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

Comparing Coro/Changes (file contents):
Revision 1.366 by root, Thu Aug 14 12:48:20 2008 UTC vs.
Revision 1.679 by root, Sun Dec 16 10:06:34 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_cancel failure (vebrose mode? show how to make backtrace?)
9TODO: scalar context for Coro::AIO wrappers?
7 10
116.54 Sun Dec 16 11:06:07 CET 2018
12 - reinstate Coro::Select::select prototype - it breaks more than it fixes.
13
146.53 Sun Dec 16 10:33:15 CET 2018
15 - the autogenerated prototypes of Coro::AIO now add ";" at the end
16 to avoid generating named unary operators and messing up parsing
17 (mostly affects aio_mlockall 4 | 8 which was parsed as
18 (aio_mlockall 4) | 8).
19 - do not use a prototype for select in Coro::Select, as CORE::select also
20 doesn't have one.
21
226.52 Tue Aug 14 18:49:43 CEST 2018
23 - libcoro update: allow sharing of coro threads using the pthreads
24 backend among different pthreads, as is required by Coro::Multicore.
25 - always allow ->call and ->eval when the coro is the current coroutine. this
26 makes it possible, among other things, to call these on $Coro::main.
27 - minor code cleanups.
28
296.514 Thu Aug 31 18:25:31 CEST 2017
30 - safe_cancel no longer croaks when called on already-destroyed
31 thread, but instead "just works".
32
336.513 Sat Jul 15 01:18:51 CEST 2017
34 - experimental 5.22 comaptibility changes (PadlistNAMES).
35 - removed attributes example and made sure the other examples
36 still work.
37 - fix some statevar types for newer perls, should be of no consequence.
38
396.512 Fri Jul 14 05:03:17 CEST 2017
40 - seems to pass test suite down till 5.8.9, so make this the
41 minimum requirement in Makefile.PL.
42 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
43 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
44 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
45 fortified longjmp variants in some libcs.
46 - re-try JIT allocation without PROT_EXEC, to hopefully improve
47 portability (still doesn't seem to work around broken selinux
48 mmap everywhere).
49 - speling fixes (Xavier Guimard).
50
516.511 Sun Jun 26 23:44:50 CEST 2016
52 - make it compile with threaded perls again.
53 - simplify and speed up __DIE__ and __WARN__ handling.
54
556.51 Sat Jun 25 20:57:20 CEST 2016
56 - 6.5 release didn't compile on most older perls due to
57 a logic error enabling experimental 5.24 code for all versions.
58 - since various broken or deficient patches float around
59 to work around the vtbl API breakage in perl, let's provide
60 our own workaround, which seems to be much less invasive
61 and more compatible than the approaches seen so far. YMMV.
62 - work around assertion failure in perl_destruct on 5.24
63 (analyzed by Dave Mitchell).
64
656.5 Wed Jun 22 22:23:50 CEST 2016
66 - swap_sv swapping order was not symmetrical, causing
67 wrong swaps when swap_sv was used multiple times on the
68 same sv in the same thread.
69 - swap_sv calls can now be undone by calling it again
70 with the same variables.
71 - swap_sv calls will now be undone in async_pool threads.
72 - split Coro::Semaphore::up/adjust into separate xs functions
73 for better error reporting, at slight codesize increase.
74 - (libcoro) arm assembly support, please test and report.
75 - adjust to PL_savestack changes in perl 5.24 (adapted from
76 the debian patch, which unfortunately gets it wrong).
77
786.49 Sat Oct 17 01:40:12 CEST 2015
79 - throwing an exception to a thread waiting in
80 Coro::Handle using Coro::EV did not stop the watchers,
81 causing the next call to fail (testcase by Martin Pritchard).
82 - bump minimum perl version to 5.10.
83
846.48 Sun Oct 4 19:03:51 CEST 2015
85 - fix memory leak when Coro::Handle uses Coro::EV internally
86 (testcase by Sten Sten).
87 - update code to libev 4 API (internally, Coro still used the
88 version 3 API calls).
89 - remove unused hv_sig.
90
916.47 Sat Jul 11 03:58:20 CEST 2015
92 - swap_svs - NVs can be in the SV head beginning with 5.20.
93
946.46 Tue Jun 30 14:36:30 CEST 2015
95 - restore portability to perls with windows fork emulation
96 (patch by Petr Písař).
97
986.45 Tue Jun 30 01:40:08 CEST 2015
99 - restore portability to perl 5.16 and below.
100
1016.44 Tue Jun 30 00:41:54 CEST 2015
102 - avoid segfaulting (or worse) when tracing a canceled thread.
103 - the Event module silently broke it's public hook API, causing
104 failures on perls compiled with -Duselongdouble. Requiring
105 the latest version of Event because I don't know which release
106 changed this (it's not mentioned in the ChangeLog).
107 - allow xs level enterleave hooks via CoroAPI.
108
1096.43 Thu Jun 4 15:38:14 CEST 2015
110 - use stability canary.
111 - port to stableperl-5.22.0-1.001.
112 - update libecb.
113
1146.42 Wed Feb 11 20:29:52 CET 2015
115 - Coro::SemaphoreSet->try did not actually work (analyzed by
116 SATO Kentaro).
117 - upgrade libecb to be C11 compliant.
118
1196.41 Sat Sep 6 22:08:46 CEST 2014
120 - restore portability to perl 5.8.x.
121 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
122 up some memory, but fixes "use feature" and similar modules
123 when used inside a Coro.
124 - allow zero as argument to Coro::Channel to mean the same thing
125 as no argument. this works with older versions as well,
126 but wasn't legal until now.
127 - slightly better c header file detection.
128
1296.39 Mon Jun 2 00:00:08 CEST 2014
130 - work around more incompatible changes in 5.20.
131
1326.38 Sun Jun 1 21:54:23 CEST 2014
133 - check that perl slots actually have enough space to hold
134 interpreter variables.
135 - untested port to perl 5.19 (64 bit tmps indices) (reported
136 by Andreas König).
137 - croak when cancel is called without a thread context.
138
1396.37 Tue Mar 4 13:27:33 CET 2014
140 - *sigh*, removed leftover debugging code from debugging a
141 perl bug, of all things.
142
1436.36 Tue Mar 4 07:11:59 CET 2014
144 - semaphores would not clear the destroy hook when interrupted
145 by ->throw, causing segfaults or worse.
146 - ->throw on a thread waiting for a semaphore did not acquire
147 the semaphore, but also didn't wake up other waiters,
148 possibly causing a deadlock.
149 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
150 to report a full stacktrace, which should help find out
151 where the actual call is.
152 - "atomically" destroy data in slf_destroy, because it is
153 the right thing to do, just in case.
154 - disable ecb.h libm dependency, to compile on ancient systems
155 or under adverse conditions.
156
1576.33 Mon Nov 18 11:26:27 CET 2013
158 - do not crash when freeing padlists with holes (in 5.18).
159 - tentative SVt_BIND 5.19 port/fix.
160
1616.32 Tue Nov 5 15:35:35 CET 2013
162 - use a new algorithm to derive padlists for perl 5.18. The old
163 one could lead to 0-pointer accesses inside perl (reported
164 by Darin McBride).
165
1666.31 Thu May 9 07:39:48 CEST 2013
167 - Coro::AIO requests would crash if the thread was ready'd
168 while the request was ongoing.
169
1706.29 Wed May 8 02:55:18 CEST 2013
171 - when an on_destroy handler destructs the coro currently being
172 destructed a perl scalar could be accessed after being freed,
173 likely causing a crash.
174
1756.28 Wed Mar 6 06:58:02 CET 2013
176 - clean remnants of existing __DIE__ and __WARN__ handlers so
177 they lose their magic and will not cause segfaults later
178 (testcase by Andrey Sagulin).
179 - improved Coro::State documentation a bit.
180 - Coro::Debug::command now flushes the output.
181 - add hack detection code for x32 abi, because the braindead slugs
182 who designed that made it look exactly like x86_64 without
183 providing proper compile time symbols to test for it. as a result,
184 this detection cannot work reliably.
185 - valgrind stack registering was broken.
186 - do not rely on Time::HiRes anymore in Coro::Debug.
187
1886.23 Fri Dec 7 23:36:37 CET 2012
189 - use experimental fiber implementation on native windows
190 perls.
191 - use sizeof (void *) as multiplication factor for stack sizes,
192 to accomodate the totally braindamaged microsoft 64 bit "os".
193 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
194 activeperl 5.16 crashes when PerlIO_define_layer is called due
195 to some bug in the perl dll, strawberry perl at least passes
196 the testsuite.
197 - implement Coro::Handle->peeraddr/host/port, for slightly
198 improved compatibility with LWP.
199 - implement 5.17 compatibility by almost blindly applying a
200 good-looking patch by Father Chrysostomos.
201 - move stack management functions into libcoro 3.
202 - libcoro version 3 "released".
203 - support magic values as timed_io_once args.
204 - recommend AnyEvent 7+ or EV 4+, also require EV
205 version 4 or newer for Coro::EV.
206
2076.10 Tue Oct 9 01:14:27 CEST 2012
208 - updated ecb.h, it had a typo that caused it to not compile on many
209 big endian systems (reported by many people).
210 - disable memory fences in ecb.h to improve portability.
211
2126.09 Sat Oct 6 23:25:02 CEST 2012
213 - Coro::EV I/O watchers were not interruptible by exceptions
214 (Coro::State::throw) (testcase by sten).
215 - ->throw now puts threads into the ready queue, as this seems to
216 be expected by existing code, and code that doesn't cope with spurious
217 wakeups needs fixing anyway.
218 - use fd -1 in mmap.
219 - cast I32 to int in error message printf.
220 - warn about broken so-called "hardened" kernels.
221
2226.08 Fri Apr 13 12:05:47 CEST 2012
223 - be more aggressive about exiting like perl does - formerly,
224 exiting from the non-main thread would not execute END blocks.
225
2266.07 Fri Nov 11 21:21:48 CET 2011
227 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
228 - update ecb.h.
229
2306.06 Mon Aug 8 23:59:48 CEST 2011
231 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
232 backend, which is an order of magnitude slower. unfortunately, cygwins
233 pthread implementation isn't very complete either, so allocate the stack
234 twice just to be sure.
235 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
236 in its ithreaded perl - best recompile cygwin and use the 'w'indows
237 backend for much better performance. also disable ithreads for
238 even better performance...).
239
2406.05 Thu Aug 4 21:36:36 CEST 2011
241 - blush, condvar values would not be propagated from send to recv anymore
242 (reported by Chip Salzenberg).
243 - use exponential increase for the readline buffer length in
244 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
245
2466.04 Wed Aug 3 17:15:45 CEST 2011
247 - use even more efficient and more compatible condvars for
248 compatibility to AnyEvent 6.x :)
249 - more inconsequential ecb.h updates.
250
2516.03 Wed Aug 3 11:41:30 CEST 2011
252 - change how Coro patches AnyEvent condvars for compatibility to
253 AnyEvent 6.x.
254 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
255 hasn't been included.
256
2576.02 Wed Jul 13 04:35:19 CEST 2011
258 - "improve portability to Gentoo" - gentoo manages to put perl variables
259 in memory areas that are farther than 2gb apart, which the jit couldn't
260 handle and barfed. now it's just a bit slower on gentoo and similar
261 systems.
262
2636.01 Sun Jul 3 12:31:14 CEST 2011
264 - workarounds are good, but the test for whether pthreads are used
265 was not good. this one should be better.
266 - check differently whether gcc generates cfi instructions itself.
267
2686.0 Wed Jun 29 19:43:35 CEST 2011
269 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
270 destroyed and cleaned up automatically (e.g. async { schedule }).
271 - implement a JIT compiler for part of the thread switch code,
272 which gives a 50% speed improvement on threaded perls, and
273 about 4% on non-threaded perls (so threaded perls now finally
274 reach about half the speed of non-threaded perls).
275 - slightly modernise Coro::Intro, add section about rouse functions.
276 - avoid DEFSV and ERRSV, giving another 10% improvement
277 in thread switching.
278 - Coro::State->is_destroyed is now called is_zombie.
279 - implement a Coro->safe_cancel method that might fail, but
280 cancels in a "safer" way if it succeeds.
281 - add preliminary support for DEBUGGING perls.
282 - get rid of two hash-accesses when initialising a new Coro - this
283 speeds up coro creation by almost a factor of two.
284 - croak when a coro that is being cancelled tries to block
285 (e.g. while executing a guard block), instead of crashing or
286 deadlocking.
287 - use a more robust and also faster method to identify Coro::State
288 objects - speeds up everything a bit.
289 - implement Coro->cancel in XS for a 20% speed improvement, and to
290 be able to implement mutual cancellation.
291 - speed up context switches by a percent or two by more efficiently
292 allocating context stack entries.
293 - implement Coro->join and Coro->on_destroy in XS for a speedup and
294 a reduction in memory use.
295 - cancelling a coro while it itself is cancelling another coro is
296 now supported and working, instead of triggering an assertion.
297 - be a bit more crash-resistant when calling (buggy) on_destroy
298 callbacks (best effort).
299 - move on_destroy into the slf_frame, to allow extension slf
300 functions to have destructors.
301 - get rid if coro refcounting - simply crash in other interpreter
302 threads by nulling the pointers on clone.
303 - simplify warn/die hook handling when loading Coro - the convoluted
304 logic seems to be no longer neccessary.
305 - use libecb instead of our own home-grown gcc hacks.
306 - document alternatives to Coro::LWP. Please use them :)
307 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
308 sigaltstack. Really. wine suffers from it, erlang suffers from it,
309 and it's known since at least 2006.
310
3115.372 Wed Feb 23 06:14:30 CET 2011
312 - apparently mingw doesn't provide a working gettimeofday, try to
313 work around that by relying on Time::HiRes (indirectly brought to
314 my attention by Max Maischein).
315 - fix some portability issues when Time::HiRes was used.
316
3175.371 Mon Feb 21 14:36:08 CET 2011
318 - backport to windows process emulation code again.
319
3205.37 Sat Feb 19 07:49:44 CET 2011
321 - add a big "Coro thread life cycle" section to "man Coro".
322 - try a tentative workaround against the breakage that 5.13 has
323 introduced without depreciation period. sigh.
324 - no longer use Time::HiRes if gettimeofday is available, which
325 saves quite a lot of memory.
326
3275.36 Sun Feb 13 05:33:41 CET 2011
328 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
329 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
330 method call.
331 - undocument Coro::Timer::sleep and obsolete whole module.
332 - optimise Coro::Timer::timeout memory and cpu usage.
333 - slightly updated Coro::Intro for recent changes.
334 - do not initialise PL_dirty anymore.
335
3365.25 Thu Nov 11 01:08:39 CET 2010
337 - try a different approach on netbsd - netbsd 5 finally has marginally
338 working pthreads, but still broken ucontext/sigaltstack.
339 - openbsd 4.8 finally got their act together, Coro works out of the box
340 with asm, setjmp and pthreads (no change, just informational).
341
3425.24 Sat Oct 23 11:27:12 CEST 2010
343 - port to the EV 4.0 API.
344 - work around bugs in mingw32, making strawberry perl work
345 out of the box.
346 - correctly modify Coro::AIO function prototypes
347 so that they reflect the "no optional parameters" rule.
348 - "ported" libcoro to C++.
349
3505.23 Mon May 17 18:50:42 CEST 2010
351 - be more resistant to ordering changes when initialising
352 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
353 Waldorf).
354 - document that perl 5.12 deliberately removed support for cloning.
355
3565.22 Wed Apr 14 03:55:35 CEST 2010
357 - correctly return udnef on errors in Coro::Handle::read/write
358 (testcase by Marc Mims).
359 - convert Coro::Util into a "perl compatibility wrapper" - the functions
360 are less useful now, but are drop-in replacements for existing
361 functions, listing better alternatives in the documentation. This also
362 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
363 with Coro::Util::inet_aton.
364 - do not override $Coro::idle unconditionally in Coro.pm, as other
365 modules could have provided their own idle coro already
366 (for exmaple, Coro::AnyEvent).
367 - fix Coro::Util::gethost* functions.
368 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
369
3705.21 Wed Dec 16 07:19:51 CET 2009
371 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
372 together.
373 - add some examples on how to combine other event loops with Coro in
374 Coro::AnyEvent, and how to run it (and not to block). Seems to be
375 the most common source of confusion.
376 - try to catch people naively blocking in an event callback.
377 - work around the perl filehandle bug issue in conjunction with
378 older common::sense (as indirectly pointed out by ZSystem).
379 - clarify the "not from signal handlers" section.
380
3815.2 Sun Oct 4 14:54:24 CEST 2009
382 - Coro::Storable destroyed the prototypes of the functions it wrapped.
383 - export rouse_cb and rouse_wait by default now.
384 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
385 - new method $state->swap_sv.
386 - added section on "windows process emulation" to the manpage,
387 after a not-so-fruitful (nor-friendly) "discussion" with chip
388 salzenberg (discussion implies arguments, but his only arguments
389 were ad-hominems, one wonders why he started it in the first
390 place). I hope this explains it well enough for him to understand,
391 and maybe well enough for others to understand.
392 - use common::sense everywhere now.
393 - idle callbacks are no longer supported, use idle coros instead.
394 - print a thread listing when a deadlock is detected.
395
3965.17 Sat Aug 22 23:09:31 CEST 2009
397 - work around a bug in the perl debugger causing crashes
398 when running under the debugger by marking _pool_handler as nodebug.
399 - speed up Coro::async considerably.
400 - try some hacks to get netbsd to work "more often" - their broken
401 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
402 - convert Coro to AE 5.0 API.
403
4045.162 Tue Jul 28 04:04:03 CEST 2009
405 - perl 5.8.2 is now minimum requirement.
406 - skip t/19_handle.t on broken windows perls.
407
4085.161 Wed Jul 22 04:47:38 CEST 2009
409 - Coro::AnyEvent::poll could have a different prototype when EV was
410 used as backend (analyzed by Tatsuhiko Miyagawa).
411 - Coro::AnyEvent errornously initialised the event loop when loaded,
412 not on demand.
413 - try to workaround rare */t/01_unblock.t failures.
414
4155.16 Tue Jul 21 01:44:37 CEST 2009
416 - Coro::AnyEvent failed to hook into the event loop
417 when no threads had been readied between detecting
418 the event loop and actually running it.
419 - considerably speed up Coro::Select by taking avdantage
420 of AnyEvent > 4.8 and some other optimisations.
421 - implement paragraph readline mode in Coro::Handle
422 (based on patches by Zsbán Ambrus).
423 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
424 and analyzed by Yasuhiro MATSUMOTO).
425 - clarified libcoro license and copyright.
426 - someone stole my EXTRA_META!!!
427 - implement Coro::Select::patch_pp_sselect and it's brother,
428 for hardcode select overriding.
429
4305.151 Mon Jul 6 05:41:57 CEST 2009
431 - backport to windows process emulation code again (patch by
432 Yasuhiro MATSUMOTO).
433 - slightly update Coro::MakeMaker.
434
4355.15 Tue Jun 30 10:28:06 CEST 2009
436 - deprecate Coro::Socket, document how to get ipv6 support via
437 AnyEvent::Socket instead.
438 - implement signal->wait ($cb) interface, similar to semaphores.
439 - work around SvOK not supporting getmagic, so we have to getmagic
440 to test for undef :/ (reported by Matthias Waldorf).
441 - load Coro::AnyEvent in all modules using AnyEvent.
442 - work around perl corrupting our internal data structures,
443 reported by Tokuhiro Matsuno.
444 - enable per-coroutine real and cpu time gathering
445 (Coro::State::enable_times).
446
4475.14 Wed Jun 24 01:37:48 CEST 2009
448 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
449 as this seems to have been a difficult concept (poll, sleep, idle,
450 idle_upto).
451 - add Coro::AnyEvent::readable/writable functions.
452 - clarify Coro::EV/Event/AnyEvent manpages.
453 - free per-thread global scalars in the thread calling ->cancel, to
454 avoid crashes when $_, $@ etc., are magical but some of those
455 had already been freed.
456 - "unexperimentalise" the callback interface for Coro::Semaphore.
457 - speed up ready queue management/context switching by using a linked
458 list instead of an array (~5-10%).
459 - implement "watch" command in Coro::Debug shells.
460 - for fun, implement time-slicing as an example in the manpage.
461 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
462 Coro::Event, use the normal AnyEvent handling instead of dieing
463 (the same is true for Coro::Handle).
464 - properly document Coro::EV::timed_io_once.
465 - avoid unneccessary ->cancel calls in Coro::Handle.
466 - maybe make it work on mingw32 with win32 backend
467 (based on patch by Yasuhiro Matsumoto).
468
4695.132 Fri May 29 09:00:39 CEST 2009
470 - do not keep a reference to the argument itself in
471 Coro::Semaphore::guard, as it could change later.
472 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
473 should support a prepare callback.
474
4755.131 Mon Mar 16 23:20:37 CET 2009
476 - implement and document Coro->suspend, Coro->resume.
477 - fix Coro::Select implementation to not (often) close
478 the passed file descriptors (testcase provided by pippijn).
479
4805.13 Mon Dec 15 21:51:42 CET 2008
481 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
482 - don't set diehook to C<undef> but instead to NULL, to avoid
483 spurious warnings.
484 - fix a lot of bugs in Coro::SemaphoreSet.
485 - Coro::SemaphoreSet will less often create a semaphore needlessly.
486 - add Coro::SemaphoreSet::count and wait methods.
487 - take advantage of the new Guard module.
488 - deprecate Coro::guard.
489 - try to fix the dreaded 01_unblock tests once more. I hate it when
490 testsuites need more fixing than the code they are supposed to test.
491 - croak in more cases when a required callback isn't resolvable.
492 - fix some minor issues in Coro::State->call/eval.
493 - use current coroutine context instead of a temporary one
494 when temporarily switching to another coroutine.
495 - do not call C-level on_destroy handlers during global destruction,
496 to avoid needless segfaults.
497
4985.12 Sun Dec 7 13:30:38 CET 2008
499 - add default config for MirOS, which seems to be a bug-to-bug
500 compatible fork of openbsd ("world domination by releasing
501 openbsd cvs before the openbsd folks do it" or so :).
502 - free_padlist did destroy the names pad, not good, but didn't
503 seem to bother perl - this could fix issues such as eval ""
504 inside a function called from multiple coroutines.
505 - use a different method to detect destruction time.
506 - be more careful when freeing padlists just before global
507 destruction.
508 - fixed and expanded the call/cc example.
509 - renamed _terminate to _coro_run.
510 - new method Coro::Channel->shutdown.
511 - try pthreads on openbsd <4.4 (broken sigaltstack, will
512 pthreads fare better?).
513 - be less picky about destroying "a" running coroutine.
514
5155.11 Tue Nov 25 21:49:05 CET 2008
516 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
517 working code with failed assertions, introducing more bugs than
518 it fixes, requiring elaborate workarounds :(
519
5205.1 Mon Nov 24 08:54:59 CET 2008
521 - wrote a small introductory tutorial - Coro::Intro.
522 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
523 - Coro::Select did errornously dup the file descriptors
524 and didn't work with all AnyEvent backends.
525 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
526 LWP data transfers.
527 - disassociate c contexts from coro objects - this is agruably more
528 correct, but mostly allows sharing of cctxs between coro and state
529 objects, for added memory savings and speed increases.
530 - bumped $Coro::POOL_RSS up to 32kb by default.
531 - no longer set the optype to OP_CUSTOM, as B::* understandably
532 doesn't like this very much (and we *are* a type of entersub).
533 - implement state cloning, just to prove that call/cc can be done.
534 - automatically load Coro::AnyEvent in Coro::Handle.
535 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
536 do not have this method (and don't need it either).
537 - speed up generic anyevent methods in Coro::Handle by using rouse
538 callbacks.
539 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
540 others. It also makes the debugger happier, as you can trace
541 through the idle threads now.
542 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
543 - eg/event was pretty broken.
544 - better 5.8.6 compatibility.
545
5465.0 Thu Nov 20 10:35:05 CET 2008
547 - NEW ARCHITECTURE: use the latest 4.x version if you experience
548 stability issues.
549 - bump API version to 7 - all dependents must be recompiled.
550 - removed timed_* functions - they were not being used anyways
551 and should be replaced by a more generic mechanism -
552 and were annoying to support anyways :)
553 - removed SemaphoreSet's waiter method - use sem method instead.
554 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
555 - async_pool did free a scalar value twice
556 ("Attempt to unreference...").
557 - fix a longstanding bug where calling terminate on a coro that
558 was waiting for a semaphore that was just becoming available
559 would cause a deadlock (semaphore would get into a state where
560 it was available but waiters were still blocked).
561 - calling throw on a coroutine that is waiting for a semaphore will
562 no longer make it acquire the semaphore (and thus leak a count).
563 - perl's process emulation is now not even theoretically supported
564 anymore.
565 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
566 conversion of callback-style to blocking-style.
567 - new methods $coro->schedule_to and ->cede_to, to specifically
568 schedule or cede to a specific coroutine.
569 - new function Coro::Semaphore::wait.
570 - use named constants in Coro::Channel (Richard Hundt).
571 - directly patch the entersub opcode calling SLF functions (cede,
572 transfer and so on). this does speed up context switching, but
573 more importanly, it frees us from the hardcoded behaviour of
574 entersub, so we might actually be able to return something from
575 those functions and atcually create new ones.
576 - take advantage of __builtin_frame_address on gcc.
577 - expose THX in coroapi (not sure whether this was a wise decision,
578 as "threaded" perls are running at half speed anyways).
579 - implement execute_slf (schedule-like-function) interface that makes
580 it possible to implement schedule-like-functions in XS.
581 - use new SLF interface to massively speed up Coro::EV by roughly a
582 factor of two.
583 - used new SLF interface to massively speed up Coro::Semaphore by a
584 factor of three.
585 - used new SLF interface to speed up Coro::AIO by roughly a factor of
586 four and reduce its memory usage considerably.
587 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
588 for a nice speedup and vastly more correct behaviour. Also implement
589 a new method "sem" to get at the underlying semaphore object.
590 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
591 (in comparison) 20-40% speedup.
592 - used new SLF interface to reimplement Coro::Signal gaining
593 some unknown (because I was too lazy), but certain, speedup, and also
594 making signals reliable for the first time.
595 - used new SLF interface and other optimisations to speed up async_pool
596 by a factor of two. It also doesn't rely on perl's exception mechanism
597 to exit anymore. The overhead for terminating an async_pool, coro over
598 a normal async is now very small.
599 - sped up coroutine creation/destruction by 40%.
600 - forgot to include Coro/libcoro/README in the dist for all these years.
601 - work around a freebsd pthreads bug (manual testcancel is required as
602 pthread_cond_wait isn't a cancellation point on freebsd).
603 - use new rouse functions to speed up and simplify Coro::BDB.
604 - make "prefer perl native functions" work with threaded perls.
605 - condense Coro::Debug ps output, hint at v and w flags.
606 - (libcoro) lots of minor cleanups and portability improvements.
607
6084.914 Wed Nov 19 12:54:18 CET 2008
609 - fix a disastrous bug in the readline optimisation
610 introduced in 4.801.
611
6124.913 Sat Nov 15 07:58:28 CET 2008
613 - async_pool did free a scalar value twice
614 ("Attempt to unreference...").
615
6164.912 Thu Nov 13 18:31:23 CET 2008
617 - minor cleanups.
618 - use much larger stacks on linux and perl < 5.8.8.
619 - Coro::Debug::new_unix_server did not unlink the socket
620 when destroyed.
621
6224.911 Tue Nov 11 04:26:17 CET 2008
623 - "port" to threaded perls.
624
6254.91 Mon Nov 10 05:36:38 CET 2008
626 - the ->throw exception object no longer leaks.
627 - creating a new cctx leaked a scopestack entry (memleak).
628 - new coroutines didn't get created with a zero flags field
629 (unknown impact).
630 - calling ->throw on a not-yet-started coroutine should now work
631 instead of being ignored.
632 - ->throw is now supported on Coro::State objects.
633 - clean up cctx creation code a bit.
634 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
635
6364.9 Sat Nov 8 17:45:27 CET 2008
637 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
638 - (libcoro) no longer leak threads in the experimental pthread backend,
639 also speed it up considerably.
640 - (libcoro) do not rely on makecontext passing void *'s unscathed.
641 - fix compiletime dependencies on libcoro in the Makefile.
642 - cctx_count wasn't always updated properly.
643 - Coro::State::cctx_stacksize wasn't applied correctly.
644 - new function Coro::State::cctx_max_idle.
645 - the default max number of idle C contexts is now 4.
646 - (libcoro) try harder to get _setjmp/_longjmp.
647 - (libcoro) cleanup and extend the libcoro API to officially
648 allow the creation of empty source contexts.
649 - very experimental workaround for the totally broken netbsd platform.
650 - tried to hack around openbsd bugs.
651
6524.804 Wed Nov 5 16:36:00 CET 2008
653 - Coro::Debug::new_unix_server would not create a non-blocking listening
654 socket, sometimes causing freezes.
655 - (libcoro) fix misaligned stack points for setjmp and assembly
656 methods, which can cause crashes on x86/x86_64 with a sufficiently
657 aggressive compiler.
658 - new function: Coro::Debug::new_tcp_server.
659 - move ->throw into the Coro class because it only works on coro objects.
660
6614.803 Mon Nov 3 17:16:12 CET 2008
662 - (libcoro) use a global asm statement to become independent of gcc
663 otpimisations for CORO_ASM (thanks to pippijn for the idea).
664 - try to workaround yet another broken bsd, this time dragonfly.
665
6664.802 Thu Oct 30 10:56:12 CET 2008
667 - support -fno-omit-frame-pointer on x86 with the assembly method.
668 - tune 01_unblock.t tests a bit.
669
6704.801 Wed Oct 22 18:33:37 CEST 2008
671 - improve readline speed for very long "lines".
672 - backport to 5.8.8.
673
6744.8 Thu Oct 2 13:34:40 CEST 2008
675 - new function Coro::AIO::aio_wait.
676 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
677 - greatly speed up and reduce memory usage of Coro::AIO requests.
678 - implement some other µ-optimisations.
679
6804.749 Mon Sep 29 14:40:12 CEST 2008
681 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
682
6834.748 Sat Sep 27 14:03:19 CEST 2008
684 - implement, but do not document, PerlIO::cede(granularity).
685 - Coro::Storable forgot to wrap Storable::pstore.
686 - work around the multitude of leaks and memory corruption
687 bugs in PerlIO::via by using our own C-level perliol. As a side
688 effect, Coro::Storable is now much, much, much faster.
689
6904.747 Tue Sep 23 01:59:41 CEST 2008
691 - fix a per-cv memleak (one empty array was leaked per
692 code reference).
693 - avoid a crash in coro->call|rss when the coroutine was already
694 destroyed (most noticably when using Coro::Debug::ps :)
695 - also protect *Storable::FILE.
696 - push up default storable granularity to 20ms.
697
6984.746 Sun Sep 21 03:22:20 CEST 2008
699 - be more insistent on locking Storable against reentrancy
700 in Coro::Storable.
8 - move swap_def?v and throw to Coro::State, as documented. 701 - move swap_def?v and throw to Coro::State, as documented.
9 702
104.745 Thu Jul 24 00:14:38 CEST 2008 7034.745 Thu Jul 24 00:14:38 CEST 2008
11 - remove debugging code related to MgPV_nolen_const, also try to 704 - remove debugging code related to MgPV_nolen_const, also try to
12 make it compile with perl 5.8.6 (yes, apple apparently loves 705 make it compile with perl 5.8.6 (yes, apple apparently loves
13 outdated software). Reported by John S. 706 outdated software). Reported by John S.
14 707
154.744 Tue Jul 8 22:06:35 CEST 2008 7084.744 Tue Jul 8 22:06:35 CEST 2008
16 - correctly provide default DIE/WARN handlers as documented. 709 - correctly provide default DIE/WARN handlers as documented.
17 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 710 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
18 implementations inside perl work fine for us. 711 implementations inside perl work fine for us.
19 712
204.743 Mon Jun 16 00:21:57 CEST 2008 7134.743 Mon Jun 16 00:21:57 CEST 2008
21 - when using Coro::EV without running EV::loop it could 714 - when using Coro::EV without running EV::loop it could
… …
43 - sprinkle "no warnings" freely over everything, also suppress 736 - sprinkle "no warnings" freely over everything, also suppress
44 warnings for some other modules. 737 warnings for some other modules.
45 - fix typo in WSAEWOULDBLOCK. 738 - fix typo in WSAEWOULDBLOCK.
46 739
474.72 Sun May 25 05:14:36 CEST 2008 7404.72 Sun May 25 05:14:36 CEST 2008
48 - tweak META.yaml a bit, unfortunately, there is no documented way 741 - tweak META.yaml a bit, unfortunately, there is no documented way
49 to have optional dependencies with CPAN. doh :( 742 to have optional dependencies with CPAN. doh :(
50 - avoid running some tests on windows because they would fail due to 743 - avoid running some tests on windows because they would fail due to
51 perl bug (broken fork, broken pipes...). 744 perl bug (broken fork, broken pipes...).
52 - work around perl on windows bugs where perl returns undocumented 745 - work around perl on windows bugs where perl returns undocumented
53 error codes for sysread, syswrite etc. by taking advantage 746 error codes for sysread, syswrite etc. by taking advantage
… …
57 - use unix domain sockets in testsuite to work around 750 - use unix domain sockets in testsuite to work around
58 common perl implementation bugs on widows (they are emulated by 751 common perl implementation bugs on widows (they are emulated by
59 tcp sockets on windows. ugh.) 752 tcp sockets on windows. ugh.)
60 753
614.71 Sat May 24 20:01:27 CEST 2008 7544.71 Sat May 24 20:01:27 CEST 2008
62 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 755 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
63 - take advantage of async name resolution of AnyEvent::Util. 756 - take advantage of async name resolution of AnyEvent::Util.
64 - work around brutal inet_aton override in Coro::LWP. 757 - work around brutal inet_aton override in Coro::LWP.
65 - take advantage of the readyhook in Coro::EV, for smoother 758 - take advantage of the readyhook in Coro::EV, for smoother
66 scheduling. 759 scheduling.
67 760
684.7 Sun May 11 00:32:19 CEST 2008 7614.7 Sun May 11 00:32:19 CEST 2008
69 - completely reworked the Coro manpage. 762 - completely reworked the Coro manpage.
70 - added Coro::AnyEvent, generic event loop integration. 763 - added Coro::AnyEvent, generic event loop integration.
71 - implement cancel, ready and kill commands in Coro::Debug. 764 - implement cancel, ready and kill commands in Coro::Debug.
72 - document find_coro in Coro::Debug. 765 - document find_coro in Coro::Debug.
73 - incompatible change: rename has_stack to has_cctx. 766 - incompatible change: rename has_stack to has_cctx.
74 - Coro::AIO and Coro::BDB no longer force event model detection, 767 - Coro::AIO and Coro::BDB no longer force event model detection,
75 use AnyEvent::AIO and AnyEvent::BDB. 768 use AnyEvent::AIO and AnyEvent::BDB.
… …
199 892
2004.11 Thu Oct 11 02:40:24 CEST 2007 8934.11 Thu Oct 11 02:40:24 CEST 2007
201 - port to threaded perls. 894 - port to threaded perls.
202 895
2034.1 Thu Oct 11 02:38:16 CEST 2007 8964.1 Thu Oct 11 02:38:16 CEST 2007
204 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 897 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
205 be local to each coro (see Coro::State). 898 be local to each coro (see Coro::State).
206 - incompatible change: for very deep reasons, cede and cede_notself 899 - incompatible change: for very deep reasons, cede and cede_notself
207 cannot return anything, so nothing will be returned. 900 cannot return anything, so nothing will be returned.
208 - possibly bring back 5.10 compatibility (untested). 901 - possibly bring back 5.10 compatibility (untested).
209 - work around stupid (and wrong) warning on 5.10 :(. 902 - work around stupid (and wrong) warning on 5.10 :(.
… …
281 but stateful semantics. 974 but stateful semantics.
282 - fixed a lot of typos in Coro.pm (patch submitted by David 975 - fixed a lot of typos in Coro.pm (patch submitted by David
283 Steinbrunner, which applied flawlessly). 976 Steinbrunner, which applied flawlessly).
284 977
2853.6 Sat Apr 14 17:13:31 CEST 2007 9783.6 Sat Apr 14 17:13:31 CEST 2007
286 - added some bugfixes to get eg/myhttpd working again. 979 - added some bugfixes to get eg/myhttpd working again.
287 - added Coro::Storable for often-cede'ing freeze/thaw. 980 - added Coro::Storable for often-cede'ing freeze/thaw.
288 - try to do a clean exit when a coroutine calls exit 981 - try to do a clean exit when a coroutine calls exit
289 (EXPERIMENTAL). 982 (EXPERIMENTAL).
290 - got rid of indirect call through _coro_init. 983 - got rid of indirect call through _coro_init.
291 - updated the partly antique examples in eg/ to 984 - updated the partly antique examples in eg/ to

Diff Legend

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