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

Comparing Coro/Changes (file contents):
Revision 1.363 by root, Mon Jul 21 02:39:15 2008 UTC vs.
Revision 1.673 by root, Sun Sep 3 23:36:01 2017 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?)
7 9
10 - minor code cleanups.
11
126.514 Thu Aug 31 18:25:31 CEST 2017
13 - safe_cancel no longer croaks when called on already-destroyed
14 thread, but instead "just works".
15
166.513 Sat Jul 15 01:18:51 CEST 2017
17 - experimental 5.22 comaptibility changes (PadlistNAMES).
18 - removed attributes example and made sure the other examples
19 still work.
20 - fix some statevar types for newer perls, should be of no consequence.
21
226.512 Fri Jul 14 05:03:17 CEST 2017
23 - seems to pass test suite down till 5.8.9, so make this the
24 minimum requirement in Makefile.PL.
25 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
26 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
27 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
28 fortified longjmp variants in some libcs.
29 - re-try JIT allocation without PROT_EXEC, to hopefully improve
30 portability (still doesn't seem to work around broken selinux
31 mmap everywhere).
32 - speling fixes (Xavier Guimard).
33
346.511 Sun Jun 26 23:44:50 CEST 2016
35 - make it compile with threaded perls again.
36 - simplify and speed up __DIE__ and __WARN__ handling.
37
386.51 Sat Jun 25 20:57:20 CEST 2016
39 - 6.5 release didn't compile on most older perls due to
40 a logic error enabling experimental 5.24 code for all versions.
41 - since various broken or deficient patches float around
42 to work around the vtbl API breakage in perl, let's provide
43 our own workaround, which seems to be much less invasive
44 and more compatible than the approaches seen so far. YMMV.
45 - work around assertion failure in perl_destruct on 5.24
46 (analyzed by Dave Mitchell).
47
486.5 Wed Jun 22 22:23:50 CEST 2016
49 - swap_sv swapping order was not symmetrical, causing
50 wrong swaps when swap_sv was used multiple times on the
51 same sv in the same thread.
52 - swap_sv calls can now be undone by calling it again
53 with the same variables.
54 - swap_sv calls will now be undone in async_pool threads.
55 - split Coro::Semaphore::up/adjust into separate xs functions
56 for better error reporting, at slight codesize increase.
57 - (libcoro) arm assembly support, please test and report.
58 - adjust to PL_savestack changes in perl 5.24 (adapted from
59 the debian patch, which unfortunately gets it wrong).
60
616.49 Sat Oct 17 01:40:12 CEST 2015
62 - throwing an exception to a thread waiting in
63 Coro::Handle using Coro::EV did not stop the watchers,
64 causing the next call to fail (testcase by Martin Pritchard).
65 - bump minimum perl version to 5.10.
66
676.48 Sun Oct 4 19:03:51 CEST 2015
68 - fix memory leak when Coro::Handle uses Coro::EV internally
69 (testcase by Sten Sten).
70 - update code to libev 4 API (internally, Coro still used the
71 version 3 API calls).
72 - remove unused hv_sig.
73
746.47 Sat Jul 11 03:58:20 CEST 2015
75 - swap_svs - NVs can be in the SV head beginning with 5.20.
76
776.46 Tue Jun 30 14:36:30 CEST 2015
78 - restore portability to perls with windows fork emulation
79 (patch by Petr Písař).
80
816.45 Tue Jun 30 01:40:08 CEST 2015
82 - restore portability to perl 5.16 and below.
83
846.44 Tue Jun 30 00:41:54 CEST 2015
85 - avoid segfaulting (or worse) when tracing a canceled thread.
86 - the Event module silently broke it's public hook API, causing
87 failures on perls compiled with -Duselongdouble. Requiring
88 the latest version of Event because I don't know which release
89 changed this (it's not mentioned in the ChangeLog).
90 - allow xs level enterleave hooks via CoroAPI.
91
926.43 Thu Jun 4 15:38:14 CEST 2015
93 - use stability canary.
94 - port to stableperl-5.22.0-1.001.
95 - update libecb.
96
976.42 Wed Feb 11 20:29:52 CET 2015
98 - Coro::SemaphoreSet->try did not actually work (analyzed by
99 SATO Kentaro).
100 - upgrade libecb to be C11 compliant.
101
1026.41 Sat Sep 6 22:08:46 CEST 2014
103 - restore portability to perl 5.8.x.
104 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
105 up some memory, but fixes "use feature" and similar modules
106 when used inside a Coro.
107 - allow zero as argument to Coro::Channel to mean the same thing
108 as no argument. this works with older versions as well,
109 but wasn't legal until now.
110 - slightly better c header file detection.
111
1126.39 Mon Jun 2 00:00:08 CEST 2014
113 - work around more incompatible changes in 5.20.
114
1156.38 Sun Jun 1 21:54:23 CEST 2014
116 - check that perl slots actually have enough space to hold
117 interpreter variables.
118 - untested port to perl 5.19 (64 bit tmps indices) (reported
119 by Andreas König).
120 - croak when cancel is called without a thread context.
121
1226.37 Tue Mar 4 13:27:33 CET 2014
123 - *sigh*, removed leftover debugging code from debugging a
124 perl bug, of all things.
125
1266.36 Tue Mar 4 07:11:59 CET 2014
127 - semaphores would not clear the destroy hook when interrupted
128 by ->throw, causing segfaults or worse.
129 - ->throw on a thread waiting for a semaphore did not acquire
130 the semaphore, but also didn't wake up other waiters,
131 possibly causing a deadlock.
132 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
133 to report a full stacktrace, which should help find out
134 where the actual call is.
135 - "atomically" destroy data in slf_destroy, because it is
136 the right thing to do, just in case.
137 - disable ecb.h libm dependency, to compile on ancient systems
138 or under adverse conditions.
139
1406.33 Mon Nov 18 11:26:27 CET 2013
141 - do not crash when freeing padlists with holes (in 5.18).
142 - tentative SVt_BIND 5.19 port/fix.
143
1446.32 Tue Nov 5 15:35:35 CET 2013
145 - use a new algorithm to derive padlists for perl 5.18. The old
146 one could lead to 0-pointer accesses inside perl (reported
147 by Darin McBride).
148
1496.31 Thu May 9 07:39:48 CEST 2013
150 - Coro::AIO requests would crash if the thread was ready'd
151 while the request was ongoing.
152
1536.29 Wed May 8 02:55:18 CEST 2013
154 - when an on_destroy handler destructs the coro currently being
155 destructed a perl scalar could be accessed after being freed,
156 likely causing a crash.
157
1586.28 Wed Mar 6 06:58:02 CET 2013
159 - clean remnants of existing __DIE__ and __WARN__ handlers so
160 they lose their magic and will not cause segfaults later
161 (testcase by Andrey Sagulin).
162 - improved Coro::State documentation a bit.
163 - Coro::Debug::command now flushes the output.
164 - add hack detection code for x32 abi, because the braindead slugs
165 who designed that made it look exactly like x86_64 without
166 providing proper compile time symbols to test for it. as a result,
167 this detection cannot work reliably.
168 - valgrind stack registering was broken.
169 - do not rely on Time::HiRes anymore in Coro::Debug.
170
1716.23 Fri Dec 7 23:36:37 CET 2012
172 - use experimental fiber implementation on native windows
173 perls.
174 - use sizeof (void *) as multiplication factor for stack sizes,
175 to accomodate the totally braindamaged microsoft 64 bit "os".
176 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
177 activeperl 5.16 crashes when PerlIO_define_layer is called due
178 to some bug in the perl dll, strawberry perl at least passes
179 the testsuite.
180 - implement Coro::Handle->peeraddr/host/port, for slightly
181 improved compatibility with LWP.
182 - implement 5.17 compatibility by almost blindly applying a
183 good-looking patch by Father Chrysostomos.
184 - move stack management functions into libcoro 3.
185 - libcoro version 3 "released".
186 - support magic values as timed_io_once args.
187 - recommend AnyEvent 7+ or EV 4+, also require EV
188 version 4 or newer for Coro::EV.
189
1906.10 Tue Oct 9 01:14:27 CEST 2012
191 - updated ecb.h, it had a typo that caused it to not compile on many
192 big endian systems (reported by many people).
193 - disable memory fences in ecb.h to improve portability.
194
1956.09 Sat Oct 6 23:25:02 CEST 2012
196 - Coro::EV I/O watchers were not interruptible by exceptions
197 (Coro::State::throw) (testcase by sten).
198 - ->throw now puts threads into the ready queue, as this seems to
199 be expected by existing code, and code that doesn't cope with spurious
200 wakeups needs fixing anyway.
201 - use fd -1 in mmap.
202 - cast I32 to int in error message printf.
203 - warn about broken so-called "hardened" kernels.
204
2056.08 Fri Apr 13 12:05:47 CEST 2012
206 - be more aggressive about exiting like perl does - formerly,
207 exiting from the non-main thread would not execute END blocks.
208
2096.07 Fri Nov 11 21:21:48 CET 2011
210 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
211 - update ecb.h.
212
2136.06 Mon Aug 8 23:59:48 CEST 2011
214 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
215 backend, which is an order of magnitude slower. unfortunately, cygwins
216 pthread implementation isn't very complete either, so allocate the stack
217 twice just to be sure.
218 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
219 in its ithreaded perl - best recompile cygwin and use the 'w'indows
220 backend for much better performance. also disable ithreads for
221 even better performance...).
222
2236.05 Thu Aug 4 21:36:36 CEST 2011
224 - blush, condvar values would not be propagated from send to recv anymore
225 (reported by Chip Salzenberg).
226 - use exponential increase for the readline buffer length in
227 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
228
2296.04 Wed Aug 3 17:15:45 CEST 2011
230 - use even more efficient and more compatible condvars for
231 compatibility to AnyEvent 6.x :)
232 - more inconsequential ecb.h updates.
233
2346.03 Wed Aug 3 11:41:30 CEST 2011
235 - change how Coro patches AnyEvent condvars for compatibility to
236 AnyEvent 6.x.
237 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
238 hasn't been included.
239
2406.02 Wed Jul 13 04:35:19 CEST 2011
241 - "improve portability to Gentoo" - gentoo manages to put perl variables
242 in memory areas that are farther than 2gb apart, which the jit couldn't
243 handle and barfed. now it's just a bit slower on gentoo and similar
244 systems.
245
2466.01 Sun Jul 3 12:31:14 CEST 2011
247 - workarounds are good, but the test for whether pthreads are used
248 was not good. this one should be better.
249 - check differently whether gcc generates cfi instructions itself.
250
2516.0 Wed Jun 29 19:43:35 CEST 2011
252 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
253 destroyed and cleaned up automatically (e.g. async { schedule }).
254 - implement a JIT compiler for part of the thread switch code,
255 which gives a 50% speed improvement on threaded perls, and
256 about 4% on non-threaded perls (so threaded perls now finally
257 reach about half the speed of non-threaded perls).
258 - slightly modernise Coro::Intro, add section about rouse functions.
259 - avoid DEFSV and ERRSV, giving another 10% improvement
260 in thread switching.
261 - Coro::State->is_destroyed is now called is_zombie.
262 - implement a Coro->safe_cancel method that might fail, but
263 cancels in a "safer" way if it succeeds.
264 - add preliminary support for DEBUGGING perls.
265 - get rid of two hash-accesses when initialising a new Coro - this
266 speeds up coro creation by almost a factor of two.
267 - croak when a coro that is being cancelled tries to block
268 (e.g. while executing a guard block), instead of crashing or
269 deadlocking.
270 - use a more robust and also faster method to identify Coro::State
271 objects - speeds up everything a bit.
272 - implement Coro->cancel in XS for a 20% speed improvement, and to
273 be able to implement mutual cancellation.
274 - speed up context switches by a percent or two by more efficiently
275 allocating context stack entries.
276 - implement Coro->join and Coro->on_destroy in XS for a speedup and
277 a reduction in memory use.
278 - cancelling a coro while it itself is cancelling another coro is
279 now supported and working, instead of triggering an assertion.
280 - be a bit more crash-resistant when calling (buggy) on_destroy
281 callbacks (best effort).
282 - move on_destroy into the slf_frame, to allow extension slf
283 functions to have destructors.
284 - get rid if coro refcounting - simply crash in other interpreter
285 threads by nulling the pointers on clone.
286 - simplify warn/die hook handling when loading Coro - the convoluted
287 logic seems to be no longer neccessary.
288 - use libecb instead of our own home-grown gcc hacks.
289 - document alternatives to Coro::LWP. Please use them :)
290 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
291 sigaltstack. Really. wine suffers from it, erlang suffers from it,
292 and it's known since at least 2006.
293
2945.372 Wed Feb 23 06:14:30 CET 2011
295 - apparently mingw doesn't provide a working gettimeofday, try to
296 work around that by relying on Time::HiRes (indirectly brought to
297 my attention by Max Maischein).
298 - fix some portability issues when Time::HiRes was used.
299
3005.371 Mon Feb 21 14:36:08 CET 2011
301 - backport to windows process emulation code again.
302
3035.37 Sat Feb 19 07:49:44 CET 2011
304 - add a big "Coro thread life cycle" section to "man Coro".
305 - try a tentative workaround against the breakage that 5.13 has
306 introduced without depreciation period. sigh.
307 - no longer use Time::HiRes if gettimeofday is available, which
308 saves quite a lot of memory.
309
3105.36 Sun Feb 13 05:33:41 CET 2011
311 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
312 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
313 method call.
314 - undocument Coro::Timer::sleep and obsolete whole module.
315 - optimise Coro::Timer::timeout memory and cpu usage.
316 - slightly updated Coro::Intro for recent changes.
317 - do not initialise PL_dirty anymore.
318
3195.25 Thu Nov 11 01:08:39 CET 2010
320 - try a different approach on netbsd - netbsd 5 finally has marginally
321 working pthreads, but still broken ucontext/sigaltstack.
322 - openbsd 4.8 finally got their act together, Coro works out of the box
323 with asm, setjmp and pthreads (no change, just informational).
324
3255.24 Sat Oct 23 11:27:12 CEST 2010
326 - port to the EV 4.0 API.
327 - work around bugs in mingw32, making strawberry perl work
328 out of the box.
329 - correctly modify Coro::AIO function prototypes
330 so that they reflect the "no optional parameters" rule.
331 - "ported" libcoro to C++.
332
3335.23 Mon May 17 18:50:42 CEST 2010
334 - be more resistant to ordering changes when initialising
335 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
336 Waldorf).
337 - document that perl 5.12 deliberately removed support for cloning.
338
3395.22 Wed Apr 14 03:55:35 CEST 2010
340 - correctly return udnef on errors in Coro::Handle::read/write
341 (testcase by Marc Mims).
342 - convert Coro::Util into a "perl compatibility wrapper" - the functions
343 are less useful now, but are drop-in replacements for existing
344 functions, listing better alternatives in the documentation. This also
345 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
346 with Coro::Util::inet_aton.
347 - do not override $Coro::idle unconditionally in Coro.pm, as other
348 modules could have provided their own idle coro already
349 (for exmaple, Coro::AnyEvent).
350 - fix Coro::Util::gethost* functions.
351 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
352
3535.21 Wed Dec 16 07:19:51 CET 2009
354 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
355 together.
356 - add some examples on how to combine other event loops with Coro in
357 Coro::AnyEvent, and how to run it (and not to block). Seems to be
358 the most common source of confusion.
359 - try to catch people naively blocking in an event callback.
360 - work around the perl filehandle bug issue in conjunction with
361 older common::sense (as indirectly pointed out by ZSystem).
362 - clarify the "not from signal handlers" section.
363
3645.2 Sun Oct 4 14:54:24 CEST 2009
365 - Coro::Storable destroyed the prototypes of the functions it wrapped.
366 - export rouse_cb and rouse_wait by default now.
367 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
368 - new method $state->swap_sv.
369 - added section on "windows process emulation" to the manpage,
370 after a not-so-fruitful (nor-friendly) "discussion" with chip
371 salzenberg (discussion implies arguments, but his only arguments
372 were ad-hominems, one wonders why he started it in the first
373 place). I hope this explains it well enough for him to understand,
374 and maybe well enough for others to understand.
375 - use common::sense everywhere now.
376 - idle callbacks are no longer supported, use idle coros instead.
377 - print a thread listing when a deadlock is detected.
378
3795.17 Sat Aug 22 23:09:31 CEST 2009
380 - work around a bug in the perl debugger causing crashes
381 when running under the debugger by marking _pool_handler as nodebug.
382 - speed up Coro::async considerably.
383 - try some hacks to get netbsd to work "more often" - their broken
384 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
385 - convert Coro to AE 5.0 API.
386
3875.162 Tue Jul 28 04:04:03 CEST 2009
388 - perl 5.8.2 is now minimum requirement.
389 - skip t/19_handle.t on broken windows perls.
390
3915.161 Wed Jul 22 04:47:38 CEST 2009
392 - Coro::AnyEvent::poll could have a different prototype when EV was
393 used as backend (analyzed by Tatsuhiko Miyagawa).
394 - Coro::AnyEvent errornously initialised the event loop when loaded,
395 not on demand.
396 - try to workaround rare */t/01_unblock.t failures.
397
3985.16 Tue Jul 21 01:44:37 CEST 2009
399 - Coro::AnyEvent failed to hook into the event loop
400 when no threads had been readied between detecting
401 the event loop and actually running it.
402 - considerably speed up Coro::Select by taking avdantage
403 of AnyEvent > 4.8 and some other optimisations.
404 - implement paragraph readline mode in Coro::Handle
405 (based on patches by Zsbán Ambrus).
406 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
407 and analyzed by Yasuhiro MATSUMOTO).
408 - clarified libcoro license and copyright.
409 - someone stole my EXTRA_META!!!
410 - implement Coro::Select::patch_pp_sselect and it's brother,
411 for hardcode select overriding.
412
4135.151 Mon Jul 6 05:41:57 CEST 2009
414 - backport to windows process emulation code again (patch by
415 Yasuhiro MATSUMOTO).
416 - slightly update Coro::MakeMaker.
417
4185.15 Tue Jun 30 10:28:06 CEST 2009
419 - deprecate Coro::Socket, document how to get ipv6 support via
420 AnyEvent::Socket instead.
421 - implement signal->wait ($cb) interface, similar to semaphores.
422 - work around SvOK not supporting getmagic, so we have to getmagic
423 to test for undef :/ (reported by Matthias Waldorf).
424 - load Coro::AnyEvent in all modules using AnyEvent.
425 - work around perl corrupting our internal data structures,
426 reported by Tokuhiro Matsuno.
427 - enable per-coroutine real and cpu time gathering
428 (Coro::State::enable_times).
429
4305.14 Wed Jun 24 01:37:48 CEST 2009
431 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
432 as this seems to have been a difficult concept (poll, sleep, idle,
433 idle_upto).
434 - add Coro::AnyEvent::readable/writable functions.
435 - clarify Coro::EV/Event/AnyEvent manpages.
436 - free per-thread global scalars in the thread calling ->cancel, to
437 avoid crashes when $_, $@ etc., are magical but some of those
438 had already been freed.
439 - "unexperimentalise" the callback interface for Coro::Semaphore.
440 - speed up ready queue management/context switching by using a linked
441 list instead of an array (~5-10%).
442 - implement "watch" command in Coro::Debug shells.
443 - for fun, implement time-slicing as an example in the manpage.
444 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
445 Coro::Event, use the normal AnyEvent handling instead of dieing
446 (the same is true for Coro::Handle).
447 - properly document Coro::EV::timed_io_once.
448 - avoid unneccessary ->cancel calls in Coro::Handle.
449 - maybe make it work on mingw32 with win32 backend
450 (based on patch by Yasuhiro Matsumoto).
451
4525.132 Fri May 29 09:00:39 CEST 2009
453 - do not keep a reference to the argument itself in
454 Coro::Semaphore::guard, as it could change later.
455 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
456 should support a prepare callback.
457
4585.131 Mon Mar 16 23:20:37 CET 2009
459 - implement and document Coro->suspend, Coro->resume.
460 - fix Coro::Select implementation to not (often) close
461 the passed file descriptors (testcase provided by pippijn).
462
4635.13 Mon Dec 15 21:51:42 CET 2008
464 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
465 - don't set diehook to C<undef> but instead to NULL, to avoid
466 spurious warnings.
467 - fix a lot of bugs in Coro::SemaphoreSet.
468 - Coro::SemaphoreSet will less often create a semaphore needlessly.
469 - add Coro::SemaphoreSet::count and wait methods.
470 - take advantage of the new Guard module.
471 - deprecate Coro::guard.
472 - try to fix the dreaded 01_unblock tests once more. I hate it when
473 testsuites need more fixing than the code they are supposed to test.
474 - croak in more cases when a required callback isn't resolvable.
475 - fix some minor issues in Coro::State->call/eval.
476 - use current coroutine context instead of a temporary one
477 when temporarily switching to another coroutine.
478 - do not call C-level on_destroy handlers during global destruction,
479 to avoid needless segfaults.
480
4815.12 Sun Dec 7 13:30:38 CET 2008
482 - add default config for MirOS, which seems to be a bug-to-bug
483 compatible fork of openbsd ("world domination by releasing
484 openbsd cvs before the openbsd folks do it" or so :).
485 - free_padlist did destroy the names pad, not good, but didn't
486 seem to bother perl - this could fix issues such as eval ""
487 inside a function called from multiple coroutines.
488 - use a different method to detect destruction time.
489 - be more careful when freeing padlists just before global
490 destruction.
491 - fixed and expanded the call/cc example.
492 - renamed _terminate to _coro_run.
493 - new method Coro::Channel->shutdown.
494 - try pthreads on openbsd <4.4 (broken sigaltstack, will
495 pthreads fare better?).
496 - be less picky about destroying "a" running coroutine.
497
4985.11 Tue Nov 25 21:49:05 CET 2008
499 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
500 working code with failed assertions, introducing more bugs than
501 it fixes, requiring elaborate workarounds :(
502
5035.1 Mon Nov 24 08:54:59 CET 2008
504 - wrote a small introductory tutorial - Coro::Intro.
505 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
506 - Coro::Select did errornously dup the file descriptors
507 and didn't work with all AnyEvent backends.
508 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
509 LWP data transfers.
510 - disassociate c contexts from coro objects - this is agruably more
511 correct, but mostly allows sharing of cctxs between coro and state
512 objects, for added memory savings and speed increases.
513 - bumped $Coro::POOL_RSS up to 32kb by default.
514 - no longer set the optype to OP_CUSTOM, as B::* understandably
515 doesn't like this very much (and we *are* a type of entersub).
516 - implement state cloning, just to prove that call/cc can be done.
517 - automatically load Coro::AnyEvent in Coro::Handle.
518 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
519 do not have this method (and don't need it either).
520 - speed up generic anyevent methods in Coro::Handle by using rouse
521 callbacks.
522 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
523 others. It also makes the debugger happier, as you can trace
524 through the idle threads now.
525 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
526 - eg/event was pretty broken.
527 - better 5.8.6 compatibility.
528
5295.0 Thu Nov 20 10:35:05 CET 2008
530 - NEW ARCHITECTURE: use the latest 4.x version if you experience
531 stability issues.
532 - bump API version to 7 - all dependents must be recompiled.
533 - removed timed_* functions - they were not being used anyways
534 and should be replaced by a more generic mechanism -
535 and were annoying to support anyways :)
536 - removed SemaphoreSet's waiter method - use sem method instead.
537 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
538 - async_pool did free a scalar value twice
539 ("Attempt to unreference...").
540 - fix a longstanding bug where calling terminate on a coro that
541 was waiting for a semaphore that was just becoming available
542 would cause a deadlock (semaphore would get into a state where
543 it was available but waiters were still blocked).
544 - calling throw on a coroutine that is waiting for a semaphore will
545 no longer make it acquire the semaphore (and thus leak a count).
546 - perl's process emulation is now not even theoretically supported
547 anymore.
548 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
549 conversion of callback-style to blocking-style.
550 - new methods $coro->schedule_to and ->cede_to, to specifically
551 schedule or cede to a specific coroutine.
552 - new function Coro::Semaphore::wait.
553 - use named constants in Coro::Channel (Richard Hundt).
554 - directly patch the entersub opcode calling SLF functions (cede,
555 transfer and so on). this does speed up context switching, but
556 more importanly, it frees us from the hardcoded behaviour of
557 entersub, so we might actually be able to return something from
558 those functions and atcually create new ones.
559 - take advantage of __builtin_frame_address on gcc.
560 - expose THX in coroapi (not sure whether this was a wise decision,
561 as "threaded" perls are running at half speed anyways).
562 - implement execute_slf (schedule-like-function) interface that makes
563 it possible to implement schedule-like-functions in XS.
564 - use new SLF interface to massively speed up Coro::EV by roughly a
565 factor of two.
566 - used new SLF interface to massively speed up Coro::Semaphore by a
567 factor of three.
568 - used new SLF interface to speed up Coro::AIO by roughly a factor of
569 four and reduce its memory usage considerably.
570 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
571 for a nice speedup and vastly more correct behaviour. Also implement
572 a new method "sem" to get at the underlying semaphore object.
573 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
574 (in comparison) 20-40% speedup.
575 - used new SLF interface to reimplement Coro::Signal gaining
576 some unknown (because I was too lazy), but certain, speedup, and also
577 making signals reliable for the first time.
578 - used new SLF interface and other optimisations to speed up async_pool
579 by a factor of two. It also doesn't rely on perl's exception mechanism
580 to exit anymore. The overhead for terminating an async_pool, coro over
581 a normal async is now very small.
582 - sped up coroutine creation/destruction by 40%.
583 - forgot to include Coro/libcoro/README in the dist for all these years.
584 - work around a freebsd pthreads bug (manual testcancel is required as
585 pthread_cond_wait isn't a cancellation point on freebsd).
586 - use new rouse functions to speed up and simplify Coro::BDB.
587 - make "prefer perl native functions" work with threaded perls.
588 - condense Coro::Debug ps output, hint at v and w flags.
589 - (libcoro) lots of minor cleanups and portability improvements.
590
5914.914 Wed Nov 19 12:54:18 CET 2008
592 - fix a disastrous bug in the readline optimisation
593 introduced in 4.801.
594
5954.913 Sat Nov 15 07:58:28 CET 2008
596 - async_pool did free a scalar value twice
597 ("Attempt to unreference...").
598
5994.912 Thu Nov 13 18:31:23 CET 2008
600 - minor cleanups.
601 - use much larger stacks on linux and perl < 5.8.8.
602 - Coro::Debug::new_unix_server did not unlink the socket
603 when destroyed.
604
6054.911 Tue Nov 11 04:26:17 CET 2008
606 - "port" to threaded perls.
607
6084.91 Mon Nov 10 05:36:38 CET 2008
609 - the ->throw exception object no longer leaks.
610 - creating a new cctx leaked a scopestack entry (memleak).
611 - new coroutines didn't get created with a zero flags field
612 (unknown impact).
613 - calling ->throw on a not-yet-started coroutine should now work
614 instead of being ignored.
615 - ->throw is now supported on Coro::State objects.
616 - clean up cctx creation code a bit.
617 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
618
6194.9 Sat Nov 8 17:45:27 CET 2008
620 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
621 - (libcoro) no longer leak threads in the experimental pthread backend,
622 also speed it up considerably.
623 - (libcoro) do not rely on makecontext passing void *'s unscathed.
624 - fix compiletime dependencies on libcoro in the Makefile.
625 - cctx_count wasn't always updated properly.
626 - Coro::State::cctx_stacksize wasn't applied correctly.
627 - new function Coro::State::cctx_max_idle.
628 - the default max number of idle C contexts is now 4.
629 - (libcoro) try harder to get _setjmp/_longjmp.
630 - (libcoro) cleanup and extend the libcoro API to officially
631 allow the creation of empty source contexts.
632 - very experimental workaround for the totally broken netbsd platform.
633 - tried to hack around openbsd bugs.
634
6354.804 Wed Nov 5 16:36:00 CET 2008
636 - Coro::Debug::new_unix_server would not create a non-blocking listening
637 socket, sometimes causing freezes.
638 - (libcoro) fix misaligned stack points for setjmp and assembly
639 methods, which can cause crashes on x86/x86_64 with a sufficiently
640 aggressive compiler.
641 - new function: Coro::Debug::new_tcp_server.
642 - move ->throw into the Coro class because it only works on coro objects.
643
6444.803 Mon Nov 3 17:16:12 CET 2008
645 - (libcoro) use a global asm statement to become independent of gcc
646 otpimisations for CORO_ASM (thanks to pippijn for the idea).
647 - try to workaround yet another broken bsd, this time dragonfly.
648
6494.802 Thu Oct 30 10:56:12 CET 2008
650 - support -fno-omit-frame-pointer on x86 with the assembly method.
651 - tune 01_unblock.t tests a bit.
652
6534.801 Wed Oct 22 18:33:37 CEST 2008
654 - improve readline speed for very long "lines".
655 - backport to 5.8.8.
656
6574.8 Thu Oct 2 13:34:40 CEST 2008
658 - new function Coro::AIO::aio_wait.
659 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
660 - greatly speed up and reduce memory usage of Coro::AIO requests.
661 - implement some other µ-optimisations.
662
6634.749 Mon Sep 29 14:40:12 CEST 2008
664 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
665
6664.748 Sat Sep 27 14:03:19 CEST 2008
667 - implement, but do not document, PerlIO::cede(granularity).
668 - Coro::Storable forgot to wrap Storable::pstore.
669 - work around the multitude of leaks and memory corruption
670 bugs in PerlIO::via by using our own C-level perliol. As a side
671 effect, Coro::Storable is now much, much, much faster.
672
6734.747 Tue Sep 23 01:59:41 CEST 2008
674 - fix a per-cv memleak (one empty array was leaked per
675 code reference).
676 - avoid a crash in coro->call|rss when the coroutine was already
677 destroyed (most noticably when using Coro::Debug::ps :)
678 - also protect *Storable::FILE.
679 - push up default storable granularity to 20ms.
680
6814.746 Sun Sep 21 03:22:20 CEST 2008
682 - be more insistent on locking Storable against reentrancy
683 in Coro::Storable.
684 - move swap_def?v and throw to Coro::State, as documented.
685
6864.745 Thu Jul 24 00:14:38 CEST 2008
8 - remove debugging code related to MgPV_nolen_const. 687 - remove debugging code related to MgPV_nolen_const, also try to
688 make it compile with perl 5.8.6 (yes, apple apparently loves
689 outdated software). Reported by John S.
9 690
104.744 Tue Jul 8 22:06:35 CEST 2008 6914.744 Tue Jul 8 22:06:35 CEST 2008
11 - correctly provide default DIE/WARN handlers as documented. 692 - correctly provide default DIE/WARN handlers as documented.
12 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 693 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
13 implementations inside perl work fine for us. 694 implementations inside perl work fine for us.
14 695
154.743 Mon Jun 16 00:21:57 CEST 2008 6964.743 Mon Jun 16 00:21:57 CEST 2008
16 - when using Coro::EV without running EV::loop it could 697 - when using Coro::EV without running EV::loop it could
… …
38 - sprinkle "no warnings" freely over everything, also suppress 719 - sprinkle "no warnings" freely over everything, also suppress
39 warnings for some other modules. 720 warnings for some other modules.
40 - fix typo in WSAEWOULDBLOCK. 721 - fix typo in WSAEWOULDBLOCK.
41 722
424.72 Sun May 25 05:14:36 CEST 2008 7234.72 Sun May 25 05:14:36 CEST 2008
43 - tweak META.yaml a bit, unfortunately, there is no documented way 724 - tweak META.yaml a bit, unfortunately, there is no documented way
44 to have optional dependencies with CPAN. doh :( 725 to have optional dependencies with CPAN. doh :(
45 - avoid running some tests on windows because they would fail due to 726 - avoid running some tests on windows because they would fail due to
46 perl bug (broken fork, broken pipes...). 727 perl bug (broken fork, broken pipes...).
47 - work around perl on windows bugs where perl returns undocumented 728 - work around perl on windows bugs where perl returns undocumented
48 error codes for sysread, syswrite etc. by taking advantage 729 error codes for sysread, syswrite etc. by taking advantage
… …
52 - use unix domain sockets in testsuite to work around 733 - use unix domain sockets in testsuite to work around
53 common perl implementation bugs on widows (they are emulated by 734 common perl implementation bugs on widows (they are emulated by
54 tcp sockets on windows. ugh.) 735 tcp sockets on windows. ugh.)
55 736
564.71 Sat May 24 20:01:27 CEST 2008 7374.71 Sat May 24 20:01:27 CEST 2008
57 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 738 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
58 - take advantage of async name resolution of AnyEvent::Util. 739 - take advantage of async name resolution of AnyEvent::Util.
59 - work around brutal inet_aton override in Coro::LWP. 740 - work around brutal inet_aton override in Coro::LWP.
60 - take advantage of the readyhook in Coro::EV, for smoother 741 - take advantage of the readyhook in Coro::EV, for smoother
61 scheduling. 742 scheduling.
62 743
634.7 Sun May 11 00:32:19 CEST 2008 7444.7 Sun May 11 00:32:19 CEST 2008
64 - completely reworked the Coro manpage. 745 - completely reworked the Coro manpage.
65 - added Coro::AnyEvent, generic event loop integration. 746 - added Coro::AnyEvent, generic event loop integration.
66 - implement cancel, ready and kill commands in Coro::Debug. 747 - implement cancel, ready and kill commands in Coro::Debug.
67 - document find_coro in Coro::Debug. 748 - document find_coro in Coro::Debug.
68 - incompatible change: rename has_stack to has_cctx. 749 - incompatible change: rename has_stack to has_cctx.
69 - Coro::AIO and Coro::BDB no longer force event model detection, 750 - Coro::AIO and Coro::BDB no longer force event model detection,
70 use AnyEvent::AIO and AnyEvent::BDB. 751 use AnyEvent::AIO and AnyEvent::BDB.
… …
194 875
1954.11 Thu Oct 11 02:40:24 CEST 2007 8764.11 Thu Oct 11 02:40:24 CEST 2007
196 - port to threaded perls. 877 - port to threaded perls.
197 878
1984.1 Thu Oct 11 02:38:16 CEST 2007 8794.1 Thu Oct 11 02:38:16 CEST 2007
199 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 880 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
200 be local to each coro (see Coro::State). 881 be local to each coro (see Coro::State).
201 - incompatible change: for very deep reasons, cede and cede_notself 882 - incompatible change: for very deep reasons, cede and cede_notself
202 cannot return anything, so nothing will be returned. 883 cannot return anything, so nothing will be returned.
203 - possibly bring back 5.10 compatibility (untested). 884 - possibly bring back 5.10 compatibility (untested).
204 - work around stupid (and wrong) warning on 5.10 :(. 885 - work around stupid (and wrong) warning on 5.10 :(.
… …
276 but stateful semantics. 957 but stateful semantics.
277 - fixed a lot of typos in Coro.pm (patch submitted by David 958 - fixed a lot of typos in Coro.pm (patch submitted by David
278 Steinbrunner, which applied flawlessly). 959 Steinbrunner, which applied flawlessly).
279 960
2803.6 Sat Apr 14 17:13:31 CEST 2007 9613.6 Sat Apr 14 17:13:31 CEST 2007
281 - added some bugfixes to get eg/myhttpd working again. 962 - added some bugfixes to get eg/myhttpd working again.
282 - added Coro::Storable for often-cede'ing freeze/thaw. 963 - added Coro::Storable for often-cede'ing freeze/thaw.
283 - try to do a clean exit when a coroutine calls exit 964 - try to do a clean exit when a coroutine calls exit
284 (EXPERIMENTAL). 965 (EXPERIMENTAL).
285 - got rid of indirect call through _coro_init. 966 - got rid of indirect call through _coro_init.
286 - updated the partly antique examples in eg/ to 967 - updated the partly antique examples in eg/ to

Diff Legend

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