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

Comparing Coro/Changes (file contents):
Revision 1.222 by root, Sat Apr 14 15:13:00 2007 UTC vs.
Revision 1.677 by root, Sun Dec 16 09:33:01 2018 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: better (earlier) error messages on broken transfer's 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: destroy gets only called on cancel: fix(?) 4TODO: channel->maxsize(newsize)?
5TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
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?
6 10
11 - the autogenerated prototypes of Coro::AIO now add ";" at the end
12 to avoid generating named unary operators and messing up parsing
13 (mostly affects aio_mlockall 4 | 8 which was parsed as
14 (aio_mlockall 4) | 8).
15 - do not use a prototype for select in Coro::Select, as CORE::select also
16 doesn't have one.
17
186.52 Tue Aug 14 18:49:43 CEST 2018
19 - libcoro update: allow sharing of coro threads using the pthreads
20 backend among different pthreads, as is required by Coro::Multicore.
21 - always allow ->call and ->eval when the coro is the current coroutine. this
22 makes it possible, among other things, to call these on $Coro::main.
23 - minor code cleanups.
24
256.514 Thu Aug 31 18:25:31 CEST 2017
26 - safe_cancel no longer croaks when called on already-destroyed
27 thread, but instead "just works".
28
296.513 Sat Jul 15 01:18:51 CEST 2017
30 - experimental 5.22 comaptibility changes (PadlistNAMES).
31 - removed attributes example and made sure the other examples
32 still work.
33 - fix some statevar types for newer perls, should be of no consequence.
34
356.512 Fri Jul 14 05:03:17 CEST 2017
36 - seems to pass test suite down till 5.8.9, so make this the
37 minimum requirement in Makefile.PL.
38 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
39 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
40 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
41 fortified longjmp variants in some libcs.
42 - re-try JIT allocation without PROT_EXEC, to hopefully improve
43 portability (still doesn't seem to work around broken selinux
44 mmap everywhere).
45 - speling fixes (Xavier Guimard).
46
476.511 Sun Jun 26 23:44:50 CEST 2016
48 - make it compile with threaded perls again.
49 - simplify and speed up __DIE__ and __WARN__ handling.
50
516.51 Sat Jun 25 20:57:20 CEST 2016
52 - 6.5 release didn't compile on most older perls due to
53 a logic error enabling experimental 5.24 code for all versions.
54 - since various broken or deficient patches float around
55 to work around the vtbl API breakage in perl, let's provide
56 our own workaround, which seems to be much less invasive
57 and more compatible than the approaches seen so far. YMMV.
58 - work around assertion failure in perl_destruct on 5.24
59 (analyzed by Dave Mitchell).
60
616.5 Wed Jun 22 22:23:50 CEST 2016
62 - swap_sv swapping order was not symmetrical, causing
63 wrong swaps when swap_sv was used multiple times on the
64 same sv in the same thread.
65 - swap_sv calls can now be undone by calling it again
66 with the same variables.
67 - swap_sv calls will now be undone in async_pool threads.
68 - split Coro::Semaphore::up/adjust into separate xs functions
69 for better error reporting, at slight codesize increase.
70 - (libcoro) arm assembly support, please test and report.
71 - adjust to PL_savestack changes in perl 5.24 (adapted from
72 the debian patch, which unfortunately gets it wrong).
73
746.49 Sat Oct 17 01:40:12 CEST 2015
75 - throwing an exception to a thread waiting in
76 Coro::Handle using Coro::EV did not stop the watchers,
77 causing the next call to fail (testcase by Martin Pritchard).
78 - bump minimum perl version to 5.10.
79
806.48 Sun Oct 4 19:03:51 CEST 2015
81 - fix memory leak when Coro::Handle uses Coro::EV internally
82 (testcase by Sten Sten).
83 - update code to libev 4 API (internally, Coro still used the
84 version 3 API calls).
85 - remove unused hv_sig.
86
876.47 Sat Jul 11 03:58:20 CEST 2015
88 - swap_svs - NVs can be in the SV head beginning with 5.20.
89
906.46 Tue Jun 30 14:36:30 CEST 2015
91 - restore portability to perls with windows fork emulation
92 (patch by Petr Písař).
93
946.45 Tue Jun 30 01:40:08 CEST 2015
95 - restore portability to perl 5.16 and below.
96
976.44 Tue Jun 30 00:41:54 CEST 2015
98 - avoid segfaulting (or worse) when tracing a canceled thread.
99 - the Event module silently broke it's public hook API, causing
100 failures on perls compiled with -Duselongdouble. Requiring
101 the latest version of Event because I don't know which release
102 changed this (it's not mentioned in the ChangeLog).
103 - allow xs level enterleave hooks via CoroAPI.
104
1056.43 Thu Jun 4 15:38:14 CEST 2015
106 - use stability canary.
107 - port to stableperl-5.22.0-1.001.
108 - update libecb.
109
1106.42 Wed Feb 11 20:29:52 CET 2015
111 - Coro::SemaphoreSet->try did not actually work (analyzed by
112 SATO Kentaro).
113 - upgrade libecb to be C11 compliant.
114
1156.41 Sat Sep 6 22:08:46 CEST 2014
116 - restore portability to perl 5.8.x.
117 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
118 up some memory, but fixes "use feature" and similar modules
119 when used inside a Coro.
120 - allow zero as argument to Coro::Channel to mean the same thing
121 as no argument. this works with older versions as well,
122 but wasn't legal until now.
123 - slightly better c header file detection.
124
1256.39 Mon Jun 2 00:00:08 CEST 2014
126 - work around more incompatible changes in 5.20.
127
1286.38 Sun Jun 1 21:54:23 CEST 2014
129 - check that perl slots actually have enough space to hold
130 interpreter variables.
131 - untested port to perl 5.19 (64 bit tmps indices) (reported
132 by Andreas König).
133 - croak when cancel is called without a thread context.
134
1356.37 Tue Mar 4 13:27:33 CET 2014
136 - *sigh*, removed leftover debugging code from debugging a
137 perl bug, of all things.
138
1396.36 Tue Mar 4 07:11:59 CET 2014
140 - semaphores would not clear the destroy hook when interrupted
141 by ->throw, causing segfaults or worse.
142 - ->throw on a thread waiting for a semaphore did not acquire
143 the semaphore, but also didn't wake up other waiters,
144 possibly causing a deadlock.
145 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
146 to report a full stacktrace, which should help find out
147 where the actual call is.
148 - "atomically" destroy data in slf_destroy, because it is
149 the right thing to do, just in case.
150 - disable ecb.h libm dependency, to compile on ancient systems
151 or under adverse conditions.
152
1536.33 Mon Nov 18 11:26:27 CET 2013
154 - do not crash when freeing padlists with holes (in 5.18).
155 - tentative SVt_BIND 5.19 port/fix.
156
1576.32 Tue Nov 5 15:35:35 CET 2013
158 - use a new algorithm to derive padlists for perl 5.18. The old
159 one could lead to 0-pointer accesses inside perl (reported
160 by Darin McBride).
161
1626.31 Thu May 9 07:39:48 CEST 2013
163 - Coro::AIO requests would crash if the thread was ready'd
164 while the request was ongoing.
165
1666.29 Wed May 8 02:55:18 CEST 2013
167 - when an on_destroy handler destructs the coro currently being
168 destructed a perl scalar could be accessed after being freed,
169 likely causing a crash.
170
1716.28 Wed Mar 6 06:58:02 CET 2013
172 - clean remnants of existing __DIE__ and __WARN__ handlers so
173 they lose their magic and will not cause segfaults later
174 (testcase by Andrey Sagulin).
175 - improved Coro::State documentation a bit.
176 - Coro::Debug::command now flushes the output.
177 - add hack detection code for x32 abi, because the braindead slugs
178 who designed that made it look exactly like x86_64 without
179 providing proper compile time symbols to test for it. as a result,
180 this detection cannot work reliably.
181 - valgrind stack registering was broken.
182 - do not rely on Time::HiRes anymore in Coro::Debug.
183
1846.23 Fri Dec 7 23:36:37 CET 2012
185 - use experimental fiber implementation on native windows
186 perls.
187 - use sizeof (void *) as multiplication factor for stack sizes,
188 to accomodate the totally braindamaged microsoft 64 bit "os".
189 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
190 activeperl 5.16 crashes when PerlIO_define_layer is called due
191 to some bug in the perl dll, strawberry perl at least passes
192 the testsuite.
193 - implement Coro::Handle->peeraddr/host/port, for slightly
194 improved compatibility with LWP.
195 - implement 5.17 compatibility by almost blindly applying a
196 good-looking patch by Father Chrysostomos.
197 - move stack management functions into libcoro 3.
198 - libcoro version 3 "released".
199 - support magic values as timed_io_once args.
200 - recommend AnyEvent 7+ or EV 4+, also require EV
201 version 4 or newer for Coro::EV.
202
2036.10 Tue Oct 9 01:14:27 CEST 2012
204 - updated ecb.h, it had a typo that caused it to not compile on many
205 big endian systems (reported by many people).
206 - disable memory fences in ecb.h to improve portability.
207
2086.09 Sat Oct 6 23:25:02 CEST 2012
209 - Coro::EV I/O watchers were not interruptible by exceptions
210 (Coro::State::throw) (testcase by sten).
211 - ->throw now puts threads into the ready queue, as this seems to
212 be expected by existing code, and code that doesn't cope with spurious
213 wakeups needs fixing anyway.
214 - use fd -1 in mmap.
215 - cast I32 to int in error message printf.
216 - warn about broken so-called "hardened" kernels.
217
2186.08 Fri Apr 13 12:05:47 CEST 2012
219 - be more aggressive about exiting like perl does - formerly,
220 exiting from the non-main thread would not execute END blocks.
221
2226.07 Fri Nov 11 21:21:48 CET 2011
223 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
224 - update ecb.h.
225
2266.06 Mon Aug 8 23:59:48 CEST 2011
227 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
228 backend, which is an order of magnitude slower. unfortunately, cygwins
229 pthread implementation isn't very complete either, so allocate the stack
230 twice just to be sure.
231 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
232 in its ithreaded perl - best recompile cygwin and use the 'w'indows
233 backend for much better performance. also disable ithreads for
234 even better performance...).
235
2366.05 Thu Aug 4 21:36:36 CEST 2011
237 - blush, condvar values would not be propagated from send to recv anymore
238 (reported by Chip Salzenberg).
239 - use exponential increase for the readline buffer length in
240 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
241
2426.04 Wed Aug 3 17:15:45 CEST 2011
243 - use even more efficient and more compatible condvars for
244 compatibility to AnyEvent 6.x :)
245 - more inconsequential ecb.h updates.
246
2476.03 Wed Aug 3 11:41:30 CEST 2011
248 - change how Coro patches AnyEvent condvars for compatibility to
249 AnyEvent 6.x.
250 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
251 hasn't been included.
252
2536.02 Wed Jul 13 04:35:19 CEST 2011
254 - "improve portability to Gentoo" - gentoo manages to put perl variables
255 in memory areas that are farther than 2gb apart, which the jit couldn't
256 handle and barfed. now it's just a bit slower on gentoo and similar
257 systems.
258
2596.01 Sun Jul 3 12:31:14 CEST 2011
260 - workarounds are good, but the test for whether pthreads are used
261 was not good. this one should be better.
262 - check differently whether gcc generates cfi instructions itself.
263
2646.0 Wed Jun 29 19:43:35 CEST 2011
265 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
266 destroyed and cleaned up automatically (e.g. async { schedule }).
267 - implement a JIT compiler for part of the thread switch code,
268 which gives a 50% speed improvement on threaded perls, and
269 about 4% on non-threaded perls (so threaded perls now finally
270 reach about half the speed of non-threaded perls).
271 - slightly modernise Coro::Intro, add section about rouse functions.
272 - avoid DEFSV and ERRSV, giving another 10% improvement
273 in thread switching.
274 - Coro::State->is_destroyed is now called is_zombie.
275 - implement a Coro->safe_cancel method that might fail, but
276 cancels in a "safer" way if it succeeds.
277 - add preliminary support for DEBUGGING perls.
278 - get rid of two hash-accesses when initialising a new Coro - this
279 speeds up coro creation by almost a factor of two.
280 - croak when a coro that is being cancelled tries to block
281 (e.g. while executing a guard block), instead of crashing or
282 deadlocking.
283 - use a more robust and also faster method to identify Coro::State
284 objects - speeds up everything a bit.
285 - implement Coro->cancel in XS for a 20% speed improvement, and to
286 be able to implement mutual cancellation.
287 - speed up context switches by a percent or two by more efficiently
288 allocating context stack entries.
289 - implement Coro->join and Coro->on_destroy in XS for a speedup and
290 a reduction in memory use.
291 - cancelling a coro while it itself is cancelling another coro is
292 now supported and working, instead of triggering an assertion.
293 - be a bit more crash-resistant when calling (buggy) on_destroy
294 callbacks (best effort).
295 - move on_destroy into the slf_frame, to allow extension slf
296 functions to have destructors.
297 - get rid if coro refcounting - simply crash in other interpreter
298 threads by nulling the pointers on clone.
299 - simplify warn/die hook handling when loading Coro - the convoluted
300 logic seems to be no longer neccessary.
301 - use libecb instead of our own home-grown gcc hacks.
302 - document alternatives to Coro::LWP. Please use them :)
303 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
304 sigaltstack. Really. wine suffers from it, erlang suffers from it,
305 and it's known since at least 2006.
306
3075.372 Wed Feb 23 06:14:30 CET 2011
308 - apparently mingw doesn't provide a working gettimeofday, try to
309 work around that by relying on Time::HiRes (indirectly brought to
310 my attention by Max Maischein).
311 - fix some portability issues when Time::HiRes was used.
312
3135.371 Mon Feb 21 14:36:08 CET 2011
314 - backport to windows process emulation code again.
315
3165.37 Sat Feb 19 07:49:44 CET 2011
317 - add a big "Coro thread life cycle" section to "man Coro".
318 - try a tentative workaround against the breakage that 5.13 has
319 introduced without depreciation period. sigh.
320 - no longer use Time::HiRes if gettimeofday is available, which
321 saves quite a lot of memory.
322
3235.36 Sun Feb 13 05:33:41 CET 2011
324 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
325 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
326 method call.
327 - undocument Coro::Timer::sleep and obsolete whole module.
328 - optimise Coro::Timer::timeout memory and cpu usage.
329 - slightly updated Coro::Intro for recent changes.
330 - do not initialise PL_dirty anymore.
331
3325.25 Thu Nov 11 01:08:39 CET 2010
333 - try a different approach on netbsd - netbsd 5 finally has marginally
334 working pthreads, but still broken ucontext/sigaltstack.
335 - openbsd 4.8 finally got their act together, Coro works out of the box
336 with asm, setjmp and pthreads (no change, just informational).
337
3385.24 Sat Oct 23 11:27:12 CEST 2010
339 - port to the EV 4.0 API.
340 - work around bugs in mingw32, making strawberry perl work
341 out of the box.
342 - correctly modify Coro::AIO function prototypes
343 so that they reflect the "no optional parameters" rule.
344 - "ported" libcoro to C++.
345
3465.23 Mon May 17 18:50:42 CEST 2010
347 - be more resistant to ordering changes when initialising
348 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
349 Waldorf).
350 - document that perl 5.12 deliberately removed support for cloning.
351
3525.22 Wed Apr 14 03:55:35 CEST 2010
353 - correctly return udnef on errors in Coro::Handle::read/write
354 (testcase by Marc Mims).
355 - convert Coro::Util into a "perl compatibility wrapper" - the functions
356 are less useful now, but are drop-in replacements for existing
357 functions, listing better alternatives in the documentation. This also
358 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
359 with Coro::Util::inet_aton.
360 - do not override $Coro::idle unconditionally in Coro.pm, as other
361 modules could have provided their own idle coro already
362 (for exmaple, Coro::AnyEvent).
363 - fix Coro::Util::gethost* functions.
364 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
365
3665.21 Wed Dec 16 07:19:51 CET 2009
367 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
368 together.
369 - add some examples on how to combine other event loops with Coro in
370 Coro::AnyEvent, and how to run it (and not to block). Seems to be
371 the most common source of confusion.
372 - try to catch people naively blocking in an event callback.
373 - work around the perl filehandle bug issue in conjunction with
374 older common::sense (as indirectly pointed out by ZSystem).
375 - clarify the "not from signal handlers" section.
376
3775.2 Sun Oct 4 14:54:24 CEST 2009
378 - Coro::Storable destroyed the prototypes of the functions it wrapped.
379 - export rouse_cb and rouse_wait by default now.
380 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
381 - new method $state->swap_sv.
382 - added section on "windows process emulation" to the manpage,
383 after a not-so-fruitful (nor-friendly) "discussion" with chip
384 salzenberg (discussion implies arguments, but his only arguments
385 were ad-hominems, one wonders why he started it in the first
386 place). I hope this explains it well enough for him to understand,
387 and maybe well enough for others to understand.
388 - use common::sense everywhere now.
389 - idle callbacks are no longer supported, use idle coros instead.
390 - print a thread listing when a deadlock is detected.
391
3925.17 Sat Aug 22 23:09:31 CEST 2009
393 - work around a bug in the perl debugger causing crashes
394 when running under the debugger by marking _pool_handler as nodebug.
395 - speed up Coro::async considerably.
396 - try some hacks to get netbsd to work "more often" - their broken
397 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
398 - convert Coro to AE 5.0 API.
399
4005.162 Tue Jul 28 04:04:03 CEST 2009
401 - perl 5.8.2 is now minimum requirement.
402 - skip t/19_handle.t on broken windows perls.
403
4045.161 Wed Jul 22 04:47:38 CEST 2009
405 - Coro::AnyEvent::poll could have a different prototype when EV was
406 used as backend (analyzed by Tatsuhiko Miyagawa).
407 - Coro::AnyEvent errornously initialised the event loop when loaded,
408 not on demand.
409 - try to workaround rare */t/01_unblock.t failures.
410
4115.16 Tue Jul 21 01:44:37 CEST 2009
412 - Coro::AnyEvent failed to hook into the event loop
413 when no threads had been readied between detecting
414 the event loop and actually running it.
415 - considerably speed up Coro::Select by taking avdantage
416 of AnyEvent > 4.8 and some other optimisations.
417 - implement paragraph readline mode in Coro::Handle
418 (based on patches by Zsbán Ambrus).
419 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
420 and analyzed by Yasuhiro MATSUMOTO).
421 - clarified libcoro license and copyright.
422 - someone stole my EXTRA_META!!!
423 - implement Coro::Select::patch_pp_sselect and it's brother,
424 for hardcode select overriding.
425
4265.151 Mon Jul 6 05:41:57 CEST 2009
427 - backport to windows process emulation code again (patch by
428 Yasuhiro MATSUMOTO).
429 - slightly update Coro::MakeMaker.
430
4315.15 Tue Jun 30 10:28:06 CEST 2009
432 - deprecate Coro::Socket, document how to get ipv6 support via
433 AnyEvent::Socket instead.
434 - implement signal->wait ($cb) interface, similar to semaphores.
435 - work around SvOK not supporting getmagic, so we have to getmagic
436 to test for undef :/ (reported by Matthias Waldorf).
437 - load Coro::AnyEvent in all modules using AnyEvent.
438 - work around perl corrupting our internal data structures,
439 reported by Tokuhiro Matsuno.
440 - enable per-coroutine real and cpu time gathering
441 (Coro::State::enable_times).
442
4435.14 Wed Jun 24 01:37:48 CEST 2009
444 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
445 as this seems to have been a difficult concept (poll, sleep, idle,
446 idle_upto).
447 - add Coro::AnyEvent::readable/writable functions.
448 - clarify Coro::EV/Event/AnyEvent manpages.
449 - free per-thread global scalars in the thread calling ->cancel, to
450 avoid crashes when $_, $@ etc., are magical but some of those
451 had already been freed.
452 - "unexperimentalise" the callback interface for Coro::Semaphore.
453 - speed up ready queue management/context switching by using a linked
454 list instead of an array (~5-10%).
455 - implement "watch" command in Coro::Debug shells.
456 - for fun, implement time-slicing as an example in the manpage.
457 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
458 Coro::Event, use the normal AnyEvent handling instead of dieing
459 (the same is true for Coro::Handle).
460 - properly document Coro::EV::timed_io_once.
461 - avoid unneccessary ->cancel calls in Coro::Handle.
462 - maybe make it work on mingw32 with win32 backend
463 (based on patch by Yasuhiro Matsumoto).
464
4655.132 Fri May 29 09:00:39 CEST 2009
466 - do not keep a reference to the argument itself in
467 Coro::Semaphore::guard, as it could change later.
468 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
469 should support a prepare callback.
470
4715.131 Mon Mar 16 23:20:37 CET 2009
472 - implement and document Coro->suspend, Coro->resume.
473 - fix Coro::Select implementation to not (often) close
474 the passed file descriptors (testcase provided by pippijn).
475
4765.13 Mon Dec 15 21:51:42 CET 2008
477 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
478 - don't set diehook to C<undef> but instead to NULL, to avoid
479 spurious warnings.
480 - fix a lot of bugs in Coro::SemaphoreSet.
481 - Coro::SemaphoreSet will less often create a semaphore needlessly.
482 - add Coro::SemaphoreSet::count and wait methods.
483 - take advantage of the new Guard module.
484 - deprecate Coro::guard.
485 - try to fix the dreaded 01_unblock tests once more. I hate it when
486 testsuites need more fixing than the code they are supposed to test.
487 - croak in more cases when a required callback isn't resolvable.
488 - fix some minor issues in Coro::State->call/eval.
489 - use current coroutine context instead of a temporary one
490 when temporarily switching to another coroutine.
491 - do not call C-level on_destroy handlers during global destruction,
492 to avoid needless segfaults.
493
4945.12 Sun Dec 7 13:30:38 CET 2008
495 - add default config for MirOS, which seems to be a bug-to-bug
496 compatible fork of openbsd ("world domination by releasing
497 openbsd cvs before the openbsd folks do it" or so :).
498 - free_padlist did destroy the names pad, not good, but didn't
499 seem to bother perl - this could fix issues such as eval ""
500 inside a function called from multiple coroutines.
501 - use a different method to detect destruction time.
502 - be more careful when freeing padlists just before global
503 destruction.
504 - fixed and expanded the call/cc example.
505 - renamed _terminate to _coro_run.
506 - new method Coro::Channel->shutdown.
507 - try pthreads on openbsd <4.4 (broken sigaltstack, will
508 pthreads fare better?).
509 - be less picky about destroying "a" running coroutine.
510
5115.11 Tue Nov 25 21:49:05 CET 2008
512 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
513 working code with failed assertions, introducing more bugs than
514 it fixes, requiring elaborate workarounds :(
515
5165.1 Mon Nov 24 08:54:59 CET 2008
517 - wrote a small introductory tutorial - Coro::Intro.
518 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
519 - Coro::Select did errornously dup the file descriptors
520 and didn't work with all AnyEvent backends.
521 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
522 LWP data transfers.
523 - disassociate c contexts from coro objects - this is agruably more
524 correct, but mostly allows sharing of cctxs between coro and state
525 objects, for added memory savings and speed increases.
526 - bumped $Coro::POOL_RSS up to 32kb by default.
527 - no longer set the optype to OP_CUSTOM, as B::* understandably
528 doesn't like this very much (and we *are* a type of entersub).
529 - implement state cloning, just to prove that call/cc can be done.
530 - automatically load Coro::AnyEvent in Coro::Handle.
531 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
532 do not have this method (and don't need it either).
533 - speed up generic anyevent methods in Coro::Handle by using rouse
534 callbacks.
535 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
536 others. It also makes the debugger happier, as you can trace
537 through the idle threads now.
538 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
539 - eg/event was pretty broken.
540 - better 5.8.6 compatibility.
541
5425.0 Thu Nov 20 10:35:05 CET 2008
543 - NEW ARCHITECTURE: use the latest 4.x version if you experience
544 stability issues.
545 - bump API version to 7 - all dependents must be recompiled.
546 - removed timed_* functions - they were not being used anyways
547 and should be replaced by a more generic mechanism -
548 and were annoying to support anyways :)
549 - removed SemaphoreSet's waiter method - use sem method instead.
550 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
551 - async_pool did free a scalar value twice
552 ("Attempt to unreference...").
553 - fix a longstanding bug where calling terminate on a coro that
554 was waiting for a semaphore that was just becoming available
555 would cause a deadlock (semaphore would get into a state where
556 it was available but waiters were still blocked).
557 - calling throw on a coroutine that is waiting for a semaphore will
558 no longer make it acquire the semaphore (and thus leak a count).
559 - perl's process emulation is now not even theoretically supported
560 anymore.
561 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
562 conversion of callback-style to blocking-style.
563 - new methods $coro->schedule_to and ->cede_to, to specifically
564 schedule or cede to a specific coroutine.
565 - new function Coro::Semaphore::wait.
566 - use named constants in Coro::Channel (Richard Hundt).
567 - directly patch the entersub opcode calling SLF functions (cede,
568 transfer and so on). this does speed up context switching, but
569 more importanly, it frees us from the hardcoded behaviour of
570 entersub, so we might actually be able to return something from
571 those functions and atcually create new ones.
572 - take advantage of __builtin_frame_address on gcc.
573 - expose THX in coroapi (not sure whether this was a wise decision,
574 as "threaded" perls are running at half speed anyways).
575 - implement execute_slf (schedule-like-function) interface that makes
576 it possible to implement schedule-like-functions in XS.
577 - use new SLF interface to massively speed up Coro::EV by roughly a
578 factor of two.
579 - used new SLF interface to massively speed up Coro::Semaphore by a
580 factor of three.
581 - used new SLF interface to speed up Coro::AIO by roughly a factor of
582 four and reduce its memory usage considerably.
583 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
584 for a nice speedup and vastly more correct behaviour. Also implement
585 a new method "sem" to get at the underlying semaphore object.
586 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
587 (in comparison) 20-40% speedup.
588 - used new SLF interface to reimplement Coro::Signal gaining
589 some unknown (because I was too lazy), but certain, speedup, and also
590 making signals reliable for the first time.
591 - used new SLF interface and other optimisations to speed up async_pool
592 by a factor of two. It also doesn't rely on perl's exception mechanism
593 to exit anymore. The overhead for terminating an async_pool, coro over
594 a normal async is now very small.
595 - sped up coroutine creation/destruction by 40%.
596 - forgot to include Coro/libcoro/README in the dist for all these years.
597 - work around a freebsd pthreads bug (manual testcancel is required as
598 pthread_cond_wait isn't a cancellation point on freebsd).
599 - use new rouse functions to speed up and simplify Coro::BDB.
600 - make "prefer perl native functions" work with threaded perls.
601 - condense Coro::Debug ps output, hint at v and w flags.
602 - (libcoro) lots of minor cleanups and portability improvements.
603
6044.914 Wed Nov 19 12:54:18 CET 2008
605 - fix a disastrous bug in the readline optimisation
606 introduced in 4.801.
607
6084.913 Sat Nov 15 07:58:28 CET 2008
609 - async_pool did free a scalar value twice
610 ("Attempt to unreference...").
611
6124.912 Thu Nov 13 18:31:23 CET 2008
613 - minor cleanups.
614 - use much larger stacks on linux and perl < 5.8.8.
615 - Coro::Debug::new_unix_server did not unlink the socket
616 when destroyed.
617
6184.911 Tue Nov 11 04:26:17 CET 2008
619 - "port" to threaded perls.
620
6214.91 Mon Nov 10 05:36:38 CET 2008
622 - the ->throw exception object no longer leaks.
623 - creating a new cctx leaked a scopestack entry (memleak).
624 - new coroutines didn't get created with a zero flags field
625 (unknown impact).
626 - calling ->throw on a not-yet-started coroutine should now work
627 instead of being ignored.
628 - ->throw is now supported on Coro::State objects.
629 - clean up cctx creation code a bit.
630 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
631
6324.9 Sat Nov 8 17:45:27 CET 2008
633 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
634 - (libcoro) no longer leak threads in the experimental pthread backend,
635 also speed it up considerably.
636 - (libcoro) do not rely on makecontext passing void *'s unscathed.
637 - fix compiletime dependencies on libcoro in the Makefile.
638 - cctx_count wasn't always updated properly.
639 - Coro::State::cctx_stacksize wasn't applied correctly.
640 - new function Coro::State::cctx_max_idle.
641 - the default max number of idle C contexts is now 4.
642 - (libcoro) try harder to get _setjmp/_longjmp.
643 - (libcoro) cleanup and extend the libcoro API to officially
644 allow the creation of empty source contexts.
645 - very experimental workaround for the totally broken netbsd platform.
646 - tried to hack around openbsd bugs.
647
6484.804 Wed Nov 5 16:36:00 CET 2008
649 - Coro::Debug::new_unix_server would not create a non-blocking listening
650 socket, sometimes causing freezes.
651 - (libcoro) fix misaligned stack points for setjmp and assembly
652 methods, which can cause crashes on x86/x86_64 with a sufficiently
653 aggressive compiler.
654 - new function: Coro::Debug::new_tcp_server.
655 - move ->throw into the Coro class because it only works on coro objects.
656
6574.803 Mon Nov 3 17:16:12 CET 2008
658 - (libcoro) use a global asm statement to become independent of gcc
659 otpimisations for CORO_ASM (thanks to pippijn for the idea).
660 - try to workaround yet another broken bsd, this time dragonfly.
661
6624.802 Thu Oct 30 10:56:12 CET 2008
663 - support -fno-omit-frame-pointer on x86 with the assembly method.
664 - tune 01_unblock.t tests a bit.
665
6664.801 Wed Oct 22 18:33:37 CEST 2008
667 - improve readline speed for very long "lines".
668 - backport to 5.8.8.
669
6704.8 Thu Oct 2 13:34:40 CEST 2008
671 - new function Coro::AIO::aio_wait.
672 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
673 - greatly speed up and reduce memory usage of Coro::AIO requests.
674 - implement some other µ-optimisations.
675
6764.749 Mon Sep 29 14:40:12 CEST 2008
677 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
678
6794.748 Sat Sep 27 14:03:19 CEST 2008
680 - implement, but do not document, PerlIO::cede(granularity).
681 - Coro::Storable forgot to wrap Storable::pstore.
682 - work around the multitude of leaks and memory corruption
683 bugs in PerlIO::via by using our own C-level perliol. As a side
684 effect, Coro::Storable is now much, much, much faster.
685
6864.747 Tue Sep 23 01:59:41 CEST 2008
687 - fix a per-cv memleak (one empty array was leaked per
688 code reference).
689 - avoid a crash in coro->call|rss when the coroutine was already
690 destroyed (most noticably when using Coro::Debug::ps :)
691 - also protect *Storable::FILE.
692 - push up default storable granularity to 20ms.
693
6944.746 Sun Sep 21 03:22:20 CEST 2008
695 - be more insistent on locking Storable against reentrancy
696 in Coro::Storable.
697 - move swap_def?v and throw to Coro::State, as documented.
698
6994.745 Thu Jul 24 00:14:38 CEST 2008
700 - remove debugging code related to MgPV_nolen_const, also try to
701 make it compile with perl 5.8.6 (yes, apple apparently loves
702 outdated software). Reported by John S.
703
7044.744 Tue Jul 8 22:06:35 CEST 2008
705 - correctly provide default DIE/WARN handlers as documented.
706 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
707 implementations inside perl work fine for us.
708
7094.743 Mon Jun 16 00:21:57 CEST 2008
710 - when using Coro::EV without running EV::loop it could
711 result in busy-waiting for events, this has been fixed.
712 - reduce codesize and improve performance by using EV_DEFAULT_UC.
713
7144.742 Sat May 31 14:10:21 CEST 2008
715 - implement a workaround for (some) perl <5.8.8 versions.
716 - require EV 3.3+.
717
7184.741 Fri May 30 23:33:09 CEST 2008
719 - tell netbsd how utterly broken their imitation of an OS is
720 and refuse to build by default if pthreads are in use.
721 - switch to "s" method on all bsd's by default, as their ucontext
722 stuff seems just too broken.
723 - fix a bug in Coro::Select.
724
7254.74 Thu May 29 20:05:31 CEST 2008
726 - do not test Coro::LWP for lack of dependencies.
727 - do not test Coro::Select for lack of working perls.
728
7294.73 Thu May 29 2008
730 - fix a bug in Coro::EV which would cause it to block despite
731 there being runnable coroutines.
732 - sprinkle "no warnings" freely over everything, also suppress
733 warnings for some other modules.
734 - fix typo in WSAEWOULDBLOCK.
735
7364.72 Sun May 25 05:14:36 CEST 2008
737 - tweak META.yaml a bit, unfortunately, there is no documented way
738 to have optional dependencies with CPAN. doh :(
739 - avoid running some tests on windows because they would fail due to
740 perl bug (broken fork, broken pipes...).
741 - work around perl on windows bugs where perl returns undocumented
742 error codes for sysread, syswrite etc. by taking advantage
743 of AnyEvent's workaround for that problem.
744 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
745 a common perl implementation bug on windows.
746 - use unix domain sockets in testsuite to work around
747 common perl implementation bugs on widows (they are emulated by
748 tcp sockets on windows. ugh.)
749
7504.71 Sat May 24 20:01:27 CEST 2008
751 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
752 - take advantage of async name resolution of AnyEvent::Util.
753 - work around brutal inet_aton override in Coro::LWP.
754 - take advantage of the readyhook in Coro::EV, for smoother
755 scheduling.
756
7574.7 Sun May 11 00:32:19 CEST 2008
758 - completely reworked the Coro manpage.
759 - added Coro::AnyEvent, generic event loop integration.
760 - implement cancel, ready and kill commands in Coro::Debug.
761 - document find_coro in Coro::Debug.
762 - incompatible change: rename has_stack to has_cctx.
763 - Coro::AIO and Coro::BDB no longer force event model detection,
764 use AnyEvent::AIO and AnyEvent::BDB.
765
7664.6 Sat Apr 26 10:05:14 CEST 2008
767 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
768 by the Coro module. It was a mistake to have it in the first place.
769 - (experimental) support for activestate perl 5.10 (method "w").
770 - (experimental) support for strawberry perl (method "a").
771 - coro_sigelem_set did not return a value although it had to,
772 actual impact unknown.
773
7744.51 Mon Apr 14 13:28:27 CEST 2008
775 - make it compile again on 5.8.
776
7774.50 Thu Apr 10 09:43:17 CEST 2008
778 - I did it twice! (see 4.49).
779
7804.49 Sun Apr 6 21:23:31 CEST 2008
781 - grr, instead of compiling the recent changes on 5.10 only they
782 were compiled on 5.8 only.
783
7844.48 Sun Apr 6 20:36:46 CEST 2008
785 - allow coroutine switches during eval's under 5.10.x, as it seems
786 the parser is a per-interpreter option now, so this is safe
787 (this might fix the odd crashing bug).
788 - drop support for 5.9.x versions: they are dead, jim.
789
7904.47 Sun Apr 6 00:37:52 CEST 2008
791 - force cctx allocation on API calls: we know we need to force one
792 and gcc actually manages to confuse our heuristic nowadays,
793 leading to crashes and worse.
794 - document force_cctx.
795
7964.46 Fri Apr 4 22:05:43 CEST 2008
797 - upgrade libcoro, resulting in pthread-backend (which was only created
798 to fulfill the rules of the programming languages shootout).
799
8004.45 Thu Mar 13 11:55:36 CET 2008
801 - fix a file leaking bug in eg/mhyttpd that would allow
802 downloading of any file (reported by oesi).
803 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
804 (also add testcase).
805 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
806 apparently having loads of fun with that).
807 - implement $state->swap_defsv, ->swap_defav and document ->throw.
808
8094.4 Wed Feb 13 16:44:29 CET 2008
810 - only assume .cfi pseudo isns exist on GNU/Linux.
811 - add get_timed method to Coro::Channels.
812 - fixed Coro::Debug manpage.
813 - perl 5.11 compatbility improvement based on patch by
814 Andreas König.
815
8164.37 Sun Jan 20 11:25:23 CET 2008
817 - soften the check for an active parser for perl 5.10 (avoiding
818 "Coro::State::transfer called while parsing" in many cases).
819
8204.36 Sun Jan 13 10:53:56 CET 2008
821 - reset diehook when terminating from an async_pool as to not
822 trigger any __DIE__ handlers.
823
8244.35 Sun Jan 13 04:14:13 CET 2008
825 - "bt" debug command now displays any exceptions
826 from longmess and also skips the innermost
827 stackframes, giving more useufl output.
828 - allow backtraces in coroutines blocked in Coro::EV,
829 at a <1% speed hit.
830 - handle localising of $SIG{__WARN__} and __DIE__
831 properly (with a proper amount of dirty hacking).
832
8334.34 Sat Dec 22 17:49:53 CET 2007
834 - upgrade to EV version 2.0 API.
835
8364.33 Mon Dec 17 08:36:12 CET 2007
837 - make Coro::AIO etc. loadable in the absence of EV.
838
8394.32 Mon Dec 17 07:46:02 CET 2007
840 - majorly improved Coro::Handle's performance with Coro::EV.
841 - implemented the until now mythical Coro::BDB module.
842 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
843
8444.31 Wed Dec 5 12:32:39 CET 2007
845 - remove warn statement form Coro::Util.
846
8474.3 Tue Dec 4 20:33:14 CET 2007
848 - calls to the idle function could cause stack corruption
849 when the stack changed.
850 - do no longer rely on the presence of EV::DNS if EV is used
851 (because it is gone), but instead take avdantage of EV::ADNS
852 if available.
853 - add ($) prototypes to all functions in Coro::Storable.
854 - use a conventional (safer) idle callback in Coro::EV.
855 - do accept quickly in Coro::Debug to avoid endless loops.
856
8574.22 Fri Nov 30 16:04:04 CET 2007
858 - really use optimised versions for Event and EV in Coro::Util
859 and Coro::Handle.
860
8614.21 Sun Nov 25 10:48:59 CET 2007
862 - fix a spurious memory read.
863 - Coro::EV no longer keeps the eventloop "alive".
864
8654.2 Fri Nov 9 20:47:05 CET 2007
866 - enable/disable tracing from a new coroutine, not a pooled one.
867 - fix a memleak in Coro::Event.
868 - removed killall call from fork_eval.
869 - made sure store_fd is already loaded so that fork_eval
870 does not have to parse autoload in each subprocess.
871 - only use assembly method if -O switch is in $Config{optimize}.
872 - add (optional) Coro::EV module, so far the best event loop module
873 directly supported by Coro.
874 - if the event model is EV, use EV::DNS to resolve
875 stuff in Coro::Util.
876 - don't get confused by multiple event notifications in Coro::Handle.
877 - initial support for EV (libevent interface).
878 - require Event and EV using configure_requires, to force their existance.
879
8804.13 Wed Oct 24 07:26:45 CEST 2007
881 - add Coro::Storable::blocking_thaw.
882 - use a vastly more complicated technique to localise
883 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
884 - use a coroutine for the idle callback Coro::Event,
885 instead of running Event in the current coroutine context.
886 This also catches recursive invocations.
887 - actually report fork errors in gethostbyname and inet_aton.
888
8894.11 Thu Oct 11 02:40:24 CEST 2007
890 - port to threaded perls.
891
8924.1 Thu Oct 11 02:38:16 CEST 2007
893 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
894 be local to each coro (see Coro::State).
895 - incompatible change: for very deep reasons, cede and cede_notself
896 cannot return anything, so nothing will be returned.
897 - possibly bring back 5.10 compatibility (untested).
898 - work around stupid (and wrong) warning on 5.10 :(.
899 - overlay the saved state over the context stack. This saves
900 a few hundred bytes per coroutine on average and also
901 speeds up context switching a bit.
902 - further tune default stack sizes.
903 - (more) correctly calculate stack usage in coro_rss.
904 - Coro::Storable::blocking_* did not properly lock
905 resulting in races between coroutines.
906 - added Coro::Storable::guard.
907 - stopping to trace a coroutine could destroy the cctx of
908 an unrelated coroutine.
909 - explain the relationship between Perl and C coroutines in
910 more detail in Coro::State.
911 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
912 causing a fork per resolve. This also affected Coro::Socket.
913 - switch to a separate stack in $coro->call/eval to avoid
914 invalidating pointers.
915
9164.03 Sat Oct 6 21:24:00 CEST 2007
917 - added Coro::throw method.
918 - minor code cleanups.
919
9204.02 Sat Oct 6 02:36:47 CEST 2007
921 - fix a very minor per-coroutine memleak (a single codereference).
922 - fixed a bug where the currently in-use c context would be freed
923 prematurely (can happen only when programs change the stacksize
924 or use tracing).
925 - tracing can no longer keep a coro alive after it terminated.
926 - do static branch prediction in the common path for gcc. gives
927 about 2-5% speed improvement here.
928
9294.01 Fri Oct 5 22:10:49 CEST 2007
930 - instead of recreating *a* standard output handle we simply
931 use STDOUT, which is faster and hopefully more robust.
932
9334.0 Fri Oct 5 12:56:00 CEST 2007
934 - incompatibly changed Coro::Storable::freeze.
935 - major new feature: added Coro::Debug, for interactive coroutine
936 debugging, tracing and much more.
937 - major bug fix: unbelievable, but true: $_, $/ and many other
938 "saved" variables actually weren't being saved. This has been fixed,
939 of course, while increasing performance while losing all the save
940 flags.
941 - save flags are gone, and all the api functions dealing with them.
942 - added Coro::Semaphore::adjust.
943 - added Coro::Util::fork_eval.
944 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
945 - added Coro::killall.
946 - reduce initial stack sizes to allow for "micro-coroutines".
947 - better async_pool resource management, moved parts of async_pool
948 handling to XS (major speed improvement).
949 - actually croak before modifying important data structures.
950 - refuse to transfer while compiling.
951 - possibly support eval EXPR better now.
952 - enable assembly per default on linux+bsd x86+amd64.
953 - all internal members were renamed _something for easier subclassing.
954 - many minor tweaks.
955
9563.63 Wed May 16 14:10:06 CEST 2007
957 - implement handcoded assembly for x86/amd64 SVR ABI.
958
9593.62 Fri Apr 27 21:36:06 CEST 2007
960 - upgrade libcoro (which might set unwind info correctly).
961 - change default on linux to setjmp/longjmp.
962
9633.61 Thu Apr 19 12:36:18 CEST 2007
964 - Coro::Storable caused an endless loop when thawing invalid
965 pst's sometimes.
966 - use a Semaphore in Coro::Storable, as Storable doesn't
967 seem to be reentrant (although it is documented to
968 be threadsafe...).
969 - fix Coro::Signal to bring back the original unreliable
970 but stateful semantics.
971 - fixed a lot of typos in Coro.pm (patch submitted by David
972 Steinbrunner, which applied flawlessly).
973
9743.6 Sat Apr 14 17:13:31 CEST 2007
7 - added some bugfixes to get eg/myhttpd working again. 975 - added some bugfixes to get eg/myhttpd working again.
8 - added Coro::Storable for often-cede'ing freeze/thaw. 976 - added Coro::Storable for often-cede'ing freeze/thaw.
9 - try to do a clean exit when a coroutine calls exit 977 - try to do a clean exit when a coroutine calls exit
10 (EXPERIMENTAL). 978 (EXPERIMENTAL).
11 - got rid of indirect call through _coro_init. 979 - got rid of indirect call through _coro_init.
12 - updated the partly antique examples in eg/ to 980 - updated the partly antique examples in eg/ to

Diff Legend

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