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

Comparing Coro/Changes (file contents):
Revision 1.333 by root, Sun Apr 6 19:23:49 2008 UTC vs.
Revision 1.674 by root, Tue Aug 14 15:46:03 2018 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_cancle failure (vebrose mode? show how to make backtrace?)
9TODO: scalar context for Coro::AIO wrappers?
10
11TODO: perl -e 'use Coro; my $a = async {}; cede; $Coro::main->call(sub { warn 1 });' # works
12TODO: perl -e 'use Coro; $Coro::main->call(sub { warn 1 });' # doesn't work
13
14 - libcoro update: allow sharing of coro threads using the pthreads
15 backend among different pthreads, as is required by Coro::Multicore.
16 - minor code cleanups.
17
186.514 Thu Aug 31 18:25:31 CEST 2017
19 - safe_cancel no longer croaks when called on already-destroyed
20 thread, but instead "just works".
21
226.513 Sat Jul 15 01:18:51 CEST 2017
23 - experimental 5.22 comaptibility changes (PadlistNAMES).
24 - removed attributes example and made sure the other examples
25 still work.
26 - fix some statevar types for newer perls, should be of no consequence.
27
286.512 Fri Jul 14 05:03:17 CEST 2017
29 - seems to pass test suite down till 5.8.9, so make this the
30 minimum requirement in Makefile.PL.
31 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
32 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
33 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
34 fortified longjmp variants in some libcs.
35 - re-try JIT allocation without PROT_EXEC, to hopefully improve
36 portability (still doesn't seem to work around broken selinux
37 mmap everywhere).
38 - speling fixes (Xavier Guimard).
39
406.511 Sun Jun 26 23:44:50 CEST 2016
41 - make it compile with threaded perls again.
42 - simplify and speed up __DIE__ and __WARN__ handling.
43
446.51 Sat Jun 25 20:57:20 CEST 2016
45 - 6.5 release didn't compile on most older perls due to
46 a logic error enabling experimental 5.24 code for all versions.
47 - since various broken or deficient patches float around
48 to work around the vtbl API breakage in perl, let's provide
49 our own workaround, which seems to be much less invasive
50 and more compatible than the approaches seen so far. YMMV.
51 - work around assertion failure in perl_destruct on 5.24
52 (analyzed by Dave Mitchell).
53
546.5 Wed Jun 22 22:23:50 CEST 2016
55 - swap_sv swapping order was not symmetrical, causing
56 wrong swaps when swap_sv was used multiple times on the
57 same sv in the same thread.
58 - swap_sv calls can now be undone by calling it again
59 with the same variables.
60 - swap_sv calls will now be undone in async_pool threads.
61 - split Coro::Semaphore::up/adjust into separate xs functions
62 for better error reporting, at slight codesize increase.
63 - (libcoro) arm assembly support, please test and report.
64 - adjust to PL_savestack changes in perl 5.24 (adapted from
65 the debian patch, which unfortunately gets it wrong).
66
676.49 Sat Oct 17 01:40:12 CEST 2015
68 - throwing an exception to a thread waiting in
69 Coro::Handle using Coro::EV did not stop the watchers,
70 causing the next call to fail (testcase by Martin Pritchard).
71 - bump minimum perl version to 5.10.
72
736.48 Sun Oct 4 19:03:51 CEST 2015
74 - fix memory leak when Coro::Handle uses Coro::EV internally
75 (testcase by Sten Sten).
76 - update code to libev 4 API (internally, Coro still used the
77 version 3 API calls).
78 - remove unused hv_sig.
79
806.47 Sat Jul 11 03:58:20 CEST 2015
81 - swap_svs - NVs can be in the SV head beginning with 5.20.
82
836.46 Tue Jun 30 14:36:30 CEST 2015
84 - restore portability to perls with windows fork emulation
85 (patch by Petr Písař).
86
876.45 Tue Jun 30 01:40:08 CEST 2015
88 - restore portability to perl 5.16 and below.
89
906.44 Tue Jun 30 00:41:54 CEST 2015
91 - avoid segfaulting (or worse) when tracing a canceled thread.
92 - the Event module silently broke it's public hook API, causing
93 failures on perls compiled with -Duselongdouble. Requiring
94 the latest version of Event because I don't know which release
95 changed this (it's not mentioned in the ChangeLog).
96 - allow xs level enterleave hooks via CoroAPI.
97
986.43 Thu Jun 4 15:38:14 CEST 2015
99 - use stability canary.
100 - port to stableperl-5.22.0-1.001.
101 - update libecb.
102
1036.42 Wed Feb 11 20:29:52 CET 2015
104 - Coro::SemaphoreSet->try did not actually work (analyzed by
105 SATO Kentaro).
106 - upgrade libecb to be C11 compliant.
107
1086.41 Sat Sep 6 22:08:46 CEST 2014
109 - restore portability to perl 5.8.x.
110 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
111 up some memory, but fixes "use feature" and similar modules
112 when used inside a Coro.
113 - allow zero as argument to Coro::Channel to mean the same thing
114 as no argument. this works with older versions as well,
115 but wasn't legal until now.
116 - slightly better c header file detection.
117
1186.39 Mon Jun 2 00:00:08 CEST 2014
119 - work around more incompatible changes in 5.20.
120
1216.38 Sun Jun 1 21:54:23 CEST 2014
122 - check that perl slots actually have enough space to hold
123 interpreter variables.
124 - untested port to perl 5.19 (64 bit tmps indices) (reported
125 by Andreas König).
126 - croak when cancel is called without a thread context.
127
1286.37 Tue Mar 4 13:27:33 CET 2014
129 - *sigh*, removed leftover debugging code from debugging a
130 perl bug, of all things.
131
1326.36 Tue Mar 4 07:11:59 CET 2014
133 - semaphores would not clear the destroy hook when interrupted
134 by ->throw, causing segfaults or worse.
135 - ->throw on a thread waiting for a semaphore did not acquire
136 the semaphore, but also didn't wake up other waiters,
137 possibly causing a deadlock.
138 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
139 to report a full stacktrace, which should help find out
140 where the actual call is.
141 - "atomically" destroy data in slf_destroy, because it is
142 the right thing to do, just in case.
143 - disable ecb.h libm dependency, to compile on ancient systems
144 or under adverse conditions.
145
1466.33 Mon Nov 18 11:26:27 CET 2013
147 - do not crash when freeing padlists with holes (in 5.18).
148 - tentative SVt_BIND 5.19 port/fix.
149
1506.32 Tue Nov 5 15:35:35 CET 2013
151 - use a new algorithm to derive padlists for perl 5.18. The old
152 one could lead to 0-pointer accesses inside perl (reported
153 by Darin McBride).
154
1556.31 Thu May 9 07:39:48 CEST 2013
156 - Coro::AIO requests would crash if the thread was ready'd
157 while the request was ongoing.
158
1596.29 Wed May 8 02:55:18 CEST 2013
160 - when an on_destroy handler destructs the coro currently being
161 destructed a perl scalar could be accessed after being freed,
162 likely causing a crash.
163
1646.28 Wed Mar 6 06:58:02 CET 2013
165 - clean remnants of existing __DIE__ and __WARN__ handlers so
166 they lose their magic and will not cause segfaults later
167 (testcase by Andrey Sagulin).
168 - improved Coro::State documentation a bit.
169 - Coro::Debug::command now flushes the output.
170 - add hack detection code for x32 abi, because the braindead slugs
171 who designed that made it look exactly like x86_64 without
172 providing proper compile time symbols to test for it. as a result,
173 this detection cannot work reliably.
174 - valgrind stack registering was broken.
175 - do not rely on Time::HiRes anymore in Coro::Debug.
176
1776.23 Fri Dec 7 23:36:37 CET 2012
178 - use experimental fiber implementation on native windows
179 perls.
180 - use sizeof (void *) as multiplication factor for stack sizes,
181 to accomodate the totally braindamaged microsoft 64 bit "os".
182 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
183 activeperl 5.16 crashes when PerlIO_define_layer is called due
184 to some bug in the perl dll, strawberry perl at least passes
185 the testsuite.
186 - implement Coro::Handle->peeraddr/host/port, for slightly
187 improved compatibility with LWP.
188 - implement 5.17 compatibility by almost blindly applying a
189 good-looking patch by Father Chrysostomos.
190 - move stack management functions into libcoro 3.
191 - libcoro version 3 "released".
192 - support magic values as timed_io_once args.
193 - recommend AnyEvent 7+ or EV 4+, also require EV
194 version 4 or newer for Coro::EV.
195
1966.10 Tue Oct 9 01:14:27 CEST 2012
197 - updated ecb.h, it had a typo that caused it to not compile on many
198 big endian systems (reported by many people).
199 - disable memory fences in ecb.h to improve portability.
200
2016.09 Sat Oct 6 23:25:02 CEST 2012
202 - Coro::EV I/O watchers were not interruptible by exceptions
203 (Coro::State::throw) (testcase by sten).
204 - ->throw now puts threads into the ready queue, as this seems to
205 be expected by existing code, and code that doesn't cope with spurious
206 wakeups needs fixing anyway.
207 - use fd -1 in mmap.
208 - cast I32 to int in error message printf.
209 - warn about broken so-called "hardened" kernels.
210
2116.08 Fri Apr 13 12:05:47 CEST 2012
212 - be more aggressive about exiting like perl does - formerly,
213 exiting from the non-main thread would not execute END blocks.
214
2156.07 Fri Nov 11 21:21:48 CET 2011
216 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
217 - update ecb.h.
218
2196.06 Mon Aug 8 23:59:48 CEST 2011
220 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
221 backend, which is an order of magnitude slower. unfortunately, cygwins
222 pthread implementation isn't very complete either, so allocate the stack
223 twice just to be sure.
224 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
225 in its ithreaded perl - best recompile cygwin and use the 'w'indows
226 backend for much better performance. also disable ithreads for
227 even better performance...).
228
2296.05 Thu Aug 4 21:36:36 CEST 2011
230 - blush, condvar values would not be propagated from send to recv anymore
231 (reported by Chip Salzenberg).
232 - use exponential increase for the readline buffer length in
233 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
234
2356.04 Wed Aug 3 17:15:45 CEST 2011
236 - use even more efficient and more compatible condvars for
237 compatibility to AnyEvent 6.x :)
238 - more inconsequential ecb.h updates.
239
2406.03 Wed Aug 3 11:41:30 CEST 2011
241 - change how Coro patches AnyEvent condvars for compatibility to
242 AnyEvent 6.x.
243 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
244 hasn't been included.
245
2466.02 Wed Jul 13 04:35:19 CEST 2011
247 - "improve portability to Gentoo" - gentoo manages to put perl variables
248 in memory areas that are farther than 2gb apart, which the jit couldn't
249 handle and barfed. now it's just a bit slower on gentoo and similar
250 systems.
251
2526.01 Sun Jul 3 12:31:14 CEST 2011
253 - workarounds are good, but the test for whether pthreads are used
254 was not good. this one should be better.
255 - check differently whether gcc generates cfi instructions itself.
256
2576.0 Wed Jun 29 19:43:35 CEST 2011
258 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
259 destroyed and cleaned up automatically (e.g. async { schedule }).
260 - implement a JIT compiler for part of the thread switch code,
261 which gives a 50% speed improvement on threaded perls, and
262 about 4% on non-threaded perls (so threaded perls now finally
263 reach about half the speed of non-threaded perls).
264 - slightly modernise Coro::Intro, add section about rouse functions.
265 - avoid DEFSV and ERRSV, giving another 10% improvement
266 in thread switching.
267 - Coro::State->is_destroyed is now called is_zombie.
268 - implement a Coro->safe_cancel method that might fail, but
269 cancels in a "safer" way if it succeeds.
270 - add preliminary support for DEBUGGING perls.
271 - get rid of two hash-accesses when initialising a new Coro - this
272 speeds up coro creation by almost a factor of two.
273 - croak when a coro that is being cancelled tries to block
274 (e.g. while executing a guard block), instead of crashing or
275 deadlocking.
276 - use a more robust and also faster method to identify Coro::State
277 objects - speeds up everything a bit.
278 - implement Coro->cancel in XS for a 20% speed improvement, and to
279 be able to implement mutual cancellation.
280 - speed up context switches by a percent or two by more efficiently
281 allocating context stack entries.
282 - implement Coro->join and Coro->on_destroy in XS for a speedup and
283 a reduction in memory use.
284 - cancelling a coro while it itself is cancelling another coro is
285 now supported and working, instead of triggering an assertion.
286 - be a bit more crash-resistant when calling (buggy) on_destroy
287 callbacks (best effort).
288 - move on_destroy into the slf_frame, to allow extension slf
289 functions to have destructors.
290 - get rid if coro refcounting - simply crash in other interpreter
291 threads by nulling the pointers on clone.
292 - simplify warn/die hook handling when loading Coro - the convoluted
293 logic seems to be no longer neccessary.
294 - use libecb instead of our own home-grown gcc hacks.
295 - document alternatives to Coro::LWP. Please use them :)
296 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
297 sigaltstack. Really. wine suffers from it, erlang suffers from it,
298 and it's known since at least 2006.
299
3005.372 Wed Feb 23 06:14:30 CET 2011
301 - apparently mingw doesn't provide a working gettimeofday, try to
302 work around that by relying on Time::HiRes (indirectly brought to
303 my attention by Max Maischein).
304 - fix some portability issues when Time::HiRes was used.
305
3065.371 Mon Feb 21 14:36:08 CET 2011
307 - backport to windows process emulation code again.
308
3095.37 Sat Feb 19 07:49:44 CET 2011
310 - add a big "Coro thread life cycle" section to "man Coro".
311 - try a tentative workaround against the breakage that 5.13 has
312 introduced without depreciation period. sigh.
313 - no longer use Time::HiRes if gettimeofday is available, which
314 saves quite a lot of memory.
315
3165.36 Sun Feb 13 05:33:41 CET 2011
317 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
318 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
319 method call.
320 - undocument Coro::Timer::sleep and obsolete whole module.
321 - optimise Coro::Timer::timeout memory and cpu usage.
322 - slightly updated Coro::Intro for recent changes.
323 - do not initialise PL_dirty anymore.
324
3255.25 Thu Nov 11 01:08:39 CET 2010
326 - try a different approach on netbsd - netbsd 5 finally has marginally
327 working pthreads, but still broken ucontext/sigaltstack.
328 - openbsd 4.8 finally got their act together, Coro works out of the box
329 with asm, setjmp and pthreads (no change, just informational).
330
3315.24 Sat Oct 23 11:27:12 CEST 2010
332 - port to the EV 4.0 API.
333 - work around bugs in mingw32, making strawberry perl work
334 out of the box.
335 - correctly modify Coro::AIO function prototypes
336 so that they reflect the "no optional parameters" rule.
337 - "ported" libcoro to C++.
338
3395.23 Mon May 17 18:50:42 CEST 2010
340 - be more resistant to ordering changes when initialising
341 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
342 Waldorf).
343 - document that perl 5.12 deliberately removed support for cloning.
344
3455.22 Wed Apr 14 03:55:35 CEST 2010
346 - correctly return udnef on errors in Coro::Handle::read/write
347 (testcase by Marc Mims).
348 - convert Coro::Util into a "perl compatibility wrapper" - the functions
349 are less useful now, but are drop-in replacements for existing
350 functions, listing better alternatives in the documentation. This also
351 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
352 with Coro::Util::inet_aton.
353 - do not override $Coro::idle unconditionally in Coro.pm, as other
354 modules could have provided their own idle coro already
355 (for exmaple, Coro::AnyEvent).
356 - fix Coro::Util::gethost* functions.
357 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
358
3595.21 Wed Dec 16 07:19:51 CET 2009
360 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
361 together.
362 - add some examples on how to combine other event loops with Coro in
363 Coro::AnyEvent, and how to run it (and not to block). Seems to be
364 the most common source of confusion.
365 - try to catch people naively blocking in an event callback.
366 - work around the perl filehandle bug issue in conjunction with
367 older common::sense (as indirectly pointed out by ZSystem).
368 - clarify the "not from signal handlers" section.
369
3705.2 Sun Oct 4 14:54:24 CEST 2009
371 - Coro::Storable destroyed the prototypes of the functions it wrapped.
372 - export rouse_cb and rouse_wait by default now.
373 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
374 - new method $state->swap_sv.
375 - added section on "windows process emulation" to the manpage,
376 after a not-so-fruitful (nor-friendly) "discussion" with chip
377 salzenberg (discussion implies arguments, but his only arguments
378 were ad-hominems, one wonders why he started it in the first
379 place). I hope this explains it well enough for him to understand,
380 and maybe well enough for others to understand.
381 - use common::sense everywhere now.
382 - idle callbacks are no longer supported, use idle coros instead.
383 - print a thread listing when a deadlock is detected.
384
3855.17 Sat Aug 22 23:09:31 CEST 2009
386 - work around a bug in the perl debugger causing crashes
387 when running under the debugger by marking _pool_handler as nodebug.
388 - speed up Coro::async considerably.
389 - try some hacks to get netbsd to work "more often" - their broken
390 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
391 - convert Coro to AE 5.0 API.
392
3935.162 Tue Jul 28 04:04:03 CEST 2009
394 - perl 5.8.2 is now minimum requirement.
395 - skip t/19_handle.t on broken windows perls.
396
3975.161 Wed Jul 22 04:47:38 CEST 2009
398 - Coro::AnyEvent::poll could have a different prototype when EV was
399 used as backend (analyzed by Tatsuhiko Miyagawa).
400 - Coro::AnyEvent errornously initialised the event loop when loaded,
401 not on demand.
402 - try to workaround rare */t/01_unblock.t failures.
403
4045.16 Tue Jul 21 01:44:37 CEST 2009
405 - Coro::AnyEvent failed to hook into the event loop
406 when no threads had been readied between detecting
407 the event loop and actually running it.
408 - considerably speed up Coro::Select by taking avdantage
409 of AnyEvent > 4.8 and some other optimisations.
410 - implement paragraph readline mode in Coro::Handle
411 (based on patches by Zsbán Ambrus).
412 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
413 and analyzed by Yasuhiro MATSUMOTO).
414 - clarified libcoro license and copyright.
415 - someone stole my EXTRA_META!!!
416 - implement Coro::Select::patch_pp_sselect and it's brother,
417 for hardcode select overriding.
418
4195.151 Mon Jul 6 05:41:57 CEST 2009
420 - backport to windows process emulation code again (patch by
421 Yasuhiro MATSUMOTO).
422 - slightly update Coro::MakeMaker.
423
4245.15 Tue Jun 30 10:28:06 CEST 2009
425 - deprecate Coro::Socket, document how to get ipv6 support via
426 AnyEvent::Socket instead.
427 - implement signal->wait ($cb) interface, similar to semaphores.
428 - work around SvOK not supporting getmagic, so we have to getmagic
429 to test for undef :/ (reported by Matthias Waldorf).
430 - load Coro::AnyEvent in all modules using AnyEvent.
431 - work around perl corrupting our internal data structures,
432 reported by Tokuhiro Matsuno.
433 - enable per-coroutine real and cpu time gathering
434 (Coro::State::enable_times).
435
4365.14 Wed Jun 24 01:37:48 CEST 2009
437 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
438 as this seems to have been a difficult concept (poll, sleep, idle,
439 idle_upto).
440 - add Coro::AnyEvent::readable/writable functions.
441 - clarify Coro::EV/Event/AnyEvent manpages.
442 - free per-thread global scalars in the thread calling ->cancel, to
443 avoid crashes when $_, $@ etc., are magical but some of those
444 had already been freed.
445 - "unexperimentalise" the callback interface for Coro::Semaphore.
446 - speed up ready queue management/context switching by using a linked
447 list instead of an array (~5-10%).
448 - implement "watch" command in Coro::Debug shells.
449 - for fun, implement time-slicing as an example in the manpage.
450 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
451 Coro::Event, use the normal AnyEvent handling instead of dieing
452 (the same is true for Coro::Handle).
453 - properly document Coro::EV::timed_io_once.
454 - avoid unneccessary ->cancel calls in Coro::Handle.
455 - maybe make it work on mingw32 with win32 backend
456 (based on patch by Yasuhiro Matsumoto).
457
4585.132 Fri May 29 09:00:39 CEST 2009
459 - do not keep a reference to the argument itself in
460 Coro::Semaphore::guard, as it could change later.
461 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
462 should support a prepare callback.
463
4645.131 Mon Mar 16 23:20:37 CET 2009
465 - implement and document Coro->suspend, Coro->resume.
466 - fix Coro::Select implementation to not (often) close
467 the passed file descriptors (testcase provided by pippijn).
468
4695.13 Mon Dec 15 21:51:42 CET 2008
470 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
471 - don't set diehook to C<undef> but instead to NULL, to avoid
472 spurious warnings.
473 - fix a lot of bugs in Coro::SemaphoreSet.
474 - Coro::SemaphoreSet will less often create a semaphore needlessly.
475 - add Coro::SemaphoreSet::count and wait methods.
476 - take advantage of the new Guard module.
477 - deprecate Coro::guard.
478 - try to fix the dreaded 01_unblock tests once more. I hate it when
479 testsuites need more fixing than the code they are supposed to test.
480 - croak in more cases when a required callback isn't resolvable.
481 - fix some minor issues in Coro::State->call/eval.
482 - use current coroutine context instead of a temporary one
483 when temporarily switching to another coroutine.
484 - do not call C-level on_destroy handlers during global destruction,
485 to avoid needless segfaults.
486
4875.12 Sun Dec 7 13:30:38 CET 2008
488 - add default config for MirOS, which seems to be a bug-to-bug
489 compatible fork of openbsd ("world domination by releasing
490 openbsd cvs before the openbsd folks do it" or so :).
491 - free_padlist did destroy the names pad, not good, but didn't
492 seem to bother perl - this could fix issues such as eval ""
493 inside a function called from multiple coroutines.
494 - use a different method to detect destruction time.
495 - be more careful when freeing padlists just before global
496 destruction.
497 - fixed and expanded the call/cc example.
498 - renamed _terminate to _coro_run.
499 - new method Coro::Channel->shutdown.
500 - try pthreads on openbsd <4.4 (broken sigaltstack, will
501 pthreads fare better?).
502 - be less picky about destroying "a" running coroutine.
503
5045.11 Tue Nov 25 21:49:05 CET 2008
505 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
506 working code with failed assertions, introducing more bugs than
507 it fixes, requiring elaborate workarounds :(
508
5095.1 Mon Nov 24 08:54:59 CET 2008
510 - wrote a small introductory tutorial - Coro::Intro.
511 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
512 - Coro::Select did errornously dup the file descriptors
513 and didn't work with all AnyEvent backends.
514 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
515 LWP data transfers.
516 - disassociate c contexts from coro objects - this is agruably more
517 correct, but mostly allows sharing of cctxs between coro and state
518 objects, for added memory savings and speed increases.
519 - bumped $Coro::POOL_RSS up to 32kb by default.
520 - no longer set the optype to OP_CUSTOM, as B::* understandably
521 doesn't like this very much (and we *are* a type of entersub).
522 - implement state cloning, just to prove that call/cc can be done.
523 - automatically load Coro::AnyEvent in Coro::Handle.
524 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
525 do not have this method (and don't need it either).
526 - speed up generic anyevent methods in Coro::Handle by using rouse
527 callbacks.
528 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
529 others. It also makes the debugger happier, as you can trace
530 through the idle threads now.
531 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
532 - eg/event was pretty broken.
533 - better 5.8.6 compatibility.
534
5355.0 Thu Nov 20 10:35:05 CET 2008
536 - NEW ARCHITECTURE: use the latest 4.x version if you experience
537 stability issues.
538 - bump API version to 7 - all dependents must be recompiled.
539 - removed timed_* functions - they were not being used anyways
540 and should be replaced by a more generic mechanism -
541 and were annoying to support anyways :)
542 - removed SemaphoreSet's waiter method - use sem method instead.
543 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
544 - async_pool did free a scalar value twice
545 ("Attempt to unreference...").
546 - fix a longstanding bug where calling terminate on a coro that
547 was waiting for a semaphore that was just becoming available
548 would cause a deadlock (semaphore would get into a state where
549 it was available but waiters were still blocked).
550 - calling throw on a coroutine that is waiting for a semaphore will
551 no longer make it acquire the semaphore (and thus leak a count).
552 - perl's process emulation is now not even theoretically supported
553 anymore.
554 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
555 conversion of callback-style to blocking-style.
556 - new methods $coro->schedule_to and ->cede_to, to specifically
557 schedule or cede to a specific coroutine.
558 - new function Coro::Semaphore::wait.
559 - use named constants in Coro::Channel (Richard Hundt).
560 - directly patch the entersub opcode calling SLF functions (cede,
561 transfer and so on). this does speed up context switching, but
562 more importanly, it frees us from the hardcoded behaviour of
563 entersub, so we might actually be able to return something from
564 those functions and atcually create new ones.
565 - take advantage of __builtin_frame_address on gcc.
566 - expose THX in coroapi (not sure whether this was a wise decision,
567 as "threaded" perls are running at half speed anyways).
568 - implement execute_slf (schedule-like-function) interface that makes
569 it possible to implement schedule-like-functions in XS.
570 - use new SLF interface to massively speed up Coro::EV by roughly a
571 factor of two.
572 - used new SLF interface to massively speed up Coro::Semaphore by a
573 factor of three.
574 - used new SLF interface to speed up Coro::AIO by roughly a factor of
575 four and reduce its memory usage considerably.
576 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
577 for a nice speedup and vastly more correct behaviour. Also implement
578 a new method "sem" to get at the underlying semaphore object.
579 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
580 (in comparison) 20-40% speedup.
581 - used new SLF interface to reimplement Coro::Signal gaining
582 some unknown (because I was too lazy), but certain, speedup, and also
583 making signals reliable for the first time.
584 - used new SLF interface and other optimisations to speed up async_pool
585 by a factor of two. It also doesn't rely on perl's exception mechanism
586 to exit anymore. The overhead for terminating an async_pool, coro over
587 a normal async is now very small.
588 - sped up coroutine creation/destruction by 40%.
589 - forgot to include Coro/libcoro/README in the dist for all these years.
590 - work around a freebsd pthreads bug (manual testcancel is required as
591 pthread_cond_wait isn't a cancellation point on freebsd).
592 - use new rouse functions to speed up and simplify Coro::BDB.
593 - make "prefer perl native functions" work with threaded perls.
594 - condense Coro::Debug ps output, hint at v and w flags.
595 - (libcoro) lots of minor cleanups and portability improvements.
596
5974.914 Wed Nov 19 12:54:18 CET 2008
598 - fix a disastrous bug in the readline optimisation
599 introduced in 4.801.
600
6014.913 Sat Nov 15 07:58:28 CET 2008
602 - async_pool did free a scalar value twice
603 ("Attempt to unreference...").
604
6054.912 Thu Nov 13 18:31:23 CET 2008
606 - minor cleanups.
607 - use much larger stacks on linux and perl < 5.8.8.
608 - Coro::Debug::new_unix_server did not unlink the socket
609 when destroyed.
610
6114.911 Tue Nov 11 04:26:17 CET 2008
612 - "port" to threaded perls.
613
6144.91 Mon Nov 10 05:36:38 CET 2008
615 - the ->throw exception object no longer leaks.
616 - creating a new cctx leaked a scopestack entry (memleak).
617 - new coroutines didn't get created with a zero flags field
618 (unknown impact).
619 - calling ->throw on a not-yet-started coroutine should now work
620 instead of being ignored.
621 - ->throw is now supported on Coro::State objects.
622 - clean up cctx creation code a bit.
623 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
624
6254.9 Sat Nov 8 17:45:27 CET 2008
626 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
627 - (libcoro) no longer leak threads in the experimental pthread backend,
628 also speed it up considerably.
629 - (libcoro) do not rely on makecontext passing void *'s unscathed.
630 - fix compiletime dependencies on libcoro in the Makefile.
631 - cctx_count wasn't always updated properly.
632 - Coro::State::cctx_stacksize wasn't applied correctly.
633 - new function Coro::State::cctx_max_idle.
634 - the default max number of idle C contexts is now 4.
635 - (libcoro) try harder to get _setjmp/_longjmp.
636 - (libcoro) cleanup and extend the libcoro API to officially
637 allow the creation of empty source contexts.
638 - very experimental workaround for the totally broken netbsd platform.
639 - tried to hack around openbsd bugs.
640
6414.804 Wed Nov 5 16:36:00 CET 2008
642 - Coro::Debug::new_unix_server would not create a non-blocking listening
643 socket, sometimes causing freezes.
644 - (libcoro) fix misaligned stack points for setjmp and assembly
645 methods, which can cause crashes on x86/x86_64 with a sufficiently
646 aggressive compiler.
647 - new function: Coro::Debug::new_tcp_server.
648 - move ->throw into the Coro class because it only works on coro objects.
649
6504.803 Mon Nov 3 17:16:12 CET 2008
651 - (libcoro) use a global asm statement to become independent of gcc
652 otpimisations for CORO_ASM (thanks to pippijn for the idea).
653 - try to workaround yet another broken bsd, this time dragonfly.
654
6554.802 Thu Oct 30 10:56:12 CET 2008
656 - support -fno-omit-frame-pointer on x86 with the assembly method.
657 - tune 01_unblock.t tests a bit.
658
6594.801 Wed Oct 22 18:33:37 CEST 2008
660 - improve readline speed for very long "lines".
661 - backport to 5.8.8.
662
6634.8 Thu Oct 2 13:34:40 CEST 2008
664 - new function Coro::AIO::aio_wait.
665 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
666 - greatly speed up and reduce memory usage of Coro::AIO requests.
667 - implement some other µ-optimisations.
668
6694.749 Mon Sep 29 14:40:12 CEST 2008
670 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
671
6724.748 Sat Sep 27 14:03:19 CEST 2008
673 - implement, but do not document, PerlIO::cede(granularity).
674 - Coro::Storable forgot to wrap Storable::pstore.
675 - work around the multitude of leaks and memory corruption
676 bugs in PerlIO::via by using our own C-level perliol. As a side
677 effect, Coro::Storable is now much, much, much faster.
678
6794.747 Tue Sep 23 01:59:41 CEST 2008
680 - fix a per-cv memleak (one empty array was leaked per
681 code reference).
682 - avoid a crash in coro->call|rss when the coroutine was already
683 destroyed (most noticably when using Coro::Debug::ps :)
684 - also protect *Storable::FILE.
685 - push up default storable granularity to 20ms.
686
6874.746 Sun Sep 21 03:22:20 CEST 2008
688 - be more insistent on locking Storable against reentrancy
689 in Coro::Storable.
690 - move swap_def?v and throw to Coro::State, as documented.
691
6924.745 Thu Jul 24 00:14:38 CEST 2008
693 - remove debugging code related to MgPV_nolen_const, also try to
694 make it compile with perl 5.8.6 (yes, apple apparently loves
695 outdated software). Reported by John S.
696
6974.744 Tue Jul 8 22:06:35 CEST 2008
698 - correctly provide default DIE/WARN handlers as documented.
699 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
700 implementations inside perl work fine for us.
701
7024.743 Mon Jun 16 00:21:57 CEST 2008
703 - when using Coro::EV without running EV::loop it could
704 result in busy-waiting for events, this has been fixed.
705 - reduce codesize and improve performance by using EV_DEFAULT_UC.
706
7074.742 Sat May 31 14:10:21 CEST 2008
708 - implement a workaround for (some) perl <5.8.8 versions.
709 - require EV 3.3+.
710
7114.741 Fri May 30 23:33:09 CEST 2008
712 - tell netbsd how utterly broken their imitation of an OS is
713 and refuse to build by default if pthreads are in use.
714 - switch to "s" method on all bsd's by default, as their ucontext
715 stuff seems just too broken.
716 - fix a bug in Coro::Select.
717
7184.74 Thu May 29 20:05:31 CEST 2008
719 - do not test Coro::LWP for lack of dependencies.
720 - do not test Coro::Select for lack of working perls.
721
7224.73 Thu May 29 2008
723 - fix a bug in Coro::EV which would cause it to block despite
724 there being runnable coroutines.
725 - sprinkle "no warnings" freely over everything, also suppress
726 warnings for some other modules.
727 - fix typo in WSAEWOULDBLOCK.
728
7294.72 Sun May 25 05:14:36 CEST 2008
730 - tweak META.yaml a bit, unfortunately, there is no documented way
731 to have optional dependencies with CPAN. doh :(
732 - avoid running some tests on windows because they would fail due to
733 perl bug (broken fork, broken pipes...).
734 - work around perl on windows bugs where perl returns undocumented
735 error codes for sysread, syswrite etc. by taking advantage
736 of AnyEvent's workaround for that problem.
737 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
738 a common perl implementation bug on windows.
739 - use unix domain sockets in testsuite to work around
740 common perl implementation bugs on widows (they are emulated by
741 tcp sockets on windows. ugh.)
742
7434.71 Sat May 24 20:01:27 CEST 2008
744 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
745 - take advantage of async name resolution of AnyEvent::Util.
746 - work around brutal inet_aton override in Coro::LWP.
747 - take advantage of the readyhook in Coro::EV, for smoother
748 scheduling.
749
7504.7 Sun May 11 00:32:19 CEST 2008
751 - completely reworked the Coro manpage.
752 - added Coro::AnyEvent, generic event loop integration.
753 - implement cancel, ready and kill commands in Coro::Debug.
754 - document find_coro in Coro::Debug.
755 - incompatible change: rename has_stack to has_cctx.
756 - Coro::AIO and Coro::BDB no longer force event model detection,
757 use AnyEvent::AIO and AnyEvent::BDB.
758
7594.6 Sat Apr 26 10:05:14 CEST 2008
760 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
761 by the Coro module. It was a mistake to have it in the first place.
762 - (experimental) support for activestate perl 5.10 (method "w").
763 - (experimental) support for strawberry perl (method "a").
764 - coro_sigelem_set did not return a value although it had to,
765 actual impact unknown.
766
7674.51 Mon Apr 14 13:28:27 CEST 2008
768 - make it compile again on 5.8.
769
7704.50 Thu Apr 10 09:43:17 CEST 2008
771 - I did it twice! (see 4.49).
7 772
84.49 Sun Apr 6 21:23:31 CEST 2008 7734.49 Sun Apr 6 21:23:31 CEST 2008
9 - grr, instead of compiling the recent changes on 5.10 only they 774 - grr, instead of compiling the recent changes on 5.10 only they
10 were compiled on 5.8 only. 775 were compiled on 5.8 only.
11 776
… …
116 881
1174.11 Thu Oct 11 02:40:24 CEST 2007 8824.11 Thu Oct 11 02:40:24 CEST 2007
118 - port to threaded perls. 883 - port to threaded perls.
119 884
1204.1 Thu Oct 11 02:38:16 CEST 2007 8854.1 Thu Oct 11 02:38:16 CEST 2007
121 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 886 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
122 be local to each coro (see Coro::State). 887 be local to each coro (see Coro::State).
123 - incompatible change: for very deep reasons, cede and cede_notself 888 - incompatible change: for very deep reasons, cede and cede_notself
124 cannot return anything, so nothing will be returned. 889 cannot return anything, so nothing will be returned.
125 - possibly bring back 5.10 compatibility (untested). 890 - possibly bring back 5.10 compatibility (untested).
126 - work around stupid (and wrong) warning on 5.10 :(. 891 - work around stupid (and wrong) warning on 5.10 :(.
… …
198 but stateful semantics. 963 but stateful semantics.
199 - fixed a lot of typos in Coro.pm (patch submitted by David 964 - fixed a lot of typos in Coro.pm (patch submitted by David
200 Steinbrunner, which applied flawlessly). 965 Steinbrunner, which applied flawlessly).
201 966
2023.6 Sat Apr 14 17:13:31 CEST 2007 9673.6 Sat Apr 14 17:13:31 CEST 2007
203 - added some bugfixes to get eg/myhttpd working again. 968 - added some bugfixes to get eg/myhttpd working again.
204 - added Coro::Storable for often-cede'ing freeze/thaw. 969 - added Coro::Storable for often-cede'ing freeze/thaw.
205 - try to do a clean exit when a coroutine calls exit 970 - try to do a clean exit when a coroutine calls exit
206 (EXPERIMENTAL). 971 (EXPERIMENTAL).
207 - got rid of indirect call through _coro_init. 972 - got rid of indirect call through _coro_init.
208 - updated the partly antique examples in eg/ to 973 - updated the partly antique examples in eg/ to

Diff Legend

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