151 Mon Jul 6 05:41:57 CEST 2009
428 - backport to windows process emulation code again (patch by
429 Yasuhiro MATSUMOTO).
430 - slightly update Coro::MakeMaker.
431
4325.15 Tue Jun 30 10:28:06 CEST 2009
433 - deprecate Coro::Socket, document how to get ipv6 support via
434 AnyEvent::Socket instead.
435 - implement signal->wait ($cb) interface, similar to semaphores.
436 - work around SvOK not supporting getmagic, so we have to getmagic
437 to test for undef :/ (reported by Matthias Waldorf).
438 - load Coro::AnyEvent in all modules using AnyEvent.
439 - work around perl corrupting our internal data structures,
440 reported by Tokuhiro Matsuno.
441 - enable per-coroutine real and cpu time gathering
442 (Coro::State::enable_times).
443
4445.14 Wed Jun 24 01:37:48 CEST 2009
445 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
446 as this seems to have been a difficult concept (poll, sleep, idle,
447 idle_upto).
448 - add Coro::AnyEvent::readable/writable functions.
449 - clarify Coro::EV/Event/AnyEvent manpages.
450 - free per-thread global scalars in the thread calling ->cancel, to
451 avoid crashes when $_, $@ etc., are magical but some of those
452 had already been freed.
453 - "unexperimentalise" the callback interface for Coro::Semaphore.
454 - speed up ready queue management/context switching by using a linked
455 list instead of an array (~5-10%).
456 - implement "watch" command in Coro::Debug shells.
457 - for fun, implement time-slicing as an example in the manpage.
458 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
459 Coro::Event, use the normal AnyEvent handling instead of dieing
460 (the same is true for Coro::Handle).
461 - properly document Coro::EV::timed_io_once.
462 - avoid unneccessary ->cancel calls in Coro::Handle.
463 - maybe make it work on mingw32 with win32 backend
464 (based on patch by Yasuhiro Matsumoto).
465
4665.132 Fri May 29 09:00:39 CEST 2009
467 - do not keep a reference to the argument itself in
468 Coro::Semaphore::guard, as it could change later.
469 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
470 should support a prepare callback.
471
4725.131 Mon Mar 16 23:20:37 CET 2009
473 - implement and document Coro->suspend, Coro->resume.
474 - fix Coro::Select implementation to not (often) close
475 the passed file descriptors (testcase provided by pippijn).
476
4775.13 Mon Dec 15 21:51:42 CET 2008
478 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
479 - don't set diehook to C<undef> but instead to NULL, to avoid
480 spurious warnings.
481 - fix a lot of bugs in Coro::SemaphoreSet.
482 - Coro::SemaphoreSet will less often create a semaphore needlessly.
483 - add Coro::SemaphoreSet::count and wait methods.
484 - take advantage of the new Guard module.
485 - deprecate Coro::guard.
486 - try to fix the dreaded 01_unblock tests once more. I hate it when
487 testsuites need more fixing than the code they are supposed to test.
488 - croak in more cases when a required callback isn't resolvable.
489 - fix some minor issues in Coro::State->call/eval.
490 - use current coroutine context instead of a temporary one
491 when temporarily switching to another coroutine.
492 - do not call C-level on_destroy handlers during global destruction,
493 to avoid needless segfaults.
494
4955.12 Sun Dec 7 13:30:38 CET 2008
496 - add default config for MirOS, which seems to be a bug-to-bug
497 compatible fork of openbsd ("world domination by releasing
498 openbsd cvs before the openbsd folks do it" or so :).
499 - free_padlist did destroy the names pad, not good, but didn't
500 seem to bother perl - this could fix issues such as eval ""
501 inside a function called from multiple coroutines.
502 - use a different method to detect destruction time.
503 - be more careful when freeing padlists just before global
504 destruction.
505 - fixed and expanded the call/cc example.
506 - renamed _terminate to _coro_run.
507 - new method Coro::Channel->shutdown.
508 - try pthreads on openbsd <4.4 (broken sigaltstack, will
509 pthreads fare better?).
510 - be less picky about destroying "a" running coroutine.
511
5125.11 Tue Nov 25 21:49:05 CET 2008
513 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
514 working code with failed assertions, introducing more bugs than
515 it fixes, requiring elaborate workarounds :(
516
5175.1 Mon Nov 24 08:54:59 CET 2008
518 - wrote a small introductory tutorial - Coro::Intro.
519 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
520 - Coro::Select did errornously dup the file descriptors
521 and didn't work with all AnyEvent backends.
522 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
523 LWP data transfers.
524 - disassociate c contexts from coro objects - this is agruably more
525 correct, but mostly allows sharing of cctxs between coro and state
526 objects, for added memory savings and speed increases.
527 - bumped $Coro::POOL_RSS up to 32kb by default.
528 - no longer set the optype to OP_CUSTOM, as B::* understandably
529 doesn't like this very much (and we *are* a type of entersub).
530 - implement state cloning, just to prove that call/cc can be done.
531 - automatically load Coro::AnyEvent in Coro::Handle.
532 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
533 do not have this method (and don't need it either).
534 - speed up generic anyevent methods in Coro::Handle by using rouse
535 callbacks.
536 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
537 others. It also makes the debugger happier, as you can trace
538 through the idle threads now.
539 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
540 - eg/event was pretty broken.
541 - better 5.8.6 compatibility.
542
5435.0 Thu Nov 20 10:35:05 CET 2008
544 - NEW ARCHITECTURE: use the latest 4.x version if you experience
545 stability issues.
546 - bump API version to 7 - all dependents must be recompiled.
547 - removed timed_* functions - they were not being used anyways
548 and should be replaced by a more generic mechanism -
549 and were annoying to support anyways :)
550 - removed SemaphoreSet's waiter method - use sem method instead.
551 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
552 - async_pool did free a scalar value twice
553 ("Attempt to unreference...").
554 - fix a longstanding bug where calling terminate on a coro that
555 was waiting for a semaphore that was just becoming available
556 would cause a deadlock (semaphore would get into a state where
557 it was available but waiters were still blocked).
558 - calling throw on a coroutine that is waiting for a semaphore will
559 no longer make it acquire the semaphore (and thus leak a count).
560 - perl's process emulation is now not even theoretically supported
561 anymore.
562 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
563 conversion of callback-style to blocking-style.
564 - new methods $coro->schedule_to and ->cede_to, to specifically
565 schedule or cede to a specific coroutine.
566 - new function Coro::Semaphore::wait.
567 - use named constants in Coro::Channel (Richard Hundt).
568 - directly patch the entersub opcode calling SLF functions (cede,
569 transfer and so on). this does speed up context switching, but
570 more importanly, it frees us from the hardcoded behaviour of
571 entersub, so we might actually be able to return something from
572 those functions and atcually create new ones.
573 - take advantage of __builtin_frame_address on gcc.
574 - expose THX in coroapi (not sure whether this was a wise decision,
575 as "threaded" perls are running at half speed anyways).
576 - implement execute_slf (schedule-like-function) interface that makes
577 it possible to implement schedule-like-functions in XS.
578 - use new SLF interface to massively speed up Coro::EV by roughly a
579 factor of two.
580 - used new SLF interface to massively speed up Coro::Semaphore by a
581 factor of three.
582 - used new SLF interface to speed up Coro::AIO by roughly a factor of
583 four and reduce its memory usage considerably.
584 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
585 for a nice speedup and vastly more correct behaviour. Also implement
586 a new method "sem" to get at the underlying semaphore object.
587 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
588 (in comparison) 20-40% speedup.
589 - used new SLF interface to reimplement Coro::Signal gaining
590 some unknown (because I was too lazy), but certain, speedup, and also
591 making signals reliable for the first time.
592 - used new SLF interface and other optimisations to speed up async_pool
593 by a factor of two. It also doesn't rely on perl's exception mechanism
594 to exit anymore. The overhead for terminating an async_pool, coro over
595 a normal async is now very small.
596 - sped up coroutine creation/destruction by 40%.
597 - forgot to include Coro/libcoro/README in the dist for all these years.
598 - work around a freebsd pthreads bug (manual testcancel is required as
599 pthread_cond_wait isn't a cancellation point on freebsd).
600 - use new rouse functions to speed up and simplify Coro::BDB.
601 - make "prefer perl native functions" work with threaded perls.
602 - condense Coro::Debug ps output, hint at v and w flags.
603 - (libcoro) lots of minor cleanups and portability improvements.
604
6054.914 Wed Nov 19 12:54:18 CET 2008
606 - fix a disastrous bug in the readline optimisation
607 introduced in 4.801.
608
6094.913 Sat Nov 15 07:58:28 CET 2008
610 - async_pool did free a scalar value twice
611 ("Attempt to unreference...").
612
6134.912 Thu Nov 13 18:31:23 CET 2008
614 - minor cleanups.
615 - use much larger stacks on linux and perl < 5.8.8.
616 - Coro::Debug::new_unix_server did not unlink the socket
617 when destroyed.
618
6194.911 Tue Nov 11 04:26:17 CET 2008
620 - "port" to threaded perls.
621
6224.91 Mon Nov 10 05:36:38 CET 2008
623 - the ->throw exception object no longer leaks.
624 - creating a new cctx leaked a scopestack entry (memleak).
625 - new coroutines didn't get created with a zero flags field
626 (unknown impact).
627 - calling ->throw on a not-yet-started coroutine should now work
628 instead of being ignored.
629 - ->throw is now supported on Coro::State objects.
630 - clean up cctx creation code a bit.
631 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
632
6334.9 Sat Nov 8 17:45:27 CET 2008
634 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
635 - (libcoro) no longer leak threads in the experimental pthread backend,
636 also speed it up considerably.
637 - (libcoro) do not rely on makecontext passing void *'s unscathed.
638 - fix compiletime dependencies on libcoro in the Makefile.
639 - cctx_count wasn't always updated properly.
640 - Coro::State::cctx_stacksize wasn't applied correctly.
641 - new function Coro::State::cctx_max_idle.
642 - the default max number of idle C contexts is now 4.
643 - (libcoro) try harder to get _setjmp/_longjmp.
644 - (libcoro) cleanup and extend the libcoro API to officially
645 allow the creation of empty source contexts.
646 - very experimental workaround for the totally broken netbsd platform.
647 - tried to hack around openbsd bugs.
648
6494.804 Wed Nov 5 16:36:00 CET 2008
650 - Coro::Debug::new_unix_server would not create a non-blocking listening
651 socket, sometimes causing freezes.
652 - (libcoro) fix misaligned stack points for setjmp and assembly
653 methods, which can cause crashes on x86/x86_64 with a sufficiently
654 aggressive compiler.
655 - new function: Coro::Debug::new_tcp_server.
656 - move ->throw into the Coro class because it only works on coro objects.
657
6584.803 Mon Nov 3 17:16:12 CET 2008
659 - (libcoro) use a global asm statement to become independent of gcc
660 otpimisations for CORO_ASM (thanks to pippijn for the idea).
661 - try to workaround yet another broken bsd, this time dragonfly.
662
6634.802 Thu Oct 30 10:56:12 CET 2008
664 - support -fno-omit-frame-pointer on x86 with the assembly method.
665 - tune 01_unblock.t tests a bit.
666
6674.801 Wed Oct 22 18:33:37 CEST 2008
668 - improve readline speed for very long "lines".
669 - backport to 5.8.8.
670
6714.8 Thu Oct 2 13:34:40 CEST 2008
672 - new function Coro::AIO::aio_wait.
673 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
674 - greatly speed up and reduce memory usage of Coro::AIO requests.
675 - implement some other µ-optimisations.
676
6774.749 Mon Sep 29 14:40:12 CEST 2008
678 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
679
6804.748 Sat Sep 27 14:03:19 CEST 2008
681 - implement, but do not document, PerlIO::cede(granularity).
682 - Coro::Storable forgot to wrap Storable::pstore.
683 - work around the multitude of leaks and memory corruption
684 bugs in PerlIO::via by using our own C-level perliol. As a side
685 effect, Coro::Storable is now much, much, much faster.
686
6874.747 Tue Sep 23 01:59:41 CEST 2008
688 - fix a per-cv memleak (one empty array was leaked per
689 code reference).
690 - avoid a crash in coro->call|rss when the coroutine was already
691 destroyed (most noticably when using Coro::Debug::ps :)
692 - also protect *Storable::FILE.
693 - push up default storable granularity to 20ms.
694
6954.746 Sun Sep 21 03:22:20 CEST 2008
696 - be more insistent on locking Storable against reentrancy
697 in Coro::Storable.
698 - move swap_def?v and throw to Coro::State, as documented.
699
7004.745 Thu Jul 24 00:14:38 CEST 2008
701 - remove debugging code related to MgPV_nolen_const, also try to
702 make it compile with perl 5.8.6 (yes, apple apparently loves
703 outdated software). Reported by John S.
704
7054.744 Tue Jul 8 22:06:35 CEST 2008
706 - correctly provide default DIE/WARN handlers as documented.
707 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
708 implementations inside perl work fine for us.
709
7104.743 Mon Jun 16 00:21:57 CEST 2008
711 - when using Coro::EV without running EV::loop it could
712 result in busy-waiting for events, this has been fixed.
713 - reduce codesize and improve performance by using EV_DEFAULT_UC.
714
7154.742 Sat May 31 14:10:21 CEST 2008
716 - implement a workaround for (some) perl <5.8.8 versions.
717 - require EV 3.3+.
718
7194.741 Fri May 30 23:33:09 CEST 2008
720 - tell netbsd how utterly broken their imitation of an OS is
721 and refuse to build by default if pthreads are in use.
722 - switch to "s" method on all bsd's by default, as their ucontext
723 stuff seems just too broken.
724 - fix a bug in Coro::Select.
725
7264.74 Thu May 29 20:05:31 CEST 2008
727 - do not test Coro::LWP for lack of dependencies.
728 - do not test Coro::Select for lack of working perls.
729
7304.73 Thu May 29 2008
731 - fix a bug in Coro::EV which would cause it to block despite
732 there being runnable coroutines.
733 - sprinkle "no warnings" freely over everything, also suppress
734 warnings for some other modules.
735 - fix typo in WSAEWOULDBLOCK.
736
7374.72 Sun May 25 05:14:36 CEST 2008
738 - tweak META.yaml a bit, unfortunately, there is no documented way
739 to have optional dependencies with CPAN. doh :(
740 - avoid running some tests on windows because they would fail due to
741 perl bug (broken fork, broken pipes...).
742 - work around perl on windows bugs where perl returns undocumented
743 error codes for sysread, syswrite etc. by taking advantage
744 of AnyEvent's workaround for that problem.
745 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
746 a common perl implementation bug on windows.
747 - use unix domain sockets in testsuite to work around
748 common perl implementation bugs on widows (they are emulated by
749 tcp sockets on windows. ugh.)
750
7514.71 Sat May 24 20:01:27 CEST 2008
752 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
753 - take advantage of async name resolution of AnyEvent::Util.
754 - work around brutal inet_aton override in Coro::LWP.
755 - take advantage of the readyhook in Coro::EV, for smoother
756 scheduling.
757
7584.7 Sun May 11 00:32:19 CEST 2008
759 - completely reworked the Coro manpage.
760 - added Coro::AnyEvent, generic event loop integration.
761 - implement cancel, ready and kill commands in Coro::Debug.
762 - document find_coro in Coro::Debug.
763 - incompatible change: rename has_stack to has_cctx.
764 - Coro::AIO and Coro::BDB no longer force event model detection,
765 use AnyEvent::AIO and AnyEvent::BDB.
766
7674.6 Sat Apr 26 10:05:14 CEST 2008
768 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
769 by the Coro module. It was a mistake to have it in the first place.
770 - (experimental) support for activestate perl 5.10 (method "w").
771 - (experimental) support for strawberry perl (method "a").
772 - coro_sigelem_set did not return a value although it had to,
773 actual impact unknown.
774
7754.51 Mon Apr 14 13:28:27 CEST 2008
776 - make it compile again on 5.8.
777
7784.50 Thu Apr 10 09:43:17 CEST 2008
779 - I did it twice! (see 4.49).
780
7814.49 Sun Apr 6 21:23:31 CEST 2008
782 - grr, instead of compiling the recent changes on 5.10 only they
783 were compiled on 5.8 only.
784
7854.48 Sun Apr 6 20:36:46 CEST 2008
786 - allow coroutine switches during eval's under 5.10.x, as it seems
787 the parser is a per-interpreter option now, so this is safe
788 (this might fix the odd crashing bug).
789 - drop support for 5.9.x versions: they are dead, jim.
790
7914.47 Sun Apr 6 00:37:52 CEST 2008
792 - force cctx allocation on API calls: we know we need to force one
793 and gcc actually manages to confuse our heuristic nowadays,
794 leading to crashes and worse.
795 - document force_cctx.
796
7974.46 Fri Apr 4 22:05:43 CEST 2008
798 - upgrade libcoro, resulting in pthread-backend (which was only created
799 to fulfill the rules of the programming languages shootout).
800
8014.45 Thu Mar 13 11:55:36 CET 2008
802 - fix a file leaking bug in eg/mhyttpd that would allow
803 downloading of any file (reported by oesi).
804 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
805 (also add testcase).
806 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
807 apparently having loads of fun with that).
808 - implement $state->swap_defsv, ->swap_defav and document ->throw.
809
8104.4 Wed Feb 13 16:44:29 CET 2008
811 - only assume .cfi pseudo isns exist on GNU/Linux.
812 - add get_timed method to Coro::Channels.
813 - fixed Coro::Debug manpage.
814 - perl 5.11 compatbility improvement based on patch by
815 Andreas König.
816
8174.37 Sun Jan 20 11:25:23 CET 2008
818 - soften the check for an active parser for perl 5.10 (avoiding
819 "Coro::State::transfer called while parsing" in many cases).
820
8214.36 Sun Jan 13 10:53:56 CET 2008
822 - reset diehook when terminating from an async_pool as to not
823 trigger any __DIE__ handlers.
824
8254.35 Sun Jan 13 04:14:13 CET 2008
826 - "bt" debug command now displays any exceptions
827 from longmess and also skips the innermost
828 stackframes, giving more useufl output.
829 - allow backtraces in coroutines blocked in Coro::EV,
830 at a <1% speed hit.
831 - handle localising of $SIG{__WARN__} and __DIE__
832 properly (with a proper amount of dirty hacking).
833
8344.34 Sat Dec 22 17:49:53 CET 2007
835 - upgrade to EV version 2.0 API.
836
8374.33 Mon Dec 17 08:36:12 CET 2007
838 - make Coro::AIO etc. loadable in the absence of EV.
839
8404.32 Mon Dec 17 07:46:02 CET 2007
841 - majorly improved Coro::Handle's performance with Coro::EV.
842 - implemented the until now mythical Coro::BDB module.
843 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
844
8454.31 Wed Dec 5 12:32:39 CET 2007
846 - remove warn statement form Coro::Util.
847
8484.3 Tue Dec 4 20:33:14 CET 2007
849 - calls to the idle function could cause stack corruption
850 when the stack changed.
851 - do no longer rely on the presence of EV::DNS if EV is used
852 (because it is gone), but instead take avdantage of EV::ADNS
853 if available.
854 - add ($) prototypes to all functions in Coro::Storable.
855 - use a conventional (safer) idle callback in Coro::EV.
856 - do accept quickly in Coro::Debug to avoid endless loops.
857
8584.22 Fri Nov 30 16:04:04 CET 2007
859 - really use optimised versions for Event and EV in Coro::Util
860 and Coro::Handle.
861
8624.21 Sun Nov 25 10:48:59 CET 2007
863 - fix a spurious memory read.
864 - Coro::EV no longer keeps the eventloop "alive".
865
8664.2 Fri Nov 9 20:47:05 CET 2007
867 - enable/disable tracing from a new coroutine, not a pooled one.
868 - fix a memleak in Coro::Event.
869 - removed killall call from fork_eval.
870 - made sure store_fd is already loaded so that fork_eval
871 does not have to parse autoload in each subprocess.
872 - only use assembly method if -O switch is in $Config{optimize}.
873 - add (optional) Coro::EV module, so far the best event loop module
874 directly supported by Coro.
875 - if the event model is EV, use EV::DNS to resolve
876 stuff in Coro::Util.
877 - don't get confused by multiple event notifications in Coro::Handle.
878 - initial support for EV (libevent interface).
879 - require Event and EV using configure_requires, to force their existance.
880
8814.13 Wed Oct 24 07:26:45 CEST 2007
882 - add Coro::Storable::blocking_thaw.
883 - use a vastly more complicated technique to localise
884 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
885 - use a coroutine for the idle callback Coro::Event,
886 instead of running Event in the current coroutine context.
887 This also catches recursive invocations.
888 - actually report fork errors in gethostbyname and inet_aton.
889
8904.11 Thu Oct 11 02:40:24 CEST 2007
891 - port to threaded perls.
892
8934.1 Thu Oct 11 02:38:16 CEST 2007
894 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
895 be local to each coro (see Coro::State).
896 - incompatible change: for very deep reasons, cede and cede_notself
897 cannot return anything, so nothing will be returned.
898 - possibly bring back 5.10 compatibility (untested).
899 - work around stupid (and wrong) warning on 5.10 :(.
900 - overlay the saved state over the context stack. This saves
901 a few hundred bytes per coroutine on average and also
902 speeds up context switching a bit.
903 - further tune default stack sizes.
904 - (more) correctly calculate stack usage in coro_rss.
905 - Coro::Storable::blocking_* did not properly lock
906 resulting in races between coroutines.
907 - added Coro::Storable::guard.
908 - stopping to trace a coroutine could destroy the cctx of
909 an unrelated coroutine.
910 - explain the relationship between Perl and C coroutines in
911 more detail in Coro::State.
912 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
913 causing a fork per resolve. This also affected Coro::Socket.
914 - switch to a separate stack in $coro->call/eval to avoid
915 invalidating pointers.
916
9174.03 Sat Oct 6 21:24:00 CEST 2007
918 - added Coro::throw method.
919 - minor code cleanups.
920
9214.02 Sat Oct 6 02:36:47 CEST 2007
922 - fix a very minor per-coroutine memleak (a single codereference).
923 - fixed a bug where the currently in-use c context would be freed
924 prematurely (can happen only when programs change the stacksize
925 or use tracing).
926 - tracing can no longer keep a coro alive after it terminated.
927 - do static branch prediction in the common path for gcc. gives
928 about 2-5% speed improvement here.
929
9304.01 Fri Oct 5 22:10:49 CEST 2007
931 - instead of recreating *a* standard output handle we simply
932 use STDOUT, which is faster and hopefully more robust.
933
9344.0 Fri Oct 5 12:56:00 CEST 2007
10 - incompatibly changed Coro::Storable::freeze. 935 - incompatibly changed Coro::Storable::freeze.
11 - major new feature: added Coro::Debug, for interactive coroutine 936 - major new feature: added Coro::Debug, for interactive coroutine
12 debugging, tracing and much more. 937 debugging, tracing and much more.
938 - major bug fix: unbelievable, but true: $_, $/ and many other
939 "saved" variables actually weren't being saved. This has been fixed,
940 of course, while increasing performance while losing all the save
941 flags.
942 - save flags are gone, and all the api functions dealing with them.
13 - added Coro::Semaphore::adjust. 943 - added Coro::Semaphore::adjust.
14 - added Coro::Util::fork_eval. 944 - added Coro::Util::fork_eval.
15 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}. 945 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
946 - added Coro::killall.
16 - reduce initial stack sizes to allow for "micro-coroutines". 947 - reduce initial stack sizes to allow for "micro-coroutines".
17 - better async_pool resource management, moved parts of async_pool 948 - better async_pool resource management, moved parts of async_pool
18 handling to XS (major speed improvement). 949 handling to XS (major speed improvement).
19 - actually croak before modifying important data structures. 950 - actually croak before modifying important data structures.
20 - refuse to transfer while compiling. 951 - refuse to transfer while compiling.
21 - possibly support eval EXPR better now. 952 - possibly support eval EXPR better now.
22 - enable assembly method per default on linux+bsd x86+amd64. 953 - enable assembly per default on linux+bsd x86+amd64.
954 - all internal members were renamed _something for easier subclassing.
23 - many minor tweaks. 955 - many minor tweaks.
24 956
253.63 Wed May 16 14:10:06 CEST 2007 9573.63 Wed May 16 14:10:06 CEST 2007
26 - implement handcoded assembly for x86/amd64 SVR ABI. 958 - implement handcoded assembly for x86/amd64 SVR ABI.
27 959
… …
39 but stateful semantics. 971 but stateful semantics.
40 - fixed a lot of typos in Coro.pm (patch submitted by David 972 - fixed a lot of typos in Coro.pm (patch submitted by David
41 Steinbrunner, which applied flawlessly). 973 Steinbrunner, which applied flawlessly).
42 974
433.6 Sat Apr 14 17:13:31 CEST 2007 9753.6 Sat Apr 14 17:13:31 CEST 2007
44 - added some bugfixes to get eg/myhttpd working again. 976 - added some bugfixes to get eg/myhttpd working again.
45 - added Coro::Storable for often-cede'ing freeze/thaw. 977 - added Coro::Storable for often-cede'ing freeze/thaw.
46 - try to do a clean exit when a coroutine calls exit 978 - try to do a clean exit when a coroutine calls exit
47 (EXPERIMENTAL). 979 (EXPERIMENTAL).
48 - got rid of indirect call through _coro_init. 980 - got rid of indirect call through _coro_init.
49 - updated the partly antique examples in eg/ to 981 - updated the partly antique examples in eg/ to

Diff Legend

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