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

Comparing Coro/Changes (file contents):
Revision 1.341 by root, Fri May 9 17:29:19 2008 UTC vs.
Revision 1.684 by root, Sat Feb 29 21:40:22 2020 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: channel->maxsize(newsize)?
5 actor model (which is cool in a lot of important aspects (failures!), 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 but very lacking in others (higher level ipc)). 6TODO: swap_sv, maybe add scope_swap_sv?
7TODO: croak when async_pool tries to run canceled thread?
8TODO: how to debug safe_cancel failure (vebrose mode? show how to make backtrace?)
9TODO: scalar context for Coro::AIO wrappers?
7 10
8 TODO: Coro::AnyEvent 116.56 Thu Feb 20 19:40:07 CET 2020
12 - Coro::EV::time*once functions used ev_once, which is not cancelable.
13 This could cause problems due to I/O watchers not being stopped up
14 on exceptions/cancellation.
15
166.55 Fri Jul 19 17:28:09 CEST 2019
17 - Coro::RWLock woke up only one reader instead of all after a wrlock
18 (analyzed and patch by Maxim Skorobogatov).
19 - slightly improved documentation of Coro::RWLock and RWLock testcase.
20 - quickly (re)-ported to minix 3.3 before minix crashed again.
21
226.54 Sun Dec 16 11:06:07 CET 2018
23 - reinstate Coro::Select::select prototype - it breaks more than it fixes.
24
256.53 Sun Dec 16 10:33:15 CET 2018
26 - the autogenerated prototypes of Coro::AIO now add ";" at the end
27 to avoid generating named unary operators and messing up parsing
28 (mostly affects aio_mlockall 4 | 8 which was parsed as
29 (aio_mlockall 4) | 8).
30 - do not use a prototype for select in Coro::Select, as CORE::select also
31 doesn't have one.
32
336.52 Tue Aug 14 18:49:43 CEST 2018
34 - libcoro update: allow sharing of coro threads using the pthreads
35 backend among different pthreads, as is required by Coro::Multicore.
36 - always allow ->call and ->eval when the coro is the current coroutine. this
37 makes it possible, among other things, to call these on $Coro::main.
38 - minor code cleanups.
39
406.514 Thu Aug 31 18:25:31 CEST 2017
41 - safe_cancel no longer croaks when called on already-destroyed
42 thread, but instead "just works".
43
446.513 Sat Jul 15 01:18:51 CEST 2017
45 - experimental 5.22 comaptibility changes (PadlistNAMES).
46 - removed attributes example and made sure the other examples
47 still work.
48 - fix some statevar types for newer perls, should be of no consequence.
49
506.512 Fri Jul 14 05:03:17 CEST 2017
51 - seems to pass test suite down till 5.8.9, so make this the
52 minimum requirement in Makefile.PL.
53 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
54 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
55 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
56 fortified longjmp variants in some libcs.
57 - re-try JIT allocation without PROT_EXEC, to hopefully improve
58 portability (still doesn't seem to work around broken selinux
59 mmap everywhere).
60 - speling fixes (Xavier Guimard).
61
626.511 Sun Jun 26 23:44:50 CEST 2016
63 - make it compile with threaded perls again.
64 - simplify and speed up __DIE__ and __WARN__ handling.
65
666.51 Sat Jun 25 20:57:20 CEST 2016
67 - 6.5 release didn't compile on most older perls due to
68 a logic error enabling experimental 5.24 code for all versions.
69 - since various broken or deficient patches float around
70 to work around the vtbl API breakage in perl, let's provide
71 our own workaround, which seems to be much less invasive
72 and more compatible than the approaches seen so far. YMMV.
73 - work around assertion failure in perl_destruct on 5.24
74 (analyzed by Dave Mitchell).
75
766.5 Wed Jun 22 22:23:50 CEST 2016
77 - swap_sv swapping order was not symmetrical, causing
78 wrong swaps when swap_sv was used multiple times on the
79 same sv in the same thread.
80 - swap_sv calls can now be undone by calling it again
81 with the same variables.
82 - swap_sv calls will now be undone in async_pool threads.
83 - split Coro::Semaphore::up/adjust into separate xs functions
84 for better error reporting, at slight codesize increase.
85 - (libcoro) arm assembly support, please test and report.
86 - adjust to PL_savestack changes in perl 5.24 (adapted from
87 the debian patch, which unfortunately gets it wrong).
88
896.49 Sat Oct 17 01:40:12 CEST 2015
90 - throwing an exception to a thread waiting in
91 Coro::Handle using Coro::EV did not stop the watchers,
92 causing the next call to fail (testcase by Martin Pritchard).
93 - bump minimum perl version to 5.10.
94
956.48 Sun Oct 4 19:03:51 CEST 2015
96 - fix memory leak when Coro::Handle uses Coro::EV internally
97 (testcase by Sten Sten).
98 - update code to libev 4 API (internally, Coro still used the
99 version 3 API calls).
100 - remove unused hv_sig.
101
1026.47 Sat Jul 11 03:58:20 CEST 2015
103 - swap_svs - NVs can be in the SV head beginning with 5.20.
104
1056.46 Tue Jun 30 14:36:30 CEST 2015
106 - restore portability to perls with windows fork emulation
107 (patch by Petr Písař).
108
1096.45 Tue Jun 30 01:40:08 CEST 2015
110 - restore portability to perl 5.16 and below.
111
1126.44 Tue Jun 30 00:41:54 CEST 2015
113 - avoid segfaulting (or worse) when tracing a canceled thread.
114 - the Event module silently broke it's public hook API, causing
115 failures on perls compiled with -Duselongdouble. Requiring
116 the latest version of Event because I don't know which release
117 changed this (it's not mentioned in the ChangeLog).
118 - allow xs level enterleave hooks via CoroAPI.
119
1206.43 Thu Jun 4 15:38:14 CEST 2015
121 - use stability canary.
122 - port to stableperl-5.22.0-1.001.
123 - update libecb.
124
1256.42 Wed Feb 11 20:29:52 CET 2015
126 - Coro::SemaphoreSet->try did not actually work (analyzed by
127 SATO Kentaro).
128 - upgrade libecb to be C11 compliant.
129
1306.41 Sat Sep 6 22:08:46 CEST 2014
131 - restore portability to perl 5.8.x.
132 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
133 up some memory, but fixes "use feature" and similar modules
134 when used inside a Coro.
135 - allow zero as argument to Coro::Channel to mean the same thing
136 as no argument. this works with older versions as well,
137 but wasn't legal until now.
138 - slightly better c header file detection.
139
1406.39 Mon Jun 2 00:00:08 CEST 2014
141 - work around more incompatible changes in 5.20.
142
1436.38 Sun Jun 1 21:54:23 CEST 2014
144 - check that perl slots actually have enough space to hold
145 interpreter variables.
146 - untested port to perl 5.19 (64 bit tmps indices) (reported
147 by Andreas König).
148 - croak when cancel is called without a thread context.
149
1506.37 Tue Mar 4 13:27:33 CET 2014
151 - *sigh*, removed leftover debugging code from debugging a
152 perl bug, of all things.
153
1546.36 Tue Mar 4 07:11:59 CET 2014
155 - semaphores would not clear the destroy hook when interrupted
156 by ->throw, causing segfaults or worse.
157 - ->throw on a thread waiting for a semaphore did not acquire
158 the semaphore, but also didn't wake up other waiters,
159 possibly causing a deadlock.
160 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
161 to report a full stacktrace, which should help find out
162 where the actual call is.
163 - "atomically" destroy data in slf_destroy, because it is
164 the right thing to do, just in case.
165 - disable ecb.h libm dependency, to compile on ancient systems
166 or under adverse conditions.
167
1686.33 Mon Nov 18 11:26:27 CET 2013
169 - do not crash when freeing padlists with holes (in 5.18).
170 - tentative SVt_BIND 5.19 port/fix.
171
1726.32 Tue Nov 5 15:35:35 CET 2013
173 - use a new algorithm to derive padlists for perl 5.18. The old
174 one could lead to 0-pointer accesses inside perl (reported
175 by Darin McBride).
176
1776.31 Thu May 9 07:39:48 CEST 2013
178 - Coro::AIO requests would crash if the thread was ready'd
179 while the request was ongoing.
180
1816.29 Wed May 8 02:55:18 CEST 2013
182 - when an on_destroy handler destructs the coro currently being
183 destructed a perl scalar could be accessed after being freed,
184 likely causing a crash.
185
1866.28 Wed Mar 6 06:58:02 CET 2013
187 - clean remnants of existing __DIE__ and __WARN__ handlers so
188 they lose their magic and will not cause segfaults later
189 (testcase by Andrey Sagulin).
190 - improved Coro::State documentation a bit.
191 - Coro::Debug::command now flushes the output.
192 - add hack detection code for x32 abi, because the braindead slugs
193 who designed that made it look exactly like x86_64 without
194 providing proper compile time symbols to test for it. as a result,
195 this detection cannot work reliably.
196 - valgrind stack registering was broken.
197 - do not rely on Time::HiRes anymore in Coro::Debug.
198
1996.23 Fri Dec 7 23:36:37 CET 2012
200 - use experimental fiber implementation on native windows
201 perls.
202 - use sizeof (void *) as multiplication factor for stack sizes,
203 to accomodate the totally braindamaged microsoft 64 bit "os".
204 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
205 activeperl 5.16 crashes when PerlIO_define_layer is called due
206 to some bug in the perl dll, strawberry perl at least passes
207 the testsuite.
208 - implement Coro::Handle->peeraddr/host/port, for slightly
209 improved compatibility with LWP.
210 - implement 5.17 compatibility by almost blindly applying a
211 good-looking patch by Father Chrysostomos.
212 - move stack management functions into libcoro 3.
213 - libcoro version 3 "released".
214 - support magic values as timed_io_once args.
215 - recommend AnyEvent 7+ or EV 4+, also require EV
216 version 4 or newer for Coro::EV.
217
2186.10 Tue Oct 9 01:14:27 CEST 2012
219 - updated ecb.h, it had a typo that caused it to not compile on many
220 big endian systems (reported by many people).
221 - disable memory fences in ecb.h to improve portability.
222
2236.09 Sat Oct 6 23:25:02 CEST 2012
224 - Coro::EV I/O watchers were not interruptible by exceptions
225 (Coro::State::throw) (testcase by sten).
226 - ->throw now puts threads into the ready queue, as this seems to
227 be expected by existing code, and code that doesn't cope with spurious
228 wakeups needs fixing anyway.
229 - use fd -1 in mmap.
230 - cast I32 to int in error message printf.
231 - warn about broken so-called "hardened" kernels.
232
2336.08 Fri Apr 13 12:05:47 CEST 2012
234 - be more aggressive about exiting like perl does - formerly,
235 exiting from the non-main thread would not execute END blocks.
236
2376.07 Fri Nov 11 21:21:48 CET 2011
238 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
239 - update ecb.h.
240
2416.06 Mon Aug 8 23:59:48 CEST 2011
242 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
243 backend, which is an order of magnitude slower. unfortunately, cygwins
244 pthread implementation isn't very complete either, so allocate the stack
245 twice just to be sure.
246 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
247 in its ithreaded perl - best recompile cygwin and use the 'w'indows
248 backend for much better performance. also disable ithreads for
249 even better performance...).
250
2516.05 Thu Aug 4 21:36:36 CEST 2011
252 - blush, condvar values would not be propagated from send to recv anymore
253 (reported by Chip Salzenberg).
254 - use exponential increase for the readline buffer length in
255 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
256
2576.04 Wed Aug 3 17:15:45 CEST 2011
258 - use even more efficient and more compatible condvars for
259 compatibility to AnyEvent 6.x :)
260 - more inconsequential ecb.h updates.
261
2626.03 Wed Aug 3 11:41:30 CEST 2011
263 - change how Coro patches AnyEvent condvars for compatibility to
264 AnyEvent 6.x.
265 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
266 hasn't been included.
267
2686.02 Wed Jul 13 04:35:19 CEST 2011
269 - "improve portability to Gentoo" - gentoo manages to put perl variables
270 in memory areas that are farther than 2gb apart, which the jit couldn't
271 handle and barfed. now it's just a bit slower on gentoo and similar
272 systems.
273
2746.01 Sun Jul 3 12:31:14 CEST 2011
275 - workarounds are good, but the test for whether pthreads are used
276 was not good. this one should be better.
277 - check differently whether gcc generates cfi instructions itself.
278
2796.0 Wed Jun 29 19:43:35 CEST 2011
280 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
281 destroyed and cleaned up automatically (e.g. async { schedule }).
282 - implement a JIT compiler for part of the thread switch code,
283 which gives a 50% speed improvement on threaded perls, and
284 about 4% on non-threaded perls (so threaded perls now finally
285 reach about half the speed of non-threaded perls).
286 - slightly modernise Coro::Intro, add section about rouse functions.
287 - avoid DEFSV and ERRSV, giving another 10% improvement
288 in thread switching.
289 - Coro::State->is_destroyed is now called is_zombie.
290 - implement a Coro->safe_cancel method that might fail, but
291 cancels in a "safer" way if it succeeds.
292 - add preliminary support for DEBUGGING perls.
293 - get rid of two hash-accesses when initialising a new Coro - this
294 speeds up coro creation by almost a factor of two.
295 - croak when a coro that is being cancelled tries to block
296 (e.g. while executing a guard block), instead of crashing or
297 deadlocking.
298 - use a more robust and also faster method to identify Coro::State
299 objects - speeds up everything a bit.
300 - implement Coro->cancel in XS for a 20% speed improvement, and to
301 be able to implement mutual cancellation.
302 - speed up context switches by a percent or two by more efficiently
303 allocating context stack entries.
304 - implement Coro->join and Coro->on_destroy in XS for a speedup and
305 a reduction in memory use.
306 - cancelling a coro while it itself is cancelling another coro is
307 now supported and working, instead of triggering an assertion.
308 - be a bit more crash-resistant when calling (buggy) on_destroy
309 callbacks (best effort).
310 - move on_destroy into the slf_frame, to allow extension slf
311 functions to have destructors.
312 - get rid if coro refcounting - simply crash in other interpreter
313 threads by nulling the pointers on clone.
314 - simplify warn/die hook handling when loading Coro - the convoluted
315 logic seems to be no longer neccessary.
316 - use libecb instead of our own home-grown gcc hacks.
317 - document alternatives to Coro::LWP. Please use them :)
318 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
319 sigaltstack. Really. wine suffers from it, erlang suffers from it,
320 and it's known since at least 2006.
321
3225.372 Wed Feb 23 06:14:30 CET 2011
323 - apparently mingw doesn't provide a working gettimeofday, try to
324 work around that by relying on Time::HiRes (indirectly brought to
325 my attention by Max Maischein).
326 - fix some portability issues when Time::HiRes was used.
327
3285.371 Mon Feb 21 14:36:08 CET 2011
329 - backport to windows process emulation code again.
330
3315.37 Sat Feb 19 07:49:44 CET 2011
332 - add a big "Coro thread life cycle" section to "man Coro".
333 - try a tentative workaround against the breakage that 5.13 has
334 introduced without depreciation period. sigh.
335 - no longer use Time::HiRes if gettimeofday is available, which
336 saves quite a lot of memory.
337
3385.36 Sun Feb 13 05:33:41 CET 2011
339 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
340 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
341 method call.
342 - undocument Coro::Timer::sleep and obsolete whole module.
343 - optimise Coro::Timer::timeout memory and cpu usage.
344 - slightly updated Coro::Intro for recent changes.
345 - do not initialise PL_dirty anymore.
346
3475.25 Thu Nov 11 01:08:39 CET 2010
348 - try a different approach on netbsd - netbsd 5 finally has marginally
349 working pthreads, but still broken ucontext/sigaltstack.
350 - openbsd 4.8 finally got their act together, Coro works out of the box
351 with asm, setjmp and pthreads (no change, just informational).
352
3535.24 Sat Oct 23 11:27:12 CEST 2010
354 - port to the EV 4.0 API.
355 - work around bugs in mingw32, making strawberry perl work
356 out of the box.
357 - correctly modify Coro::AIO function prototypes
358 so that they reflect the "no optional parameters" rule.
359 - "ported" libcoro to C++.
360
3615.23 Mon May 17 18:50:42 CEST 2010
362 - be more resistant to ordering changes when initialising
363 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
364 Waldorf).
365 - document that perl 5.12 deliberately removed support for cloning.
366
3675.22 Wed Apr 14 03:55:35 CEST 2010
368 - correctly return udnef on errors in Coro::Handle::read/write
369 (testcase by Marc Mims).
370 - convert Coro::Util into a "perl compatibility wrapper" - the functions
371 are less useful now, but are drop-in replacements for existing
372 functions, listing better alternatives in the documentation. This also
373 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
374 with Coro::Util::inet_aton.
375 - do not override $Coro::idle unconditionally in Coro.pm, as other
376 modules could have provided their own idle coro already
377 (for exmaple, Coro::AnyEvent).
378 - fix Coro::Util::gethost* functions.
379 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
380
3815.21 Wed Dec 16 07:19:51 CET 2009
382 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
383 together.
384 - add some examples on how to combine other event loops with Coro in
385 Coro::AnyEvent, and how to run it (and not to block). Seems to be
386 the most common source of confusion.
387 - try to catch people naively blocking in an event callback.
388 - work around the perl filehandle bug issue in conjunction with
389 older common::sense (as indirectly pointed out by ZSystem).
390 - clarify the "not from signal handlers" section.
391
3925.2 Sun Oct 4 14:54:24 CEST 2009
393 - Coro::Storable destroyed the prototypes of the functions it wrapped.
394 - export rouse_cb and rouse_wait by default now.
395 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
396 - new method $state->swap_sv.
397 - added section on "windows process emulation" to the manpage,
398 after a not-so-fruitful (nor-friendly) "discussion" with chip
399 salzenberg (discussion implies arguments, but his only arguments
400 were ad-hominems, one wonders why he started it in the first
401 place). I hope this explains it well enough for him to understand,
402 and maybe well enough for others to understand.
403 - use common::sense everywhere now.
404 - idle callbacks are no longer supported, use idle coros instead.
405 - print a thread listing when a deadlock is detected.
406
4075.17 Sat Aug 22 23:09:31 CEST 2009
408 - work around a bug in the perl debugger causing crashes
409 when running under the debugger by marking _pool_handler as nodebug.
410 - speed up Coro::async considerably.
411 - try some hacks to get netbsd to work "more often" - their broken
412 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
413 - convert Coro to AE 5.0 API.
414
4155.162 Tue Jul 28 04:04:03 CEST 2009
416 - perl 5.8.2 is now minimum requirement.
417 - skip t/19_handle.t on broken windows perls.
418
4195.161 Wed Jul 22 04:47:38 CEST 2009
420 - Coro::AnyEvent::poll could have a different prototype when EV was
421 used as backend (analyzed by Tatsuhiko Miyagawa).
422 - Coro::AnyEvent errornously initialised the event loop when loaded,
423 not on demand.
424 - try to workaround rare */t/01_unblock.t failures.
425
4265.16 Tue Jul 21 01:44:37 CEST 2009
427 - Coro::AnyEvent failed to hook into the event loop
428 when no threads had been readied between detecting
429 the event loop and actually running it.
430 - considerably speed up Coro::Select by taking avdantage
431 of AnyEvent > 4.8 and some other optimisations.
432 - implement paragraph readline mode in Coro::Handle
433 (based on patches by Zsbán Ambrus).
434 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
435 and analyzed by Yasuhiro MATSUMOTO).
436 - clarified libcoro license and copyright.
437 - someone stole my EXTRA_META!!!
438 - implement Coro::Select::patch_pp_sselect and it's brother,
439 for hardcode select overriding.
440
4415.151 Mon Jul 6 05:41:57 CEST 2009
442 - backport to windows process emulation code again (patch by
443 Yasuhiro MATSUMOTO).
444 - slightly update Coro::MakeMaker.
445
4465.15 Tue Jun 30 10:28:06 CEST 2009
447 - deprecate Coro::Socket, document how to get ipv6 support via
448 AnyEvent::Socket instead.
449 - implement signal->wait ($cb) interface, similar to semaphores.
450 - work around SvOK not supporting getmagic, so we have to getmagic
451 to test for undef :/ (reported by Matthias Waldorf).
452 - load Coro::AnyEvent in all modules using AnyEvent.
453 - work around perl corrupting our internal data structures,
454 reported by Tokuhiro Matsuno.
455 - enable per-coroutine real and cpu time gathering
456 (Coro::State::enable_times).
457
4585.14 Wed Jun 24 01:37:48 CEST 2009
459 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
460 as this seems to have been a difficult concept (poll, sleep, idle,
461 idle_upto).
462 - add Coro::AnyEvent::readable/writable functions.
463 - clarify Coro::EV/Event/AnyEvent manpages.
464 - free per-thread global scalars in the thread calling ->cancel, to
465 avoid crashes when $_, $@ etc., are magical but some of those
466 had already been freed.
467 - "unexperimentalise" the callback interface for Coro::Semaphore.
468 - speed up ready queue management/context switching by using a linked
469 list instead of an array (~5-10%).
470 - implement "watch" command in Coro::Debug shells.
471 - for fun, implement time-slicing as an example in the manpage.
472 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
473 Coro::Event, use the normal AnyEvent handling instead of dieing
474 (the same is true for Coro::Handle).
475 - properly document Coro::EV::timed_io_once.
476 - avoid unneccessary ->cancel calls in Coro::Handle.
477 - maybe make it work on mingw32 with win32 backend
478 (based on patch by Yasuhiro Matsumoto).
479
4805.132 Fri May 29 09:00:39 CEST 2009
481 - do not keep a reference to the argument itself in
482 Coro::Semaphore::guard, as it could change later.
483 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
484 should support a prepare callback.
485
4865.131 Mon Mar 16 23:20:37 CET 2009
487 - implement and document Coro->suspend, Coro->resume.
488 - fix Coro::Select implementation to not (often) close
489 the passed file descriptors (testcase provided by pippijn).
490
4915.13 Mon Dec 15 21:51:42 CET 2008
492 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
493 - don't set diehook to C<undef> but instead to NULL, to avoid
494 spurious warnings.
495 - fix a lot of bugs in Coro::SemaphoreSet.
496 - Coro::SemaphoreSet will less often create a semaphore needlessly.
497 - add Coro::SemaphoreSet::count and wait methods.
498 - take advantage of the new Guard module.
499 - deprecate Coro::guard.
500 - try to fix the dreaded 01_unblock tests once more. I hate it when
501 testsuites need more fixing than the code they are supposed to test.
502 - croak in more cases when a required callback isn't resolvable.
503 - fix some minor issues in Coro::State->call/eval.
504 - use current coroutine context instead of a temporary one
505 when temporarily switching to another coroutine.
506 - do not call C-level on_destroy handlers during global destruction,
507 to avoid needless segfaults.
508
5095.12 Sun Dec 7 13:30:38 CET 2008
510 - add default config for MirOS, which seems to be a bug-to-bug
511 compatible fork of openbsd ("world domination by releasing
512 openbsd cvs before the openbsd folks do it" or so :).
513 - free_padlist did destroy the names pad, not good, but didn't
514 seem to bother perl - this could fix issues such as eval ""
515 inside a function called from multiple coroutines.
516 - use a different method to detect destruction time.
517 - be more careful when freeing padlists just before global
518 destruction.
519 - fixed and expanded the call/cc example.
520 - renamed _terminate to _coro_run.
521 - new method Coro::Channel->shutdown.
522 - try pthreads on openbsd <4.4 (broken sigaltstack, will
523 pthreads fare better?).
524 - be less picky about destroying "a" running coroutine.
525
5265.11 Tue Nov 25 21:49:05 CET 2008
527 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
528 working code with failed assertions, introducing more bugs than
529 it fixes, requiring elaborate workarounds :(
530
5315.1 Mon Nov 24 08:54:59 CET 2008
532 - wrote a small introductory tutorial - Coro::Intro.
533 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
534 - Coro::Select did errornously dup the file descriptors
535 and didn't work with all AnyEvent backends.
536 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
537 LWP data transfers.
538 - disassociate c contexts from coro objects - this is agruably more
539 correct, but mostly allows sharing of cctxs between coro and state
540 objects, for added memory savings and speed increases.
541 - bumped $Coro::POOL_RSS up to 32kb by default.
542 - no longer set the optype to OP_CUSTOM, as B::* understandably
543 doesn't like this very much (and we *are* a type of entersub).
544 - implement state cloning, just to prove that call/cc can be done.
545 - automatically load Coro::AnyEvent in Coro::Handle.
546 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
547 do not have this method (and don't need it either).
548 - speed up generic anyevent methods in Coro::Handle by using rouse
549 callbacks.
550 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
551 others. It also makes the debugger happier, as you can trace
552 through the idle threads now.
553 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
554 - eg/event was pretty broken.
555 - better 5.8.6 compatibility.
556
5575.0 Thu Nov 20 10:35:05 CET 2008
558 - NEW ARCHITECTURE: use the latest 4.x version if you experience
559 stability issues.
560 - bump API version to 7 - all dependents must be recompiled.
561 - removed timed_* functions - they were not being used anyways
562 and should be replaced by a more generic mechanism -
563 and were annoying to support anyways :)
564 - removed SemaphoreSet's waiter method - use sem method instead.
565 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
566 - async_pool did free a scalar value twice
567 ("Attempt to unreference...").
568 - fix a longstanding bug where calling terminate on a coro that
569 was waiting for a semaphore that was just becoming available
570 would cause a deadlock (semaphore would get into a state where
571 it was available but waiters were still blocked).
572 - calling throw on a coroutine that is waiting for a semaphore will
573 no longer make it acquire the semaphore (and thus leak a count).
574 - perl's process emulation is now not even theoretically supported
575 anymore.
576 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
577 conversion of callback-style to blocking-style.
578 - new methods $coro->schedule_to and ->cede_to, to specifically
579 schedule or cede to a specific coroutine.
580 - new function Coro::Semaphore::wait.
581 - use named constants in Coro::Channel (Richard Hundt).
582 - directly patch the entersub opcode calling SLF functions (cede,
583 transfer and so on). this does speed up context switching, but
584 more importanly, it frees us from the hardcoded behaviour of
585 entersub, so we might actually be able to return something from
586 those functions and atcually create new ones.
587 - take advantage of __builtin_frame_address on gcc.
588 - expose THX in coroapi (not sure whether this was a wise decision,
589 as "threaded" perls are running at half speed anyways).
590 - implement execute_slf (schedule-like-function) interface that makes
591 it possible to implement schedule-like-functions in XS.
592 - use new SLF interface to massively speed up Coro::EV by roughly a
593 factor of two.
594 - used new SLF interface to massively speed up Coro::Semaphore by a
595 factor of three.
596 - used new SLF interface to speed up Coro::AIO by roughly a factor of
597 four and reduce its memory usage considerably.
598 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
599 for a nice speedup and vastly more correct behaviour. Also implement
600 a new method "sem" to get at the underlying semaphore object.
601 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
602 (in comparison) 20-40% speedup.
603 - used new SLF interface to reimplement Coro::Signal gaining
604 some unknown (because I was too lazy), but certain, speedup, and also
605 making signals reliable for the first time.
606 - used new SLF interface and other optimisations to speed up async_pool
607 by a factor of two. It also doesn't rely on perl's exception mechanism
608 to exit anymore. The overhead for terminating an async_pool, coro over
609 a normal async is now very small.
610 - sped up coroutine creation/destruction by 40%.
611 - forgot to include Coro/libcoro/README in the dist for all these years.
612 - work around a freebsd pthreads bug (manual testcancel is required as
613 pthread_cond_wait isn't a cancellation point on freebsd).
614 - use new rouse functions to speed up and simplify Coro::BDB.
615 - make "prefer perl native functions" work with threaded perls.
616 - condense Coro::Debug ps output, hint at v and w flags.
617 - (libcoro) lots of minor cleanups and portability improvements.
618
6194.914 Wed Nov 19 12:54:18 CET 2008
620 - fix a disastrous bug in the readline optimisation
621 introduced in 4.801.
622
6234.913 Sat Nov 15 07:58:28 CET 2008
624 - async_pool did free a scalar value twice
625 ("Attempt to unreference...").
626
6274.912 Thu Nov 13 18:31:23 CET 2008
628 - minor cleanups.
629 - use much larger stacks on linux and perl < 5.8.8.
630 - Coro::Debug::new_unix_server did not unlink the socket
631 when destroyed.
632
6334.911 Tue Nov 11 04:26:17 CET 2008
634 - "port" to threaded perls.
635
6364.91 Mon Nov 10 05:36:38 CET 2008
637 - the ->throw exception object no longer leaks.
638 - creating a new cctx leaked a scopestack entry (memleak).
639 - new coroutines didn't get created with a zero flags field
640 (unknown impact).
641 - calling ->throw on a not-yet-started coroutine should now work
642 instead of being ignored.
643 - ->throw is now supported on Coro::State objects.
644 - clean up cctx creation code a bit.
645 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
646
6474.9 Sat Nov 8 17:45:27 CET 2008
648 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
649 - (libcoro) no longer leak threads in the experimental pthread backend,
650 also speed it up considerably.
651 - (libcoro) do not rely on makecontext passing void *'s unscathed.
652 - fix compiletime dependencies on libcoro in the Makefile.
653 - cctx_count wasn't always updated properly.
654 - Coro::State::cctx_stacksize wasn't applied correctly.
655 - new function Coro::State::cctx_max_idle.
656 - the default max number of idle C contexts is now 4.
657 - (libcoro) try harder to get _setjmp/_longjmp.
658 - (libcoro) cleanup and extend the libcoro API to officially
659 allow the creation of empty source contexts.
660 - very experimental workaround for the totally broken netbsd platform.
661 - tried to hack around openbsd bugs.
662
6634.804 Wed Nov 5 16:36:00 CET 2008
664 - Coro::Debug::new_unix_server would not create a non-blocking listening
665 socket, sometimes causing freezes.
666 - (libcoro) fix misaligned stack points for setjmp and assembly
667 methods, which can cause crashes on x86/x86_64 with a sufficiently
668 aggressive compiler.
669 - new function: Coro::Debug::new_tcp_server.
670 - move ->throw into the Coro class because it only works on coro objects.
671
6724.803 Mon Nov 3 17:16:12 CET 2008
673 - (libcoro) use a global asm statement to become independent of gcc
674 otpimisations for CORO_ASM (thanks to pippijn for the idea).
675 - try to workaround yet another broken bsd, this time dragonfly.
676
6774.802 Thu Oct 30 10:56:12 CET 2008
678 - support -fno-omit-frame-pointer on x86 with the assembly method.
679 - tune 01_unblock.t tests a bit.
680
6814.801 Wed Oct 22 18:33:37 CEST 2008
682 - improve readline speed for very long "lines".
683 - backport to 5.8.8.
684
6854.8 Thu Oct 2 13:34:40 CEST 2008
686 - new function Coro::AIO::aio_wait.
687 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
688 - greatly speed up and reduce memory usage of Coro::AIO requests.
689 - implement some other µ-optimisations.
690
6914.749 Mon Sep 29 14:40:12 CEST 2008
692 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
693
6944.748 Sat Sep 27 14:03:19 CEST 2008
695 - implement, but do not document, PerlIO::cede(granularity).
696 - Coro::Storable forgot to wrap Storable::pstore.
697 - work around the multitude of leaks and memory corruption
698 bugs in PerlIO::via by using our own C-level perliol. As a side
699 effect, Coro::Storable is now much, much, much faster.
700
7014.747 Tue Sep 23 01:59:41 CEST 2008
702 - fix a per-cv memleak (one empty array was leaked per
703 code reference).
704 - avoid a crash in coro->call|rss when the coroutine was already
705 destroyed (most noticably when using Coro::Debug::ps :)
706 - also protect *Storable::FILE.
707 - push up default storable granularity to 20ms.
708
7094.746 Sun Sep 21 03:22:20 CEST 2008
710 - be more insistent on locking Storable against reentrancy
711 in Coro::Storable.
712 - move swap_def?v and throw to Coro::State, as documented.
713
7144.745 Thu Jul 24 00:14:38 CEST 2008
715 - remove debugging code related to MgPV_nolen_const, also try to
716 make it compile with perl 5.8.6 (yes, apple apparently loves
717 outdated software). Reported by John S.
718
7194.744 Tue Jul 8 22:06:35 CEST 2008
720 - correctly provide default DIE/WARN handlers as documented.
721 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
722 implementations inside perl work fine for us.
723
7244.743 Mon Jun 16 00:21:57 CEST 2008
725 - when using Coro::EV without running EV::loop it could
726 result in busy-waiting for events, this has been fixed.
727 - reduce codesize and improve performance by using EV_DEFAULT_UC.
728
7294.742 Sat May 31 14:10:21 CEST 2008
730 - implement a workaround for (some) perl <5.8.8 versions.
731 - require EV 3.3+.
732
7334.741 Fri May 30 23:33:09 CEST 2008
734 - tell netbsd how utterly broken their imitation of an OS is
735 and refuse to build by default if pthreads are in use.
736 - switch to "s" method on all bsd's by default, as their ucontext
737 stuff seems just too broken.
738 - fix a bug in Coro::Select.
739
7404.74 Thu May 29 20:05:31 CEST 2008
741 - do not test Coro::LWP for lack of dependencies.
742 - do not test Coro::Select for lack of working perls.
743
7444.73 Thu May 29 2008
745 - fix a bug in Coro::EV which would cause it to block despite
746 there being runnable coroutines.
747 - sprinkle "no warnings" freely over everything, also suppress
748 warnings for some other modules.
749 - fix typo in WSAEWOULDBLOCK.
750
7514.72 Sun May 25 05:14:36 CEST 2008
752 - tweak META.yaml a bit, unfortunately, there is no documented way
753 to have optional dependencies with CPAN. doh :(
754 - avoid running some tests on windows because they would fail due to
755 perl bug (broken fork, broken pipes...).
756 - work around perl on windows bugs where perl returns undocumented
757 error codes for sysread, syswrite etc. by taking advantage
758 of AnyEvent's workaround for that problem.
759 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
760 a common perl implementation bug on windows.
761 - use unix domain sockets in testsuite to work around
762 common perl implementation bugs on widows (they are emulated by
763 tcp sockets on windows. ugh.)
764
7654.71 Sat May 24 20:01:27 CEST 2008
766 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
767 - take advantage of async name resolution of AnyEvent::Util.
768 - work around brutal inet_aton override in Coro::LWP.
769 - take advantage of the readyhook in Coro::EV, for smoother
770 scheduling.
771
7724.7 Sun May 11 00:32:19 CEST 2008
773 - completely reworked the Coro manpage.
774 - added Coro::AnyEvent, generic event loop integration.
9 - implement cancel, ready and kill commands in Coro::Debug. 775 - implement cancel, ready and kill commands in Coro::Debug.
10 - document find_coro in Coro::Debug. 776 - document find_coro in Coro::Debug.
777 - incompatible change: rename has_stack to has_cctx.
778 - Coro::AIO and Coro::BDB no longer force event model detection,
779 use AnyEvent::AIO and AnyEvent::BDB.
11 780
124.6 Sat Apr 26 10:05:14 CEST 2008 7814.6 Sat Apr 26 10:05:14 CEST 2008
13 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported 782 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
14 by the Coro module. It was a mistake to have it in the first place. 783 by the Coro module. It was a mistake to have it in the first place.
15 - (experimental) support for activestate perl 5.10 (method "w"). 784 - (experimental) support for activestate perl 5.10 (method "w").
… …
134 903
1354.11 Thu Oct 11 02:40:24 CEST 2007 9044.11 Thu Oct 11 02:40:24 CEST 2007
136 - port to threaded perls. 905 - port to threaded perls.
137 906
1384.1 Thu Oct 11 02:38:16 CEST 2007 9074.1 Thu Oct 11 02:38:16 CEST 2007
139 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 908 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
140 be local to each coro (see Coro::State). 909 be local to each coro (see Coro::State).
141 - incompatible change: for very deep reasons, cede and cede_notself 910 - incompatible change: for very deep reasons, cede and cede_notself
142 cannot return anything, so nothing will be returned. 911 cannot return anything, so nothing will be returned.
143 - possibly bring back 5.10 compatibility (untested). 912 - possibly bring back 5.10 compatibility (untested).
144 - work around stupid (and wrong) warning on 5.10 :(. 913 - work around stupid (and wrong) warning on 5.10 :(.
… …
216 but stateful semantics. 985 but stateful semantics.
217 - fixed a lot of typos in Coro.pm (patch submitted by David 986 - fixed a lot of typos in Coro.pm (patch submitted by David
218 Steinbrunner, which applied flawlessly). 987 Steinbrunner, which applied flawlessly).
219 988
2203.6 Sat Apr 14 17:13:31 CEST 2007 9893.6 Sat Apr 14 17:13:31 CEST 2007
221 - added some bugfixes to get eg/myhttpd working again. 990 - added some bugfixes to get eg/myhttpd working again.
222 - added Coro::Storable for often-cede'ing freeze/thaw. 991 - added Coro::Storable for often-cede'ing freeze/thaw.
223 - try to do a clean exit when a coroutine calls exit 992 - try to do a clean exit when a coroutine calls exit
224 (EXPERIMENTAL). 993 (EXPERIMENTAL).
225 - got rid of indirect call through _coro_init. 994 - got rid of indirect call through _coro_init.
226 - updated the partly antique examples in eg/ to 995 - updated the partly antique examples in eg/ to

Diff Legend

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