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

Comparing Coro/Changes (file contents):
Revision 1.96 by root, Tue Aug 10 01:56:30 2004 UTC vs.
Revision 1.676 by root, Thu Nov 29 21:17:06 2018 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2
3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: channel->maxsize(newsize)?
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?
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
166.52 Tue Aug 14 18:49:43 CEST 2018
17 - libcoro update: allow sharing of coro threads using the pthreads
18 backend among different pthreads, as is required by Coro::Multicore.
19 - always allow ->call and ->eval when the coro is the current coroutine. this
20 makes it possible, among other things, to call these on $Coro::main.
21 - minor code cleanups.
22
236.514 Thu Aug 31 18:25:31 CEST 2017
24 - safe_cancel no longer croaks when called on already-destroyed
25 thread, but instead "just works".
26
276.513 Sat Jul 15 01:18:51 CEST 2017
28 - experimental 5.22 comaptibility changes (PadlistNAMES).
29 - removed attributes example and made sure the other examples
30 still work.
31 - fix some statevar types for newer perls, should be of no consequence.
32
336.512 Fri Jul 14 05:03:17 CEST 2017
34 - seems to pass test suite down till 5.8.9, so make this the
35 minimum requirement in Makefile.PL.
36 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
37 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
38 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
39 fortified longjmp variants in some libcs.
40 - re-try JIT allocation without PROT_EXEC, to hopefully improve
41 portability (still doesn't seem to work around broken selinux
42 mmap everywhere).
43 - speling fixes (Xavier Guimard).
44
456.511 Sun Jun 26 23:44:50 CEST 2016
46 - make it compile with threaded perls again.
47 - simplify and speed up __DIE__ and __WARN__ handling.
48
496.51 Sat Jun 25 20:57:20 CEST 2016
50 - 6.5 release didn't compile on most older perls due to
51 a logic error enabling experimental 5.24 code for all versions.
52 - since various broken or deficient patches float around
53 to work around the vtbl API breakage in perl, let's provide
54 our own workaround, which seems to be much less invasive
55 and more compatible than the approaches seen so far. YMMV.
56 - work around assertion failure in perl_destruct on 5.24
57 (analyzed by Dave Mitchell).
58
596.5 Wed Jun 22 22:23:50 CEST 2016
60 - swap_sv swapping order was not symmetrical, causing
61 wrong swaps when swap_sv was used multiple times on the
62 same sv in the same thread.
63 - swap_sv calls can now be undone by calling it again
64 with the same variables.
65 - swap_sv calls will now be undone in async_pool threads.
66 - split Coro::Semaphore::up/adjust into separate xs functions
67 for better error reporting, at slight codesize increase.
68 - (libcoro) arm assembly support, please test and report.
69 - adjust to PL_savestack changes in perl 5.24 (adapted from
70 the debian patch, which unfortunately gets it wrong).
71
726.49 Sat Oct 17 01:40:12 CEST 2015
73 - throwing an exception to a thread waiting in
74 Coro::Handle using Coro::EV did not stop the watchers,
75 causing the next call to fail (testcase by Martin Pritchard).
76 - bump minimum perl version to 5.10.
77
786.48 Sun Oct 4 19:03:51 CEST 2015
79 - fix memory leak when Coro::Handle uses Coro::EV internally
80 (testcase by Sten Sten).
81 - update code to libev 4 API (internally, Coro still used the
82 version 3 API calls).
83 - remove unused hv_sig.
84
856.47 Sat Jul 11 03:58:20 CEST 2015
86 - swap_svs - NVs can be in the SV head beginning with 5.20.
87
886.46 Tue Jun 30 14:36:30 CEST 2015
89 - restore portability to perls with windows fork emulation
90 (patch by Petr Písař).
91
926.45 Tue Jun 30 01:40:08 CEST 2015
93 - restore portability to perl 5.16 and below.
94
956.44 Tue Jun 30 00:41:54 CEST 2015
96 - avoid segfaulting (or worse) when tracing a canceled thread.
97 - the Event module silently broke it's public hook API, causing
98 failures on perls compiled with -Duselongdouble. Requiring
99 the latest version of Event because I don't know which release
100 changed this (it's not mentioned in the ChangeLog).
101 - allow xs level enterleave hooks via CoroAPI.
102
1036.43 Thu Jun 4 15:38:14 CEST 2015
104 - use stability canary.
105 - port to stableperl-5.22.0-1.001.
106 - update libecb.
107
1086.42 Wed Feb 11 20:29:52 CET 2015
109 - Coro::SemaphoreSet->try did not actually work (analyzed by
110 SATO Kentaro).
111 - upgrade libecb to be C11 compliant.
112
1136.41 Sat Sep 6 22:08:46 CEST 2014
114 - restore portability to perl 5.8.x.
115 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
116 up some memory, but fixes "use feature" and similar modules
117 when used inside a Coro.
118 - allow zero as argument to Coro::Channel to mean the same thing
119 as no argument. this works with older versions as well,
120 but wasn't legal until now.
121 - slightly better c header file detection.
122
1236.39 Mon Jun 2 00:00:08 CEST 2014
124 - work around more incompatible changes in 5.20.
125
1266.38 Sun Jun 1 21:54:23 CEST 2014
127 - check that perl slots actually have enough space to hold
128 interpreter variables.
129 - untested port to perl 5.19 (64 bit tmps indices) (reported
130 by Andreas König).
131 - croak when cancel is called without a thread context.
132
1336.37 Tue Mar 4 13:27:33 CET 2014
134 - *sigh*, removed leftover debugging code from debugging a
135 perl bug, of all things.
136
1376.36 Tue Mar 4 07:11:59 CET 2014
138 - semaphores would not clear the destroy hook when interrupted
139 by ->throw, causing segfaults or worse.
140 - ->throw on a thread waiting for a semaphore did not acquire
141 the semaphore, but also didn't wake up other waiters,
142 possibly causing a deadlock.
143 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
144 to report a full stacktrace, which should help find out
145 where the actual call is.
146 - "atomically" destroy data in slf_destroy, because it is
147 the right thing to do, just in case.
148 - disable ecb.h libm dependency, to compile on ancient systems
149 or under adverse conditions.
150
1516.33 Mon Nov 18 11:26:27 CET 2013
152 - do not crash when freeing padlists with holes (in 5.18).
153 - tentative SVt_BIND 5.19 port/fix.
154
1556.32 Tue Nov 5 15:35:35 CET 2013
156 - use a new algorithm to derive padlists for perl 5.18. The old
157 one could lead to 0-pointer accesses inside perl (reported
158 by Darin McBride).
159
1606.31 Thu May 9 07:39:48 CEST 2013
161 - Coro::AIO requests would crash if the thread was ready'd
162 while the request was ongoing.
163
1646.29 Wed May 8 02:55:18 CEST 2013
165 - when an on_destroy handler destructs the coro currently being
166 destructed a perl scalar could be accessed after being freed,
167 likely causing a crash.
168
1696.28 Wed Mar 6 06:58:02 CET 2013
170 - clean remnants of existing __DIE__ and __WARN__ handlers so
171 they lose their magic and will not cause segfaults later
172 (testcase by Andrey Sagulin).
173 - improved Coro::State documentation a bit.
174 - Coro::Debug::command now flushes the output.
175 - add hack detection code for x32 abi, because the braindead slugs
176 who designed that made it look exactly like x86_64 without
177 providing proper compile time symbols to test for it. as a result,
178 this detection cannot work reliably.
179 - valgrind stack registering was broken.
180 - do not rely on Time::HiRes anymore in Coro::Debug.
181
1826.23 Fri Dec 7 23:36:37 CET 2012
183 - use experimental fiber implementation on native windows
184 perls.
185 - use sizeof (void *) as multiplication factor for stack sizes,
186 to accomodate the totally braindamaged microsoft 64 bit "os".
187 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
188 activeperl 5.16 crashes when PerlIO_define_layer is called due
189 to some bug in the perl dll, strawberry perl at least passes
190 the testsuite.
191 - implement Coro::Handle->peeraddr/host/port, for slightly
192 improved compatibility with LWP.
193 - implement 5.17 compatibility by almost blindly applying a
194 good-looking patch by Father Chrysostomos.
195 - move stack management functions into libcoro 3.
196 - libcoro version 3 "released".
197 - support magic values as timed_io_once args.
198 - recommend AnyEvent 7+ or EV 4+, also require EV
199 version 4 or newer for Coro::EV.
200
2016.10 Tue Oct 9 01:14:27 CEST 2012
202 - updated ecb.h, it had a typo that caused it to not compile on many
203 big endian systems (reported by many people).
204 - disable memory fences in ecb.h to improve portability.
205
2066.09 Sat Oct 6 23:25:02 CEST 2012
207 - Coro::EV I/O watchers were not interruptible by exceptions
208 (Coro::State::throw) (testcase by sten).
209 - ->throw now puts threads into the ready queue, as this seems to
210 be expected by existing code, and code that doesn't cope with spurious
211 wakeups needs fixing anyway.
212 - use fd -1 in mmap.
213 - cast I32 to int in error message printf.
214 - warn about broken so-called "hardened" kernels.
215
2166.08 Fri Apr 13 12:05:47 CEST 2012
217 - be more aggressive about exiting like perl does - formerly,
218 exiting from the non-main thread would not execute END blocks.
219
2206.07 Fri Nov 11 21:21:48 CET 2011
221 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
222 - update ecb.h.
223
2246.06 Mon Aug 8 23:59:48 CEST 2011
225 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
226 backend, which is an order of magnitude slower. unfortunately, cygwins
227 pthread implementation isn't very complete either, so allocate the stack
228 twice just to be sure.
229 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
230 in its ithreaded perl - best recompile cygwin and use the 'w'indows
231 backend for much better performance. also disable ithreads for
232 even better performance...).
233
2346.05 Thu Aug 4 21:36:36 CEST 2011
235 - blush, condvar values would not be propagated from send to recv anymore
236 (reported by Chip Salzenberg).
237 - use exponential increase for the readline buffer length in
238 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
239
2406.04 Wed Aug 3 17:15:45 CEST 2011
241 - use even more efficient and more compatible condvars for
242 compatibility to AnyEvent 6.x :)
243 - more inconsequential ecb.h updates.
244
2456.03 Wed Aug 3 11:41:30 CEST 2011
246 - change how Coro patches AnyEvent condvars for compatibility to
247 AnyEvent 6.x.
248 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
249 hasn't been included.
250
2516.02 Wed Jul 13 04:35:19 CEST 2011
252 - "improve portability to Gentoo" - gentoo manages to put perl variables
253 in memory areas that are farther than 2gb apart, which the jit couldn't
254 handle and barfed. now it's just a bit slower on gentoo and similar
255 systems.
256
2576.01 Sun Jul 3 12:31:14 CEST 2011
258 - workarounds are good, but the test for whether pthreads are used
259 was not good. this one should be better.
260 - check differently whether gcc generates cfi instructions itself.
261
2626.0 Wed Jun 29 19:43:35 CEST 2011
263 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
264 destroyed and cleaned up automatically (e.g. async { schedule }).
265 - implement a JIT compiler for part of the thread switch code,
266 which gives a 50% speed improvement on threaded perls, and
267 about 4% on non-threaded perls (so threaded perls now finally
268 reach about half the speed of non-threaded perls).
269 - slightly modernise Coro::Intro, add section about rouse functions.
270 - avoid DEFSV and ERRSV, giving another 10% improvement
271 in thread switching.
272 - Coro::State->is_destroyed is now called is_zombie.
273 - implement a Coro->safe_cancel method that might fail, but
274 cancels in a "safer" way if it succeeds.
275 - add preliminary support for DEBUGGING perls.
276 - get rid of two hash-accesses when initialising a new Coro - this
277 speeds up coro creation by almost a factor of two.
278 - croak when a coro that is being cancelled tries to block
279 (e.g. while executing a guard block), instead of crashing or
280 deadlocking.
281 - use a more robust and also faster method to identify Coro::State
282 objects - speeds up everything a bit.
283 - implement Coro->cancel in XS for a 20% speed improvement, and to
284 be able to implement mutual cancellation.
285 - speed up context switches by a percent or two by more efficiently
286 allocating context stack entries.
287 - implement Coro->join and Coro->on_destroy in XS for a speedup and
288 a reduction in memory use.
289 - cancelling a coro while it itself is cancelling another coro is
290 now supported and working, instead of triggering an assertion.
291 - be a bit more crash-resistant when calling (buggy) on_destroy
292 callbacks (best effort).
293 - move on_destroy into the slf_frame, to allow extension slf
294 functions to have destructors.
295 - get rid if coro refcounting - simply crash in other interpreter
296 threads by nulling the pointers on clone.
297 - simplify warn/die hook handling when loading Coro - the convoluted
298 logic seems to be no longer neccessary.
299 - use libecb instead of our own home-grown gcc hacks.
300 - document alternatives to Coro::LWP. Please use them :)
301 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
302 sigaltstack. Really. wine suffers from it, erlang suffers from it,
303 and it's known since at least 2006.
304
3055.372 Wed Feb 23 06:14:30 CET 2011
306 - apparently mingw doesn't provide a working gettimeofday, try to
307 work around that by relying on Time::HiRes (indirectly brought to
308 my attention by Max Maischein).
309 - fix some portability issues when Time::HiRes was used.
310
3115.371 Mon Feb 21 14:36:08 CET 2011
312 - backport to windows process emulation code again.
313
3145.37 Sat Feb 19 07:49:44 CET 2011
315 - add a big "Coro thread life cycle" section to "man Coro".
316 - try a tentative workaround against the breakage that 5.13 has
317 introduced without depreciation period. sigh.
318 - no longer use Time::HiRes if gettimeofday is available, which
319 saves quite a lot of memory.
320
3215.36 Sun Feb 13 05:33:41 CET 2011
322 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
323 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
324 method call.
325 - undocument Coro::Timer::sleep and obsolete whole module.
326 - optimise Coro::Timer::timeout memory and cpu usage.
327 - slightly updated Coro::Intro for recent changes.
328 - do not initialise PL_dirty anymore.
329
3305.25 Thu Nov 11 01:08:39 CET 2010
331 - try a different approach on netbsd - netbsd 5 finally has marginally
332 working pthreads, but still broken ucontext/sigaltstack.
333 - openbsd 4.8 finally got their act together, Coro works out of the box
334 with asm, setjmp and pthreads (no change, just informational).
335
3365.24 Sat Oct 23 11:27:12 CEST 2010
337 - port to the EV 4.0 API.
338 - work around bugs in mingw32, making strawberry perl work
339 out of the box.
340 - correctly modify Coro::AIO function prototypes
341 so that they reflect the "no optional parameters" rule.
342 - "ported" libcoro to C++.
343
3445.23 Mon May 17 18:50:42 CEST 2010
345 - be more resistant to ordering changes when initialising
346 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
347 Waldorf).
348 - document that perl 5.12 deliberately removed support for cloning.
349
3505.22 Wed Apr 14 03:55:35 CEST 2010
351 - correctly return udnef on errors in Coro::Handle::read/write
352 (testcase by Marc Mims).
353 - convert Coro::Util into a "perl compatibility wrapper" - the functions
354 are less useful now, but are drop-in replacements for existing
355 functions, listing better alternatives in the documentation. This also
356 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
357 with Coro::Util::inet_aton.
358 - do not override $Coro::idle unconditionally in Coro.pm, as other
359 modules could have provided their own idle coro already
360 (for exmaple, Coro::AnyEvent).
361 - fix Coro::Util::gethost* functions.
362 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
363
3645.21 Wed Dec 16 07:19:51 CET 2009
365 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
366 together.
367 - add some examples on how to combine other event loops with Coro in
368 Coro::AnyEvent, and how to run it (and not to block). Seems to be
369 the most common source of confusion.
370 - try to catch people naively blocking in an event callback.
371 - work around the perl filehandle bug issue in conjunction with
372 older common::sense (as indirectly pointed out by ZSystem).
373 - clarify the "not from signal handlers" section.
374
3755.2 Sun Oct 4 14:54:24 CEST 2009
376 - Coro::Storable destroyed the prototypes of the functions it wrapped.
377 - export rouse_cb and rouse_wait by default now.
378 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
379 - new method $state->swap_sv.
380 - added section on "windows process emulation" to the manpage,
381 after a not-so-fruitful (nor-friendly) "discussion" with chip
382 salzenberg (discussion implies arguments, but his only arguments
383 were ad-hominems, one wonders why he started it in the first
384 place). I hope this explains it well enough for him to understand,
385 and maybe well enough for others to understand.
386 - use common::sense everywhere now.
387 - idle callbacks are no longer supported, use idle coros instead.
388 - print a thread listing when a deadlock is detected.
389
3905.17 Sat Aug 22 23:09:31 CEST 2009
391 - work around a bug in the perl debugger causing crashes
392 when running under the debugger by marking _pool_handler as nodebug.
393 - speed up Coro::async considerably.
394 - try some hacks to get netbsd to work "more often" - their broken
395 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
396 - convert Coro to AE 5.0 API.
397
3985.162 Tue Jul 28 04:04:03 CEST 2009
399 - perl 5.8.2 is now minimum requirement.
400 - skip t/19_handle.t on broken windows perls.
401
4025.161 Wed Jul 22 04:47:38 CEST 2009
403 - Coro::AnyEvent::poll could have a different prototype when EV was
404 used as backend (analyzed by Tatsuhiko Miyagawa).
405 - Coro::AnyEvent errornously initialised the event loop when loaded,
406 not on demand.
407 - try to workaround rare */t/01_unblock.t failures.
408
4095.16 Tue Jul 21 01:44:37 CEST 2009
410 - Coro::AnyEvent failed to hook into the event loop
411 when no threads had been readied between detecting
412 the event loop and actually running it.
413 - considerably speed up Coro::Select by taking avdantage
414 of AnyEvent > 4.8 and some other optimisations.
415 - implement paragraph readline mode in Coro::Handle
416 (based on patches by Zsbán Ambrus).
417 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
418 and analyzed by Yasuhiro MATSUMOTO).
419 - clarified libcoro license and copyright.
420 - someone stole my EXTRA_META!!!
421 - implement Coro::Select::patch_pp_sselect and it's brother,
422 for hardcode select overriding.
423
4245.151 Mon Jul 6 05:41:57 CEST 2009
425 - backport to windows process emulation code again (patch by
426 Yasuhiro MATSUMOTO).
427 - slightly update Coro::MakeMaker.
428
4295.15 Tue Jun 30 10:28:06 CEST 2009
430 - deprecate Coro::Socket, document how to get ipv6 support via
431 AnyEvent::Socket instead.
432 - implement signal->wait ($cb) interface, similar to semaphores.
433 - work around SvOK not supporting getmagic, so we have to getmagic
434 to test for undef :/ (reported by Matthias Waldorf).
435 - load Coro::AnyEvent in all modules using AnyEvent.
436 - work around perl corrupting our internal data structures,
437 reported by Tokuhiro Matsuno.
438 - enable per-coroutine real and cpu time gathering
439 (Coro::State::enable_times).
440
4415.14 Wed Jun 24 01:37:48 CEST 2009
442 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
443 as this seems to have been a difficult concept (poll, sleep, idle,
444 idle_upto).
445 - add Coro::AnyEvent::readable/writable functions.
446 - clarify Coro::EV/Event/AnyEvent manpages.
447 - free per-thread global scalars in the thread calling ->cancel, to
448 avoid crashes when $_, $@ etc., are magical but some of those
449 had already been freed.
450 - "unexperimentalise" the callback interface for Coro::Semaphore.
451 - speed up ready queue management/context switching by using a linked
452 list instead of an array (~5-10%).
453 - implement "watch" command in Coro::Debug shells.
454 - for fun, implement time-slicing as an example in the manpage.
455 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
456 Coro::Event, use the normal AnyEvent handling instead of dieing
457 (the same is true for Coro::Handle).
458 - properly document Coro::EV::timed_io_once.
459 - avoid unneccessary ->cancel calls in Coro::Handle.
460 - maybe make it work on mingw32 with win32 backend
461 (based on patch by Yasuhiro Matsumoto).
462
4635.132 Fri May 29 09:00:39 CEST 2009
464 - do not keep a reference to the argument itself in
465 Coro::Semaphore::guard, as it could change later.
466 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
467 should support a prepare callback.
468
4695.131 Mon Mar 16 23:20:37 CET 2009
470 - implement and document Coro->suspend, Coro->resume.
471 - fix Coro::Select implementation to not (often) close
472 the passed file descriptors (testcase provided by pippijn).
473
4745.13 Mon Dec 15 21:51:42 CET 2008
475 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
476 - don't set diehook to C<undef> but instead to NULL, to avoid
477 spurious warnings.
478 - fix a lot of bugs in Coro::SemaphoreSet.
479 - Coro::SemaphoreSet will less often create a semaphore needlessly.
480 - add Coro::SemaphoreSet::count and wait methods.
481 - take advantage of the new Guard module.
482 - deprecate Coro::guard.
483 - try to fix the dreaded 01_unblock tests once more. I hate it when
484 testsuites need more fixing than the code they are supposed to test.
485 - croak in more cases when a required callback isn't resolvable.
486 - fix some minor issues in Coro::State->call/eval.
487 - use current coroutine context instead of a temporary one
488 when temporarily switching to another coroutine.
489 - do not call C-level on_destroy handlers during global destruction,
490 to avoid needless segfaults.
491
4925.12 Sun Dec 7 13:30:38 CET 2008
493 - add default config for MirOS, which seems to be a bug-to-bug
494 compatible fork of openbsd ("world domination by releasing
495 openbsd cvs before the openbsd folks do it" or so :).
496 - free_padlist did destroy the names pad, not good, but didn't
497 seem to bother perl - this could fix issues such as eval ""
498 inside a function called from multiple coroutines.
499 - use a different method to detect destruction time.
500 - be more careful when freeing padlists just before global
501 destruction.
502 - fixed and expanded the call/cc example.
503 - renamed _terminate to _coro_run.
504 - new method Coro::Channel->shutdown.
505 - try pthreads on openbsd <4.4 (broken sigaltstack, will
506 pthreads fare better?).
507 - be less picky about destroying "a" running coroutine.
508
5095.11 Tue Nov 25 21:49:05 CET 2008
510 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
511 working code with failed assertions, introducing more bugs than
512 it fixes, requiring elaborate workarounds :(
513
5145.1 Mon Nov 24 08:54:59 CET 2008
515 - wrote a small introductory tutorial - Coro::Intro.
516 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
517 - Coro::Select did errornously dup the file descriptors
518 and didn't work with all AnyEvent backends.
519 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
520 LWP data transfers.
521 - disassociate c contexts from coro objects - this is agruably more
522 correct, but mostly allows sharing of cctxs between coro and state
523 objects, for added memory savings and speed increases.
524 - bumped $Coro::POOL_RSS up to 32kb by default.
525 - no longer set the optype to OP_CUSTOM, as B::* understandably
526 doesn't like this very much (and we *are* a type of entersub).
527 - implement state cloning, just to prove that call/cc can be done.
528 - automatically load Coro::AnyEvent in Coro::Handle.
529 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
530 do not have this method (and don't need it either).
531 - speed up generic anyevent methods in Coro::Handle by using rouse
532 callbacks.
533 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
534 others. It also makes the debugger happier, as you can trace
535 through the idle threads now.
536 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
537 - eg/event was pretty broken.
538 - better 5.8.6 compatibility.
539
5405.0 Thu Nov 20 10:35:05 CET 2008
541 - NEW ARCHITECTURE: use the latest 4.x version if you experience
542 stability issues.
543 - bump API version to 7 - all dependents must be recompiled.
544 - removed timed_* functions - they were not being used anyways
545 and should be replaced by a more generic mechanism -
546 and were annoying to support anyways :)
547 - removed SemaphoreSet's waiter method - use sem method instead.
548 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
549 - async_pool did free a scalar value twice
550 ("Attempt to unreference...").
551 - fix a longstanding bug where calling terminate on a coro that
552 was waiting for a semaphore that was just becoming available
553 would cause a deadlock (semaphore would get into a state where
554 it was available but waiters were still blocked).
555 - calling throw on a coroutine that is waiting for a semaphore will
556 no longer make it acquire the semaphore (and thus leak a count).
557 - perl's process emulation is now not even theoretically supported
558 anymore.
559 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
560 conversion of callback-style to blocking-style.
561 - new methods $coro->schedule_to and ->cede_to, to specifically
562 schedule or cede to a specific coroutine.
563 - new function Coro::Semaphore::wait.
564 - use named constants in Coro::Channel (Richard Hundt).
565 - directly patch the entersub opcode calling SLF functions (cede,
566 transfer and so on). this does speed up context switching, but
567 more importanly, it frees us from the hardcoded behaviour of
568 entersub, so we might actually be able to return something from
569 those functions and atcually create new ones.
570 - take advantage of __builtin_frame_address on gcc.
571 - expose THX in coroapi (not sure whether this was a wise decision,
572 as "threaded" perls are running at half speed anyways).
573 - implement execute_slf (schedule-like-function) interface that makes
574 it possible to implement schedule-like-functions in XS.
575 - use new SLF interface to massively speed up Coro::EV by roughly a
576 factor of two.
577 - used new SLF interface to massively speed up Coro::Semaphore by a
578 factor of three.
579 - used new SLF interface to speed up Coro::AIO by roughly a factor of
580 four and reduce its memory usage considerably.
581 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
582 for a nice speedup and vastly more correct behaviour. Also implement
583 a new method "sem" to get at the underlying semaphore object.
584 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
585 (in comparison) 20-40% speedup.
586 - used new SLF interface to reimplement Coro::Signal gaining
587 some unknown (because I was too lazy), but certain, speedup, and also
588 making signals reliable for the first time.
589 - used new SLF interface and other optimisations to speed up async_pool
590 by a factor of two. It also doesn't rely on perl's exception mechanism
591 to exit anymore. The overhead for terminating an async_pool, coro over
592 a normal async is now very small.
593 - sped up coroutine creation/destruction by 40%.
594 - forgot to include Coro/libcoro/README in the dist for all these years.
595 - work around a freebsd pthreads bug (manual testcancel is required as
596 pthread_cond_wait isn't a cancellation point on freebsd).
597 - use new rouse functions to speed up and simplify Coro::BDB.
598 - make "prefer perl native functions" work with threaded perls.
599 - condense Coro::Debug ps output, hint at v and w flags.
600 - (libcoro) lots of minor cleanups and portability improvements.
601
6024.914 Wed Nov 19 12:54:18 CET 2008
603 - fix a disastrous bug in the readline optimisation
604 introduced in 4.801.
605
6064.913 Sat Nov 15 07:58:28 CET 2008
607 - async_pool did free a scalar value twice
608 ("Attempt to unreference...").
609
6104.912 Thu Nov 13 18:31:23 CET 2008
611 - minor cleanups.
612 - use much larger stacks on linux and perl < 5.8.8.
613 - Coro::Debug::new_unix_server did not unlink the socket
614 when destroyed.
615
6164.911 Tue Nov 11 04:26:17 CET 2008
617 - "port" to threaded perls.
618
6194.91 Mon Nov 10 05:36:38 CET 2008
620 - the ->throw exception object no longer leaks.
621 - creating a new cctx leaked a scopestack entry (memleak).
622 - new coroutines didn't get created with a zero flags field
623 (unknown impact).
624 - calling ->throw on a not-yet-started coroutine should now work
625 instead of being ignored.
626 - ->throw is now supported on Coro::State objects.
627 - clean up cctx creation code a bit.
628 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
629
6304.9 Sat Nov 8 17:45:27 CET 2008
631 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
632 - (libcoro) no longer leak threads in the experimental pthread backend,
633 also speed it up considerably.
634 - (libcoro) do not rely on makecontext passing void *'s unscathed.
635 - fix compiletime dependencies on libcoro in the Makefile.
636 - cctx_count wasn't always updated properly.
637 - Coro::State::cctx_stacksize wasn't applied correctly.
638 - new function Coro::State::cctx_max_idle.
639 - the default max number of idle C contexts is now 4.
640 - (libcoro) try harder to get _setjmp/_longjmp.
641 - (libcoro) cleanup and extend the libcoro API to officially
642 allow the creation of empty source contexts.
643 - very experimental workaround for the totally broken netbsd platform.
644 - tried to hack around openbsd bugs.
645
6464.804 Wed Nov 5 16:36:00 CET 2008
647 - Coro::Debug::new_unix_server would not create a non-blocking listening
648 socket, sometimes causing freezes.
649 - (libcoro) fix misaligned stack points for setjmp and assembly
650 methods, which can cause crashes on x86/x86_64 with a sufficiently
651 aggressive compiler.
652 - new function: Coro::Debug::new_tcp_server.
653 - move ->throw into the Coro class because it only works on coro objects.
654
6554.803 Mon Nov 3 17:16:12 CET 2008
656 - (libcoro) use a global asm statement to become independent of gcc
657 otpimisations for CORO_ASM (thanks to pippijn for the idea).
658 - try to workaround yet another broken bsd, this time dragonfly.
659
6604.802 Thu Oct 30 10:56:12 CET 2008
661 - support -fno-omit-frame-pointer on x86 with the assembly method.
662 - tune 01_unblock.t tests a bit.
663
6644.801 Wed Oct 22 18:33:37 CEST 2008
665 - improve readline speed for very long "lines".
666 - backport to 5.8.8.
667
6684.8 Thu Oct 2 13:34:40 CEST 2008
669 - new function Coro::AIO::aio_wait.
670 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
671 - greatly speed up and reduce memory usage of Coro::AIO requests.
672 - implement some other µ-optimisations.
673
6744.749 Mon Sep 29 14:40:12 CEST 2008
675 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
676
6774.748 Sat Sep 27 14:03:19 CEST 2008
678 - implement, but do not document, PerlIO::cede(granularity).
679 - Coro::Storable forgot to wrap Storable::pstore.
680 - work around the multitude of leaks and memory corruption
681 bugs in PerlIO::via by using our own C-level perliol. As a side
682 effect, Coro::Storable is now much, much, much faster.
683
6844.747 Tue Sep 23 01:59:41 CEST 2008
685 - fix a per-cv memleak (one empty array was leaked per
686 code reference).
687 - avoid a crash in coro->call|rss when the coroutine was already
688 destroyed (most noticably when using Coro::Debug::ps :)
689 - also protect *Storable::FILE.
690 - push up default storable granularity to 20ms.
691
6924.746 Sun Sep 21 03:22:20 CEST 2008
693 - be more insistent on locking Storable against reentrancy
694 in Coro::Storable.
695 - move swap_def?v and throw to Coro::State, as documented.
696
6974.745 Thu Jul 24 00:14:38 CEST 2008
698 - remove debugging code related to MgPV_nolen_const, also try to
699 make it compile with perl 5.8.6 (yes, apple apparently loves
700 outdated software). Reported by John S.
701
7024.744 Tue Jul 8 22:06:35 CEST 2008
703 - correctly provide default DIE/WARN handlers as documented.
704 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
705 implementations inside perl work fine for us.
706
7074.743 Mon Jun 16 00:21:57 CEST 2008
708 - when using Coro::EV without running EV::loop it could
709 result in busy-waiting for events, this has been fixed.
710 - reduce codesize and improve performance by using EV_DEFAULT_UC.
711
7124.742 Sat May 31 14:10:21 CEST 2008
713 - implement a workaround for (some) perl <5.8.8 versions.
714 - require EV 3.3+.
715
7164.741 Fri May 30 23:33:09 CEST 2008
717 - tell netbsd how utterly broken their imitation of an OS is
718 and refuse to build by default if pthreads are in use.
719 - switch to "s" method on all bsd's by default, as their ucontext
720 stuff seems just too broken.
721 - fix a bug in Coro::Select.
722
7234.74 Thu May 29 20:05:31 CEST 2008
724 - do not test Coro::LWP for lack of dependencies.
725 - do not test Coro::Select for lack of working perls.
726
7274.73 Thu May 29 2008
728 - fix a bug in Coro::EV which would cause it to block despite
729 there being runnable coroutines.
730 - sprinkle "no warnings" freely over everything, also suppress
731 warnings for some other modules.
732 - fix typo in WSAEWOULDBLOCK.
733
7344.72 Sun May 25 05:14:36 CEST 2008
735 - tweak META.yaml a bit, unfortunately, there is no documented way
736 to have optional dependencies with CPAN. doh :(
737 - avoid running some tests on windows because they would fail due to
738 perl bug (broken fork, broken pipes...).
739 - work around perl on windows bugs where perl returns undocumented
740 error codes for sysread, syswrite etc. by taking advantage
741 of AnyEvent's workaround for that problem.
742 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
743 a common perl implementation bug on windows.
744 - use unix domain sockets in testsuite to work around
745 common perl implementation bugs on widows (they are emulated by
746 tcp sockets on windows. ugh.)
747
7484.71 Sat May 24 20:01:27 CEST 2008
749 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
750 - take advantage of async name resolution of AnyEvent::Util.
751 - work around brutal inet_aton override in Coro::LWP.
752 - take advantage of the readyhook in Coro::EV, for smoother
753 scheduling.
754
7554.7 Sun May 11 00:32:19 CEST 2008
756 - completely reworked the Coro manpage.
757 - added Coro::AnyEvent, generic event loop integration.
758 - implement cancel, ready and kill commands in Coro::Debug.
759 - document find_coro in Coro::Debug.
760 - incompatible change: rename has_stack to has_cctx.
761 - Coro::AIO and Coro::BDB no longer force event model detection,
762 use AnyEvent::AIO and AnyEvent::BDB.
763
7644.6 Sat Apr 26 10:05:14 CEST 2008
765 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
766 by the Coro module. It was a mistake to have it in the first place.
767 - (experimental) support for activestate perl 5.10 (method "w").
768 - (experimental) support for strawberry perl (method "a").
769 - coro_sigelem_set did not return a value although it had to,
770 actual impact unknown.
771
7724.51 Mon Apr 14 13:28:27 CEST 2008
773 - make it compile again on 5.8.
774
7754.50 Thu Apr 10 09:43:17 CEST 2008
776 - I did it twice! (see 4.49).
777
7784.49 Sun Apr 6 21:23:31 CEST 2008
779 - grr, instead of compiling the recent changes on 5.10 only they
780 were compiled on 5.8 only.
781
7824.48 Sun Apr 6 20:36:46 CEST 2008
783 - allow coroutine switches during eval's under 5.10.x, as it seems
784 the parser is a per-interpreter option now, so this is safe
785 (this might fix the odd crashing bug).
786 - drop support for 5.9.x versions: they are dead, jim.
787
7884.47 Sun Apr 6 00:37:52 CEST 2008
789 - force cctx allocation on API calls: we know we need to force one
790 and gcc actually manages to confuse our heuristic nowadays,
791 leading to crashes and worse.
792 - document force_cctx.
793
7944.46 Fri Apr 4 22:05:43 CEST 2008
795 - upgrade libcoro, resulting in pthread-backend (which was only created
796 to fulfill the rules of the programming languages shootout).
797
7984.45 Thu Mar 13 11:55:36 CET 2008
799 - fix a file leaking bug in eg/mhyttpd that would allow
800 downloading of any file (reported by oesi).
801 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
802 (also add testcase).
803 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
804 apparently having loads of fun with that).
805 - implement $state->swap_defsv, ->swap_defav and document ->throw.
806
8074.4 Wed Feb 13 16:44:29 CET 2008
808 - only assume .cfi pseudo isns exist on GNU/Linux.
809 - add get_timed method to Coro::Channels.
810 - fixed Coro::Debug manpage.
811 - perl 5.11 compatbility improvement based on patch by
812 Andreas König.
813
8144.37 Sun Jan 20 11:25:23 CET 2008
815 - soften the check for an active parser for perl 5.10 (avoiding
816 "Coro::State::transfer called while parsing" in many cases).
817
8184.36 Sun Jan 13 10:53:56 CET 2008
819 - reset diehook when terminating from an async_pool as to not
820 trigger any __DIE__ handlers.
821
8224.35 Sun Jan 13 04:14:13 CET 2008
823 - "bt" debug command now displays any exceptions
824 from longmess and also skips the innermost
825 stackframes, giving more useufl output.
826 - allow backtraces in coroutines blocked in Coro::EV,
827 at a <1% speed hit.
828 - handle localising of $SIG{__WARN__} and __DIE__
829 properly (with a proper amount of dirty hacking).
830
8314.34 Sat Dec 22 17:49:53 CET 2007
832 - upgrade to EV version 2.0 API.
833
8344.33 Mon Dec 17 08:36:12 CET 2007
835 - make Coro::AIO etc. loadable in the absence of EV.
836
8374.32 Mon Dec 17 07:46:02 CET 2007
838 - majorly improved Coro::Handle's performance with Coro::EV.
839 - implemented the until now mythical Coro::BDB module.
840 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
841
8424.31 Wed Dec 5 12:32:39 CET 2007
843 - remove warn statement form Coro::Util.
844
8454.3 Tue Dec 4 20:33:14 CET 2007
846 - calls to the idle function could cause stack corruption
847 when the stack changed.
848 - do no longer rely on the presence of EV::DNS if EV is used
849 (because it is gone), but instead take avdantage of EV::ADNS
850 if available.
851 - add ($) prototypes to all functions in Coro::Storable.
852 - use a conventional (safer) idle callback in Coro::EV.
853 - do accept quickly in Coro::Debug to avoid endless loops.
854
8554.22 Fri Nov 30 16:04:04 CET 2007
856 - really use optimised versions for Event and EV in Coro::Util
857 and Coro::Handle.
858
8594.21 Sun Nov 25 10:48:59 CET 2007
860 - fix a spurious memory read.
861 - Coro::EV no longer keeps the eventloop "alive".
862
8634.2 Fri Nov 9 20:47:05 CET 2007
864 - enable/disable tracing from a new coroutine, not a pooled one.
865 - fix a memleak in Coro::Event.
866 - removed killall call from fork_eval.
867 - made sure store_fd is already loaded so that fork_eval
868 does not have to parse autoload in each subprocess.
869 - only use assembly method if -O switch is in $Config{optimize}.
870 - add (optional) Coro::EV module, so far the best event loop module
871 directly supported by Coro.
872 - if the event model is EV, use EV::DNS to resolve
873 stuff in Coro::Util.
874 - don't get confused by multiple event notifications in Coro::Handle.
875 - initial support for EV (libevent interface).
876 - require Event and EV using configure_requires, to force their existance.
877
8784.13 Wed Oct 24 07:26:45 CEST 2007
879 - add Coro::Storable::blocking_thaw.
880 - use a vastly more complicated technique to localise
881 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
882 - use a coroutine for the idle callback Coro::Event,
883 instead of running Event in the current coroutine context.
884 This also catches recursive invocations.
885 - actually report fork errors in gethostbyname and inet_aton.
886
8874.11 Thu Oct 11 02:40:24 CEST 2007
888 - port to threaded perls.
889
8904.1 Thu Oct 11 02:38:16 CEST 2007
891 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
892 be local to each coro (see Coro::State).
893 - incompatible change: for very deep reasons, cede and cede_notself
894 cannot return anything, so nothing will be returned.
895 - possibly bring back 5.10 compatibility (untested).
896 - work around stupid (and wrong) warning on 5.10 :(.
897 - overlay the saved state over the context stack. This saves
898 a few hundred bytes per coroutine on average and also
899 speeds up context switching a bit.
900 - further tune default stack sizes.
901 - (more) correctly calculate stack usage in coro_rss.
902 - Coro::Storable::blocking_* did not properly lock
903 resulting in races between coroutines.
904 - added Coro::Storable::guard.
905 - stopping to trace a coroutine could destroy the cctx of
906 an unrelated coroutine.
907 - explain the relationship between Perl and C coroutines in
908 more detail in Coro::State.
909 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
910 causing a fork per resolve. This also affected Coro::Socket.
911 - switch to a separate stack in $coro->call/eval to avoid
912 invalidating pointers.
913
9144.03 Sat Oct 6 21:24:00 CEST 2007
915 - added Coro::throw method.
916 - minor code cleanups.
917
9184.02 Sat Oct 6 02:36:47 CEST 2007
919 - fix a very minor per-coroutine memleak (a single codereference).
920 - fixed a bug where the currently in-use c context would be freed
921 prematurely (can happen only when programs change the stacksize
922 or use tracing).
923 - tracing can no longer keep a coro alive after it terminated.
924 - do static branch prediction in the common path for gcc. gives
925 about 2-5% speed improvement here.
926
9274.01 Fri Oct 5 22:10:49 CEST 2007
928 - instead of recreating *a* standard output handle we simply
929 use STDOUT, which is faster and hopefully more robust.
930
9314.0 Fri Oct 5 12:56:00 CEST 2007
932 - incompatibly changed Coro::Storable::freeze.
933 - major new feature: added Coro::Debug, for interactive coroutine
934 debugging, tracing and much more.
935 - major bug fix: unbelievable, but true: $_, $/ and many other
936 "saved" variables actually weren't being saved. This has been fixed,
937 of course, while increasing performance while losing all the save
938 flags.
939 - save flags are gone, and all the api functions dealing with them.
940 - added Coro::Semaphore::adjust.
941 - added Coro::Util::fork_eval.
942 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
943 - added Coro::killall.
944 - reduce initial stack sizes to allow for "micro-coroutines".
945 - better async_pool resource management, moved parts of async_pool
946 handling to XS (major speed improvement).
947 - actually croak before modifying important data structures.
948 - refuse to transfer while compiling.
949 - possibly support eval EXPR better now.
950 - enable assembly per default on linux+bsd x86+amd64.
951 - all internal members were renamed _something for easier subclassing.
952 - many minor tweaks.
953
9543.63 Wed May 16 14:10:06 CEST 2007
955 - implement handcoded assembly for x86/amd64 SVR ABI.
956
9573.62 Fri Apr 27 21:36:06 CEST 2007
958 - upgrade libcoro (which might set unwind info correctly).
959 - change default on linux to setjmp/longjmp.
960
9613.61 Thu Apr 19 12:36:18 CEST 2007
962 - Coro::Storable caused an endless loop when thawing invalid
963 pst's sometimes.
964 - use a Semaphore in Coro::Storable, as Storable doesn't
965 seem to be reentrant (although it is documented to
966 be threadsafe...).
967 - fix Coro::Signal to bring back the original unreliable
968 but stateful semantics.
969 - fixed a lot of typos in Coro.pm (patch submitted by David
970 Steinbrunner, which applied flawlessly).
971
9723.6 Sat Apr 14 17:13:31 CEST 2007
973 - added some bugfixes to get eg/myhttpd working again.
974 - added Coro::Storable for often-cede'ing freeze/thaw.
975 - try to do a clean exit when a coroutine calls exit
976 (EXPERIMENTAL).
977 - got rid of indirect call through _coro_init.
978 - updated the partly antique examples in eg/ to
979 work again and be a bit less magic, too.
980 - fixed Coro::Signal semantics to work as documented again.
981
9823.55 Sun Mar 25 01:20:47 CET 2007
983 - add SAVE_DEFFH to save the default filehandle and enable
984 it by default.
985 - finally move socket-operations from Coro::Socket to Coro::Handle
986 to be able to unblock foreign sockets.
987 - add Coro::State::save_also and guarded_save.
988 - add count accessor to Coro::Semaphore.
989 - add Coro::State::cctx_stacksize.
990 - just for the fun of it, do not rely on implicit context,
991 which can dramatically improve performance, but people
992 using windows-process-emulation perls do not care much about
993 performance.
994
9953.51 Sun Mar 4 14:18:04 CET 2007
996 - fixed a problem when you weakened references to Coro::State's
997 (patch by Yuval Kogman).
998
9993.501 Wed Feb 28 12:44:07 CET 2007
1000 - rename some global symbols as macosx from hell redefines
1001 them without asking.
1002
10033.5 Tue Feb 13 20:22:53 CET 2007
1004 - do AnyEvent model detection earlier, avoiding problems
1005 caused by first using AnyEvent and later Coro::Event.
1006 - implement and document Coro::Event event objects.
1007 - fix a potential problem in Coro::Event causing crashes.
1008 - initialise PL_comppad when creating a new coroutine,
1009 avoids crashes on early coro destruction.
1010
10113.41 Mon Jan 22 19:19:49 CET 2007
1012 - readline on Coro::Handle did not support undefined $/,
1013 nor did it deliver partial lines on EOF or error.
1014 - implement malloc fallback for stack allocation because
1015 stupid broken idiotic OSX has a stupid broken
1016 idiotic fits-the-whole-os mmap "implementation" and
1017 my dick feels longer if Coro is portable even to
1018 obsolete platforms.
1019
10203.4 Fri Jan 19 21:52:54 CET 2007
1021 - remove t/09_timer.t, as it isn't really testing much
1022 but was rather flaky in practise.
1023 - async_pool coro would keep arguments and callback alive until
1024 it was reused.
1025 - cancellation of a coroutine could cause spurious idle calls
1026 in cede_notself.
1027
10283.3 Sat Jan 6 03:45:00 CET 2007
1029 - implement $coro->on_destroy.
1030 - Coro::Event blocking semantics have been changed,
1031 documented and - hopefully - improved.
1032 - fix nice adding, not subtracting, from priority.
1033 - fix ->prio and api_is_ready (patch by Mark Hinds).
1034 - fixed an assert ("... == prev__cctx->idle_te")
1035 that could errronously trigger.
1036 - fix various large and small memleaks.
1037 - use a (hopefully) more stable cancel implementation
1038 that immediately frees the coroutine data.
1039 - cede/cede_notself return a status now.
1040 - added Coro::guard function.
1041 - added a global coroutine pool for jobs (on my machine,
1042 I can create and execute 48k simple coros/s with async,
1043 and 128k coros with async_pool).
1044 - Coro::AIO now uses the coroutine priority as io priority.
1045
10463.2 Fri Dec 22 05:07:09 CET 2006
1047 - improve portability to slightly older perls.
1048 - use cleaner coroutine destruction.
1049 - simplify configuration for users.
1050 - optionally (unrecommended) prefer perl functions over
1051 their coro replacements.
1052
10533.11 Tue Dec 5 13:11:24 CET 2006
1054 - fixed some bogus assert's, but as perl.h disables assert even
1055 without NDEBUG (thank you very much), not too many people should
1056 notice (that did include myself). Andreas König noticed, though :)
1057 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
1058 it. Also noticed by Andreas König :)
1059 - save/restore tainted status.
1060 - verified to pass the testsuite on my 5.9.5.
1061
10623.1 Mon Dec 4 23:03:40 CET 2006
1063 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
1064 - incompatible change: transfer flags are now per-state.
1065 - give a better error message on deadlock.
1066 - document Coro::nready.
1067 - enhanced testsuite.
1068
10693.01 Sun Dec 3 23:47:42 CET 2006
1070 - forgot to include Coro::Timer.
1071
10723.0 Sun Dec 3 22:57:25 CET 2006
1073 - the "FINALLY COMPLETELY STABLE" release (coming soon:
1074 the "FAMOUS LAST WORDS" release).
1075 - implement a new stack sharing algorithm, which uses a stack
1076 pool (size currently hardcoded).
1077 - make stack sharing mandatory (it no longer uses a heuristic).
1078 - eval/die no longer cause weird problems under heavy use.
1079 - Coro::Event could cause livelocks if it was used but
1080 no Coro::Event watchers were used.
1081 - Coro::Event now uses asynccheck as prepare does not
1082 check for changed watchers.
1083 - Coro::Event allows multiple waiting coros and will wake up one
1084 per event.
1085 - Coro::Event should be cleaner and more efficient now.
1086 - new utility function Coro::unblock_sub.
1087 - document the sad fact that Event is no longer reentrant.
1088 - putting a coroutine into the ready queue twice could under
1089 some circumstances lead to stack corruption.
1090 - minor incompatibility: subclassing Coro::State is supported
1091 directly now without going through a _coro_state member.
1092 - state/coro switching is much faster now.
1093 - very minor optimisations and code/documentation cleanup.
1094 - avoid problems due to compiler inlining.
1095 - removed timers from Coro::Timer -> use AnyEvent instead.
1096 - replaced Coro::idle coroutine by (cleaner) idle handler.
1097 - updated to newest libcoro.
1098 - implement enhanced support for valgrind.
1099 - implement is_ready and return value for ready.
1100 - removed Coro::Cont, it was a misnomer (it's generators),
1101 and it was somewhat annoying to get it right. Will come back
1102 if somebody asks for it :)
1103 - many ->wait methods and Coro::Event could return
1104 spuriously without the event having happened.
1105
11062.5 Tue Nov 7 12:22:33 CET 2006
1107 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
1108 use AnyEvent, moved them to Coro/.
1109 - added Coro::LWP which contains all the uglyness required to
1110 make LWP non-blocking.
1111 - should work with perl 5.9.x now (Andreas König made me do it).
1112 - fixed another bug in Coro::Select when the timeout was undef.
1113 - reuse PL_start_env for all coros, saves some memory per coroutine.
1114 - manage PL_top_env differently, hopefully to avoid panic: top_env.
1115 - timeout argument was not properly used in Coro::Socket.
1116 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
1117 - emulate undocumented(!) functionality of IO::Socket required
1118 by LWP(!!).
1119 - updated eg/lwp to work with newer lwp's.
1120 - remove "FATAL: uncaught exception" prefix. Coroutines that die
1121 kill the whole process, just as exceptions in the main "coroutine"
1122 did already.
1123
11242.1 Wed Nov 1 23:01:13 CET 2006
1125 - fix a long-standing bug in Coro::Select where select with
1126 zero timeout would instead change the current default filehandle.
1127 - use a simpler and hopefully more robust way to clone padlists
1128 (uses less memory and a perl function instead of our own).
1129 - coro can now create a stack guard area on many architectures.
1130 - Coro::AIO properly reexports additional functions from IO::AIO.
1131 - updated libcoro with a workaround for OS X,
1132 pach and testing by Michael Schwern.
1133
11342.0 Tue Oct 24 05:47:17 CEST 2006
1135 - support additional aio requests in Coro::AIO.
1136
11371.9 (never properly released due to a glitch)
1138
11391.8 Thu Feb 2 00:59:06 CET 2006
1140 - applied suggested patch by SAMV to avoid problems during stupid
1141 mark & sweep gc run.
1142 - applied patch by Scott Walters for 5.9.3 compatibility.
1143
11441.7 Tue Dec 27 01:41:58 CET 2005
1145 - added Coro::AIO, a thin wrapper around IO::AIO.
1146 - improved Makefile.PL explanations.
1147
11481.6 Mon Dec 12 21:30:05 CET 2005
1149 - additionally save PM_curpm might fix as-of-yet
1150 unseen problems with regex matches being attributed to the
1151 wrong package.
1152 - add t/10_bugs.t, which currently checks against imho broken
1153 perls that use 0.26MB of stack space per Perl_magic_get
1154 invocation (newer linuxes) as opposed to the 0.0002MB perl
1155 normally uses.
1156 - make stacksize configurable for the ultimate debian experience.
1157
11581.51 Mon Dec 12 18:48:36 CET 2005
1159 - remove debugging warn() accidentally left in Coro::Select.
1160
11611.5 Tue Nov 29 13:32:44 CET 2005
1162 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
1163 - fix Coro::Select 'select'.
1164 - strict'ify some modules.
1165 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
1166 - use XSLoader in selected modules.
1167 - remove some 5.6 compatibility cruft.
1168
11691.4 Tue Sep 6 00:11:05 CEST 2005
1170 - libcoro did not take into account the trampoline on amd64,
1171 when 'l'inux method was used.
1172
11731.31 Tue Aug 30 23:31:33 CEST 2005
1174 - some portability fixes/workarounds.
1175
11761.3 Sat Aug 20 03:08:56 CEST 2005
1177 - no code changes. module seems to work fine.
1178 - improve Coro::State docs, remove reference to nonexisting
1179 Coro::State::flush method.
1180 - no longer autodetect windows, present it as an option instead.
1181
11821.2 Mon May 16 02:00:55 CEST 2005
1183 - included libcoro.c earlier in State.xs, might improve portability.
1184 - use a faster and possibly more stable Coro::Cont implementation.
1185 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
1186 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
1187 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
1188
11891.11 Thu Mar 3 18:00:52 CET 2005
1190 - change of contact address.
1191
11921.1 Tue Feb 22 20:51:16 CET 2005
1193 - support [l]inux method on amd64.
1194 - allow some context switches while compiling/eval'ing.
2 1195
31.0 Tue Aug 10 03:47:27 CEST 2004 11961.0 Tue Aug 10 03:47:27 CEST 2004
4 - Event 0.89 fixes the bug again. 1197 - Event 0.89 fixes the bug again.
5 - fix the remaining(??) memleak. I didn't free the cache-AV for 1198 - fix the remaining(??) memleak. I didn't free the cache-AV for
6 padlists, so one AV per created anon-sub leaked. 1199 padlists, so one AV per created anon-sub leaked.
7 1200
80.97 Fri May 14 15:23:32 CEST 2004 12010.97 Fri May 14 15:23:32 CEST 2004
9 - remove stupid gcc extension that was an experiment only anyway. 1202 - remove stupid gcc extension that was an experiment only anyway.
10 1203
110.96 Thu May 13 18:09:29 CEST 2004 12040.96 Thu May 13 18:09:29 CEST 2004
12 - tell people if they have yet anothe rbroken Event variant. 1205 - tell people if they have yet another broken Event variant.
13 - made Coro to work _better_ with threads (it seems that, as long 1206 - made Coro to work _better_ with threads (it seems that, as long
14 as you use Coros in one thread only everything will work fine. 1207 as you use Coros in one thread only everything will work fine.
15 Looking at the perl source the problem might be missing locking 1208 Looking at the perl source the problem might be missing locking
16 between Coro and some perl internal routines. That's not easily 1209 between Coro and some perl internal routines. That's not easily
17 fixable). 1210 fixable).

Diff Legend

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