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

Comparing Coro/Changes (file contents):
Revision 1.294 by root, Fri Nov 30 15:26:41 2007 UTC vs.
Revision 1.681 by root, Fri Jul 19 15:29:46 2019 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 4TODO: channel->maxsize(newsize)?
5 (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?
10
116.55 Fri Jul 19 17:28:09 CEST 2019
12 - Coro::RWLock woke up only one reader instead of all after a wrlock
13 (analyzed and patch by Maxim Skorobogatov).
14 - slightly improved documentation of Coro::RWLock and RWLock testcase.
15 - quickly (re)-ported to minix 3.3 before minix crashed again.
16
176.54 Sun Dec 16 11:06:07 CET 2018
18 - reinstate Coro::Select::select prototype - it breaks more than it fixes.
19
206.53 Sun Dec 16 10:33:15 CET 2018
21 - the autogenerated prototypes of Coro::AIO now add ";" at the end
22 to avoid generating named unary operators and messing up parsing
23 (mostly affects aio_mlockall 4 | 8 which was parsed as
24 (aio_mlockall 4) | 8).
25 - do not use a prototype for select in Coro::Select, as CORE::select also
26 doesn't have one.
27
286.52 Tue Aug 14 18:49:43 CEST 2018
29 - libcoro update: allow sharing of coro threads using the pthreads
30 backend among different pthreads, as is required by Coro::Multicore.
31 - always allow ->call and ->eval when the coro is the current coroutine. this
32 makes it possible, among other things, to call these on $Coro::main.
33 - minor code cleanups.
34
356.514 Thu Aug 31 18:25:31 CEST 2017
36 - safe_cancel no longer croaks when called on already-destroyed
37 thread, but instead "just works".
38
396.513 Sat Jul 15 01:18:51 CEST 2017
40 - experimental 5.22 comaptibility changes (PadlistNAMES).
41 - removed attributes example and made sure the other examples
42 still work.
43 - fix some statevar types for newer perls, should be of no consequence.
44
456.512 Fri Jul 14 05:03:17 CEST 2017
46 - seems to pass test suite down till 5.8.9, so make this the
47 minimum requirement in Makefile.PL.
48 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
49 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
50 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
51 fortified longjmp variants in some libcs.
52 - re-try JIT allocation without PROT_EXEC, to hopefully improve
53 portability (still doesn't seem to work around broken selinux
54 mmap everywhere).
55 - speling fixes (Xavier Guimard).
56
576.511 Sun Jun 26 23:44:50 CEST 2016
58 - make it compile with threaded perls again.
59 - simplify and speed up __DIE__ and __WARN__ handling.
60
616.51 Sat Jun 25 20:57:20 CEST 2016
62 - 6.5 release didn't compile on most older perls due to
63 a logic error enabling experimental 5.24 code for all versions.
64 - since various broken or deficient patches float around
65 to work around the vtbl API breakage in perl, let's provide
66 our own workaround, which seems to be much less invasive
67 and more compatible than the approaches seen so far. YMMV.
68 - work around assertion failure in perl_destruct on 5.24
69 (analyzed by Dave Mitchell).
70
716.5 Wed Jun 22 22:23:50 CEST 2016
72 - swap_sv swapping order was not symmetrical, causing
73 wrong swaps when swap_sv was used multiple times on the
74 same sv in the same thread.
75 - swap_sv calls can now be undone by calling it again
76 with the same variables.
77 - swap_sv calls will now be undone in async_pool threads.
78 - split Coro::Semaphore::up/adjust into separate xs functions
79 for better error reporting, at slight codesize increase.
80 - (libcoro) arm assembly support, please test and report.
81 - adjust to PL_savestack changes in perl 5.24 (adapted from
82 the debian patch, which unfortunately gets it wrong).
83
846.49 Sat Oct 17 01:40:12 CEST 2015
85 - throwing an exception to a thread waiting in
86 Coro::Handle using Coro::EV did not stop the watchers,
87 causing the next call to fail (testcase by Martin Pritchard).
88 - bump minimum perl version to 5.10.
89
906.48 Sun Oct 4 19:03:51 CEST 2015
91 - fix memory leak when Coro::Handle uses Coro::EV internally
92 (testcase by Sten Sten).
93 - update code to libev 4 API (internally, Coro still used the
94 version 3 API calls).
95 - remove unused hv_sig.
96
976.47 Sat Jul 11 03:58:20 CEST 2015
98 - swap_svs - NVs can be in the SV head beginning with 5.20.
99
1006.46 Tue Jun 30 14:36:30 CEST 2015
101 - restore portability to perls with windows fork emulation
102 (patch by Petr Písař).
103
1046.45 Tue Jun 30 01:40:08 CEST 2015
105 - restore portability to perl 5.16 and below.
106
1076.44 Tue Jun 30 00:41:54 CEST 2015
108 - avoid segfaulting (or worse) when tracing a canceled thread.
109 - the Event module silently broke it's public hook API, causing
110 failures on perls compiled with -Duselongdouble. Requiring
111 the latest version of Event because I don't know which release
112 changed this (it's not mentioned in the ChangeLog).
113 - allow xs level enterleave hooks via CoroAPI.
114
1156.43 Thu Jun 4 15:38:14 CEST 2015
116 - use stability canary.
117 - port to stableperl-5.22.0-1.001.
118 - update libecb.
119
1206.42 Wed Feb 11 20:29:52 CET 2015
121 - Coro::SemaphoreSet->try did not actually work (analyzed by
122 SATO Kentaro).
123 - upgrade libecb to be C11 compliant.
124
1256.41 Sat Sep 6 22:08:46 CEST 2014
126 - restore portability to perl 5.8.x.
127 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
128 up some memory, but fixes "use feature" and similar modules
129 when used inside a Coro.
130 - allow zero as argument to Coro::Channel to mean the same thing
131 as no argument. this works with older versions as well,
132 but wasn't legal until now.
133 - slightly better c header file detection.
134
1356.39 Mon Jun 2 00:00:08 CEST 2014
136 - work around more incompatible changes in 5.20.
137
1386.38 Sun Jun 1 21:54:23 CEST 2014
139 - check that perl slots actually have enough space to hold
140 interpreter variables.
141 - untested port to perl 5.19 (64 bit tmps indices) (reported
142 by Andreas König).
143 - croak when cancel is called without a thread context.
144
1456.37 Tue Mar 4 13:27:33 CET 2014
146 - *sigh*, removed leftover debugging code from debugging a
147 perl bug, of all things.
148
1496.36 Tue Mar 4 07:11:59 CET 2014
150 - semaphores would not clear the destroy hook when interrupted
151 by ->throw, causing segfaults or worse.
152 - ->throw on a thread waiting for a semaphore did not acquire
153 the semaphore, but also didn't wake up other waiters,
154 possibly causing a deadlock.
155 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
156 to report a full stacktrace, which should help find out
157 where the actual call is.
158 - "atomically" destroy data in slf_destroy, because it is
159 the right thing to do, just in case.
160 - disable ecb.h libm dependency, to compile on ancient systems
161 or under adverse conditions.
162
1636.33 Mon Nov 18 11:26:27 CET 2013
164 - do not crash when freeing padlists with holes (in 5.18).
165 - tentative SVt_BIND 5.19 port/fix.
166
1676.32 Tue Nov 5 15:35:35 CET 2013
168 - use a new algorithm to derive padlists for perl 5.18. The old
169 one could lead to 0-pointer accesses inside perl (reported
170 by Darin McBride).
171
1726.31 Thu May 9 07:39:48 CEST 2013
173 - Coro::AIO requests would crash if the thread was ready'd
174 while the request was ongoing.
175
1766.29 Wed May 8 02:55:18 CEST 2013
177 - when an on_destroy handler destructs the coro currently being
178 destructed a perl scalar could be accessed after being freed,
179 likely causing a crash.
180
1816.28 Wed Mar 6 06:58:02 CET 2013
182 - clean remnants of existing __DIE__ and __WARN__ handlers so
183 they lose their magic and will not cause segfaults later
184 (testcase by Andrey Sagulin).
185 - improved Coro::State documentation a bit.
186 - Coro::Debug::command now flushes the output.
187 - add hack detection code for x32 abi, because the braindead slugs
188 who designed that made it look exactly like x86_64 without
189 providing proper compile time symbols to test for it. as a result,
190 this detection cannot work reliably.
191 - valgrind stack registering was broken.
192 - do not rely on Time::HiRes anymore in Coro::Debug.
193
1946.23 Fri Dec 7 23:36:37 CET 2012
195 - use experimental fiber implementation on native windows
196 perls.
197 - use sizeof (void *) as multiplication factor for stack sizes,
198 to accomodate the totally braindamaged microsoft 64 bit "os".
199 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
200 activeperl 5.16 crashes when PerlIO_define_layer is called due
201 to some bug in the perl dll, strawberry perl at least passes
202 the testsuite.
203 - implement Coro::Handle->peeraddr/host/port, for slightly
204 improved compatibility with LWP.
205 - implement 5.17 compatibility by almost blindly applying a
206 good-looking patch by Father Chrysostomos.
207 - move stack management functions into libcoro 3.
208 - libcoro version 3 "released".
209 - support magic values as timed_io_once args.
210 - recommend AnyEvent 7+ or EV 4+, also require EV
211 version 4 or newer for Coro::EV.
212
2136.10 Tue Oct 9 01:14:27 CEST 2012
214 - updated ecb.h, it had a typo that caused it to not compile on many
215 big endian systems (reported by many people).
216 - disable memory fences in ecb.h to improve portability.
217
2186.09 Sat Oct 6 23:25:02 CEST 2012
219 - Coro::EV I/O watchers were not interruptible by exceptions
220 (Coro::State::throw) (testcase by sten).
221 - ->throw now puts threads into the ready queue, as this seems to
222 be expected by existing code, and code that doesn't cope with spurious
223 wakeups needs fixing anyway.
224 - use fd -1 in mmap.
225 - cast I32 to int in error message printf.
226 - warn about broken so-called "hardened" kernels.
227
2286.08 Fri Apr 13 12:05:47 CEST 2012
229 - be more aggressive about exiting like perl does - formerly,
230 exiting from the non-main thread would not execute END blocks.
231
2326.07 Fri Nov 11 21:21:48 CET 2011
233 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
234 - update ecb.h.
235
2366.06 Mon Aug 8 23:59:48 CEST 2011
237 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
238 backend, which is an order of magnitude slower. unfortunately, cygwins
239 pthread implementation isn't very complete either, so allocate the stack
240 twice just to be sure.
241 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
242 in its ithreaded perl - best recompile cygwin and use the 'w'indows
243 backend for much better performance. also disable ithreads for
244 even better performance...).
245
2466.05 Thu Aug 4 21:36:36 CEST 2011
247 - blush, condvar values would not be propagated from send to recv anymore
248 (reported by Chip Salzenberg).
249 - use exponential increase for the readline buffer length in
250 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
251
2526.04 Wed Aug 3 17:15:45 CEST 2011
253 - use even more efficient and more compatible condvars for
254 compatibility to AnyEvent 6.x :)
255 - more inconsequential ecb.h updates.
256
2576.03 Wed Aug 3 11:41:30 CEST 2011
258 - change how Coro patches AnyEvent condvars for compatibility to
259 AnyEvent 6.x.
260 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
261 hasn't been included.
262
2636.02 Wed Jul 13 04:35:19 CEST 2011
264 - "improve portability to Gentoo" - gentoo manages to put perl variables
265 in memory areas that are farther than 2gb apart, which the jit couldn't
266 handle and barfed. now it's just a bit slower on gentoo and similar
267 systems.
268
2696.01 Sun Jul 3 12:31:14 CEST 2011
270 - workarounds are good, but the test for whether pthreads are used
271 was not good. this one should be better.
272 - check differently whether gcc generates cfi instructions itself.
273
2746.0 Wed Jun 29 19:43:35 CEST 2011
275 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
276 destroyed and cleaned up automatically (e.g. async { schedule }).
277 - implement a JIT compiler for part of the thread switch code,
278 which gives a 50% speed improvement on threaded perls, and
279 about 4% on non-threaded perls (so threaded perls now finally
280 reach about half the speed of non-threaded perls).
281 - slightly modernise Coro::Intro, add section about rouse functions.
282 - avoid DEFSV and ERRSV, giving another 10% improvement
283 in thread switching.
284 - Coro::State->is_destroyed is now called is_zombie.
285 - implement a Coro->safe_cancel method that might fail, but
286 cancels in a "safer" way if it succeeds.
287 - add preliminary support for DEBUGGING perls.
288 - get rid of two hash-accesses when initialising a new Coro - this
289 speeds up coro creation by almost a factor of two.
290 - croak when a coro that is being cancelled tries to block
291 (e.g. while executing a guard block), instead of crashing or
292 deadlocking.
293 - use a more robust and also faster method to identify Coro::State
294 objects - speeds up everything a bit.
295 - implement Coro->cancel in XS for a 20% speed improvement, and to
296 be able to implement mutual cancellation.
297 - speed up context switches by a percent or two by more efficiently
298 allocating context stack entries.
299 - implement Coro->join and Coro->on_destroy in XS for a speedup and
300 a reduction in memory use.
301 - cancelling a coro while it itself is cancelling another coro is
302 now supported and working, instead of triggering an assertion.
303 - be a bit more crash-resistant when calling (buggy) on_destroy
304 callbacks (best effort).
305 - move on_destroy into the slf_frame, to allow extension slf
306 functions to have destructors.
307 - get rid if coro refcounting - simply crash in other interpreter
308 threads by nulling the pointers on clone.
309 - simplify warn/die hook handling when loading Coro - the convoluted
310 logic seems to be no longer neccessary.
311 - use libecb instead of our own home-grown gcc hacks.
312 - document alternatives to Coro::LWP. Please use them :)
313 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
314 sigaltstack. Really. wine suffers from it, erlang suffers from it,
315 and it's known since at least 2006.
316
3175.372 Wed Feb 23 06:14:30 CET 2011
318 - apparently mingw doesn't provide a working gettimeofday, try to
319 work around that by relying on Time::HiRes (indirectly brought to
320 my attention by Max Maischein).
321 - fix some portability issues when Time::HiRes was used.
322
3235.371 Mon Feb 21 14:36:08 CET 2011
324 - backport to windows process emulation code again.
325
3265.37 Sat Feb 19 07:49:44 CET 2011
327 - add a big "Coro thread life cycle" section to "man Coro".
328 - try a tentative workaround against the breakage that 5.13 has
329 introduced without depreciation period. sigh.
330 - no longer use Time::HiRes if gettimeofday is available, which
331 saves quite a lot of memory.
332
3335.36 Sun Feb 13 05:33:41 CET 2011
334 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
335 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
336 method call.
337 - undocument Coro::Timer::sleep and obsolete whole module.
338 - optimise Coro::Timer::timeout memory and cpu usage.
339 - slightly updated Coro::Intro for recent changes.
340 - do not initialise PL_dirty anymore.
341
3425.25 Thu Nov 11 01:08:39 CET 2010
343 - try a different approach on netbsd - netbsd 5 finally has marginally
344 working pthreads, but still broken ucontext/sigaltstack.
345 - openbsd 4.8 finally got their act together, Coro works out of the box
346 with asm, setjmp and pthreads (no change, just informational).
347
3485.24 Sat Oct 23 11:27:12 CEST 2010
349 - port to the EV 4.0 API.
350 - work around bugs in mingw32, making strawberry perl work
351 out of the box.
352 - correctly modify Coro::AIO function prototypes
353 so that they reflect the "no optional parameters" rule.
354 - "ported" libcoro to C++.
355
3565.23 Mon May 17 18:50:42 CEST 2010
357 - be more resistant to ordering changes when initialising
358 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
359 Waldorf).
360 - document that perl 5.12 deliberately removed support for cloning.
361
3625.22 Wed Apr 14 03:55:35 CEST 2010
363 - correctly return udnef on errors in Coro::Handle::read/write
364 (testcase by Marc Mims).
365 - convert Coro::Util into a "perl compatibility wrapper" - the functions
366 are less useful now, but are drop-in replacements for existing
367 functions, listing better alternatives in the documentation. This also
368 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
369 with Coro::Util::inet_aton.
370 - do not override $Coro::idle unconditionally in Coro.pm, as other
371 modules could have provided their own idle coro already
372 (for exmaple, Coro::AnyEvent).
373 - fix Coro::Util::gethost* functions.
374 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
375
3765.21 Wed Dec 16 07:19:51 CET 2009
377 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
378 together.
379 - add some examples on how to combine other event loops with Coro in
380 Coro::AnyEvent, and how to run it (and not to block). Seems to be
381 the most common source of confusion.
382 - try to catch people naively blocking in an event callback.
383 - work around the perl filehandle bug issue in conjunction with
384 older common::sense (as indirectly pointed out by ZSystem).
385 - clarify the "not from signal handlers" section.
386
3875.2 Sun Oct 4 14:54:24 CEST 2009
388 - Coro::Storable destroyed the prototypes of the functions it wrapped.
389 - export rouse_cb and rouse_wait by default now.
390 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
391 - new method $state->swap_sv.
392 - added section on "windows process emulation" to the manpage,
393 after a not-so-fruitful (nor-friendly) "discussion" with chip
394 salzenberg (discussion implies arguments, but his only arguments
395 were ad-hominems, one wonders why he started it in the first
396 place). I hope this explains it well enough for him to understand,
397 and maybe well enough for others to understand.
398 - use common::sense everywhere now.
399 - idle callbacks are no longer supported, use idle coros instead.
400 - print a thread listing when a deadlock is detected.
401
4025.17 Sat Aug 22 23:09:31 CEST 2009
403 - work around a bug in the perl debugger causing crashes
404 when running under the debugger by marking _pool_handler as nodebug.
405 - speed up Coro::async considerably.
406 - try some hacks to get netbsd to work "more often" - their broken
407 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
408 - convert Coro to AE 5.0 API.
409
4105.162 Tue Jul 28 04:04:03 CEST 2009
411 - perl 5.8.2 is now minimum requirement.
412 - skip t/19_handle.t on broken windows perls.
413
4145.161 Wed Jul 22 04:47:38 CEST 2009
415 - Coro::AnyEvent::poll could have a different prototype when EV was
416 used as backend (analyzed by Tatsuhiko Miyagawa).
417 - Coro::AnyEvent errornously initialised the event loop when loaded,
418 not on demand.
419 - try to workaround rare */t/01_unblock.t failures.
420
4215.16 Tue Jul 21 01:44:37 CEST 2009
422 - Coro::AnyEvent failed to hook into the event loop
423 when no threads had been readied between detecting
424 the event loop and actually running it.
425 - considerably speed up Coro::Select by taking avdantage
426 of AnyEvent > 4.8 and some other optimisations.
427 - implement paragraph readline mode in Coro::Handle
428 (based on patches by Zsbán Ambrus).
429 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
430 and analyzed by Yasuhiro MATSUMOTO).
431 - clarified libcoro license and copyright.
432 - someone stole my EXTRA_META!!!
433 - implement Coro::Select::patch_pp_sselect and it's brother,
434 for hardcode select overriding.
435
4365.151 Mon Jul 6 05:41:57 CEST 2009
437 - backport to windows process emulation code again (patch by
438 Yasuhiro MATSUMOTO).
439 - slightly update Coro::MakeMaker.
440
4415.15 Tue Jun 30 10:28:06 CEST 2009
442 - deprecate Coro::Socket, document how to get ipv6 support via
443 AnyEvent::Socket instead.
444 - implement signal->wait ($cb) interface, similar to semaphores.
445 - work around SvOK not supporting getmagic, so we have to getmagic
446 to test for undef :/ (reported by Matthias Waldorf).
447 - load Coro::AnyEvent in all modules using AnyEvent.
448 - work around perl corrupting our internal data structures,
449 reported by Tokuhiro Matsuno.
450 - enable per-coroutine real and cpu time gathering
451 (Coro::State::enable_times).
452
4535.14 Wed Jun 24 01:37:48 CEST 2009
454 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
455 as this seems to have been a difficult concept (poll, sleep, idle,
456 idle_upto).
457 - add Coro::AnyEvent::readable/writable functions.
458 - clarify Coro::EV/Event/AnyEvent manpages.
459 - free per-thread global scalars in the thread calling ->cancel, to
460 avoid crashes when $_, $@ etc., are magical but some of those
461 had already been freed.
462 - "unexperimentalise" the callback interface for Coro::Semaphore.
463 - speed up ready queue management/context switching by using a linked
464 list instead of an array (~5-10%).
465 - implement "watch" command in Coro::Debug shells.
466 - for fun, implement time-slicing as an example in the manpage.
467 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
468 Coro::Event, use the normal AnyEvent handling instead of dieing
469 (the same is true for Coro::Handle).
470 - properly document Coro::EV::timed_io_once.
471 - avoid unneccessary ->cancel calls in Coro::Handle.
472 - maybe make it work on mingw32 with win32 backend
473 (based on patch by Yasuhiro Matsumoto).
474
4755.132 Fri May 29 09:00:39 CEST 2009
476 - do not keep a reference to the argument itself in
477 Coro::Semaphore::guard, as it could change later.
478 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
479 should support a prepare callback.
480
4815.131 Mon Mar 16 23:20:37 CET 2009
482 - implement and document Coro->suspend, Coro->resume.
483 - fix Coro::Select implementation to not (often) close
484 the passed file descriptors (testcase provided by pippijn).
485
4865.13 Mon Dec 15 21:51:42 CET 2008
487 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
488 - don't set diehook to C<undef> but instead to NULL, to avoid
489 spurious warnings.
490 - fix a lot of bugs in Coro::SemaphoreSet.
491 - Coro::SemaphoreSet will less often create a semaphore needlessly.
492 - add Coro::SemaphoreSet::count and wait methods.
493 - take advantage of the new Guard module.
494 - deprecate Coro::guard.
495 - try to fix the dreaded 01_unblock tests once more. I hate it when
496 testsuites need more fixing than the code they are supposed to test.
497 - croak in more cases when a required callback isn't resolvable.
498 - fix some minor issues in Coro::State->call/eval.
499 - use current coroutine context instead of a temporary one
500 when temporarily switching to another coroutine.
501 - do not call C-level on_destroy handlers during global destruction,
502 to avoid needless segfaults.
503
5045.12 Sun Dec 7 13:30:38 CET 2008
505 - add default config for MirOS, which seems to be a bug-to-bug
506 compatible fork of openbsd ("world domination by releasing
507 openbsd cvs before the openbsd folks do it" or so :).
508 - free_padlist did destroy the names pad, not good, but didn't
509 seem to bother perl - this could fix issues such as eval ""
510 inside a function called from multiple coroutines.
511 - use a different method to detect destruction time.
512 - be more careful when freeing padlists just before global
513 destruction.
514 - fixed and expanded the call/cc example.
515 - renamed _terminate to _coro_run.
516 - new method Coro::Channel->shutdown.
517 - try pthreads on openbsd <4.4 (broken sigaltstack, will
518 pthreads fare better?).
519 - be less picky about destroying "a" running coroutine.
520
5215.11 Tue Nov 25 21:49:05 CET 2008
522 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
523 working code with failed assertions, introducing more bugs than
524 it fixes, requiring elaborate workarounds :(
525
5265.1 Mon Nov 24 08:54:59 CET 2008
527 - wrote a small introductory tutorial - Coro::Intro.
528 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
529 - Coro::Select did errornously dup the file descriptors
530 and didn't work with all AnyEvent backends.
531 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
532 LWP data transfers.
533 - disassociate c contexts from coro objects - this is agruably more
534 correct, but mostly allows sharing of cctxs between coro and state
535 objects, for added memory savings and speed increases.
536 - bumped $Coro::POOL_RSS up to 32kb by default.
537 - no longer set the optype to OP_CUSTOM, as B::* understandably
538 doesn't like this very much (and we *are* a type of entersub).
539 - implement state cloning, just to prove that call/cc can be done.
540 - automatically load Coro::AnyEvent in Coro::Handle.
541 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
542 do not have this method (and don't need it either).
543 - speed up generic anyevent methods in Coro::Handle by using rouse
544 callbacks.
545 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
546 others. It also makes the debugger happier, as you can trace
547 through the idle threads now.
548 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
549 - eg/event was pretty broken.
550 - better 5.8.6 compatibility.
551
5525.0 Thu Nov 20 10:35:05 CET 2008
553 - NEW ARCHITECTURE: use the latest 4.x version if you experience
554 stability issues.
555 - bump API version to 7 - all dependents must be recompiled.
556 - removed timed_* functions - they were not being used anyways
557 and should be replaced by a more generic mechanism -
558 and were annoying to support anyways :)
559 - removed SemaphoreSet's waiter method - use sem method instead.
560 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
561 - async_pool did free a scalar value twice
562 ("Attempt to unreference...").
563 - fix a longstanding bug where calling terminate on a coro that
564 was waiting for a semaphore that was just becoming available
565 would cause a deadlock (semaphore would get into a state where
566 it was available but waiters were still blocked).
567 - calling throw on a coroutine that is waiting for a semaphore will
568 no longer make it acquire the semaphore (and thus leak a count).
569 - perl's process emulation is now not even theoretically supported
570 anymore.
571 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
572 conversion of callback-style to blocking-style.
573 - new methods $coro->schedule_to and ->cede_to, to specifically
574 schedule or cede to a specific coroutine.
575 - new function Coro::Semaphore::wait.
576 - use named constants in Coro::Channel (Richard Hundt).
577 - directly patch the entersub opcode calling SLF functions (cede,
578 transfer and so on). this does speed up context switching, but
579 more importanly, it frees us from the hardcoded behaviour of
580 entersub, so we might actually be able to return something from
581 those functions and atcually create new ones.
582 - take advantage of __builtin_frame_address on gcc.
583 - expose THX in coroapi (not sure whether this was a wise decision,
584 as "threaded" perls are running at half speed anyways).
585 - implement execute_slf (schedule-like-function) interface that makes
586 it possible to implement schedule-like-functions in XS.
587 - use new SLF interface to massively speed up Coro::EV by roughly a
588 factor of two.
589 - used new SLF interface to massively speed up Coro::Semaphore by a
590 factor of three.
591 - used new SLF interface to speed up Coro::AIO by roughly a factor of
592 four and reduce its memory usage considerably.
593 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
594 for a nice speedup and vastly more correct behaviour. Also implement
595 a new method "sem" to get at the underlying semaphore object.
596 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
597 (in comparison) 20-40% speedup.
598 - used new SLF interface to reimplement Coro::Signal gaining
599 some unknown (because I was too lazy), but certain, speedup, and also
600 making signals reliable for the first time.
601 - used new SLF interface and other optimisations to speed up async_pool
602 by a factor of two. It also doesn't rely on perl's exception mechanism
603 to exit anymore. The overhead for terminating an async_pool, coro over
604 a normal async is now very small.
605 - sped up coroutine creation/destruction by 40%.
606 - forgot to include Coro/libcoro/README in the dist for all these years.
607 - work around a freebsd pthreads bug (manual testcancel is required as
608 pthread_cond_wait isn't a cancellation point on freebsd).
609 - use new rouse functions to speed up and simplify Coro::BDB.
610 - make "prefer perl native functions" work with threaded perls.
611 - condense Coro::Debug ps output, hint at v and w flags.
612 - (libcoro) lots of minor cleanups and portability improvements.
613
6144.914 Wed Nov 19 12:54:18 CET 2008
615 - fix a disastrous bug in the readline optimisation
616 introduced in 4.801.
617
6184.913 Sat Nov 15 07:58:28 CET 2008
619 - async_pool did free a scalar value twice
620 ("Attempt to unreference...").
621
6224.912 Thu Nov 13 18:31:23 CET 2008
623 - minor cleanups.
624 - use much larger stacks on linux and perl < 5.8.8.
625 - Coro::Debug::new_unix_server did not unlink the socket
626 when destroyed.
627
6284.911 Tue Nov 11 04:26:17 CET 2008
629 - "port" to threaded perls.
630
6314.91 Mon Nov 10 05:36:38 CET 2008
632 - the ->throw exception object no longer leaks.
633 - creating a new cctx leaked a scopestack entry (memleak).
634 - new coroutines didn't get created with a zero flags field
635 (unknown impact).
636 - calling ->throw on a not-yet-started coroutine should now work
637 instead of being ignored.
638 - ->throw is now supported on Coro::State objects.
639 - clean up cctx creation code a bit.
640 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
641
6424.9 Sat Nov 8 17:45:27 CET 2008
643 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
644 - (libcoro) no longer leak threads in the experimental pthread backend,
645 also speed it up considerably.
646 - (libcoro) do not rely on makecontext passing void *'s unscathed.
647 - fix compiletime dependencies on libcoro in the Makefile.
648 - cctx_count wasn't always updated properly.
649 - Coro::State::cctx_stacksize wasn't applied correctly.
650 - new function Coro::State::cctx_max_idle.
651 - the default max number of idle C contexts is now 4.
652 - (libcoro) try harder to get _setjmp/_longjmp.
653 - (libcoro) cleanup and extend the libcoro API to officially
654 allow the creation of empty source contexts.
655 - very experimental workaround for the totally broken netbsd platform.
656 - tried to hack around openbsd bugs.
657
6584.804 Wed Nov 5 16:36:00 CET 2008
659 - Coro::Debug::new_unix_server would not create a non-blocking listening
660 socket, sometimes causing freezes.
661 - (libcoro) fix misaligned stack points for setjmp and assembly
662 methods, which can cause crashes on x86/x86_64 with a sufficiently
663 aggressive compiler.
664 - new function: Coro::Debug::new_tcp_server.
665 - move ->throw into the Coro class because it only works on coro objects.
666
6674.803 Mon Nov 3 17:16:12 CET 2008
668 - (libcoro) use a global asm statement to become independent of gcc
669 otpimisations for CORO_ASM (thanks to pippijn for the idea).
670 - try to workaround yet another broken bsd, this time dragonfly.
671
6724.802 Thu Oct 30 10:56:12 CET 2008
673 - support -fno-omit-frame-pointer on x86 with the assembly method.
674 - tune 01_unblock.t tests a bit.
675
6764.801 Wed Oct 22 18:33:37 CEST 2008
677 - improve readline speed for very long "lines".
678 - backport to 5.8.8.
679
6804.8 Thu Oct 2 13:34:40 CEST 2008
681 - new function Coro::AIO::aio_wait.
682 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
683 - greatly speed up and reduce memory usage of Coro::AIO requests.
684 - implement some other µ-optimisations.
685
6864.749 Mon Sep 29 14:40:12 CEST 2008
687 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
688
6894.748 Sat Sep 27 14:03:19 CEST 2008
690 - implement, but do not document, PerlIO::cede(granularity).
691 - Coro::Storable forgot to wrap Storable::pstore.
692 - work around the multitude of leaks and memory corruption
693 bugs in PerlIO::via by using our own C-level perliol. As a side
694 effect, Coro::Storable is now much, much, much faster.
695
6964.747 Tue Sep 23 01:59:41 CEST 2008
697 - fix a per-cv memleak (one empty array was leaked per
698 code reference).
699 - avoid a crash in coro->call|rss when the coroutine was already
700 destroyed (most noticably when using Coro::Debug::ps :)
701 - also protect *Storable::FILE.
702 - push up default storable granularity to 20ms.
703
7044.746 Sun Sep 21 03:22:20 CEST 2008
705 - be more insistent on locking Storable against reentrancy
706 in Coro::Storable.
707 - move swap_def?v and throw to Coro::State, as documented.
708
7094.745 Thu Jul 24 00:14:38 CEST 2008
710 - remove debugging code related to MgPV_nolen_const, also try to
711 make it compile with perl 5.8.6 (yes, apple apparently loves
712 outdated software). Reported by John S.
713
7144.744 Tue Jul 8 22:06:35 CEST 2008
715 - correctly provide default DIE/WARN handlers as documented.
716 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
717 implementations inside perl work fine for us.
718
7194.743 Mon Jun 16 00:21:57 CEST 2008
720 - when using Coro::EV without running EV::loop it could
721 result in busy-waiting for events, this has been fixed.
722 - reduce codesize and improve performance by using EV_DEFAULT_UC.
723
7244.742 Sat May 31 14:10:21 CEST 2008
725 - implement a workaround for (some) perl <5.8.8 versions.
726 - require EV 3.3+.
727
7284.741 Fri May 30 23:33:09 CEST 2008
729 - tell netbsd how utterly broken their imitation of an OS is
730 and refuse to build by default if pthreads are in use.
731 - switch to "s" method on all bsd's by default, as their ucontext
732 stuff seems just too broken.
733 - fix a bug in Coro::Select.
734
7354.74 Thu May 29 20:05:31 CEST 2008
736 - do not test Coro::LWP for lack of dependencies.
737 - do not test Coro::Select for lack of working perls.
738
7394.73 Thu May 29 2008
740 - fix a bug in Coro::EV which would cause it to block despite
741 there being runnable coroutines.
742 - sprinkle "no warnings" freely over everything, also suppress
743 warnings for some other modules.
744 - fix typo in WSAEWOULDBLOCK.
745
7464.72 Sun May 25 05:14:36 CEST 2008
747 - tweak META.yaml a bit, unfortunately, there is no documented way
748 to have optional dependencies with CPAN. doh :(
749 - avoid running some tests on windows because they would fail due to
750 perl bug (broken fork, broken pipes...).
751 - work around perl on windows bugs where perl returns undocumented
752 error codes for sysread, syswrite etc. by taking advantage
753 of AnyEvent's workaround for that problem.
754 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
755 a common perl implementation bug on windows.
756 - use unix domain sockets in testsuite to work around
757 common perl implementation bugs on widows (they are emulated by
758 tcp sockets on windows. ugh.)
759
7604.71 Sat May 24 20:01:27 CEST 2008
761 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
762 - take advantage of async name resolution of AnyEvent::Util.
763 - work around brutal inet_aton override in Coro::LWP.
764 - take advantage of the readyhook in Coro::EV, for smoother
765 scheduling.
766
7674.7 Sun May 11 00:32:19 CEST 2008
768 - completely reworked the Coro manpage.
769 - added Coro::AnyEvent, generic event loop integration.
770 - implement cancel, ready and kill commands in Coro::Debug.
771 - document find_coro in Coro::Debug.
772 - incompatible change: rename has_stack to has_cctx.
773 - Coro::AIO and Coro::BDB no longer force event model detection,
774 use AnyEvent::AIO and AnyEvent::BDB.
775
7764.6 Sat Apr 26 10:05:14 CEST 2008
777 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
778 by the Coro module. It was a mistake to have it in the first place.
779 - (experimental) support for activestate perl 5.10 (method "w").
780 - (experimental) support for strawberry perl (method "a").
781 - coro_sigelem_set did not return a value although it had to,
782 actual impact unknown.
783
7844.51 Mon Apr 14 13:28:27 CEST 2008
785 - make it compile again on 5.8.
786
7874.50 Thu Apr 10 09:43:17 CEST 2008
788 - I did it twice! (see 4.49).
789
7904.49 Sun Apr 6 21:23:31 CEST 2008
791 - grr, instead of compiling the recent changes on 5.10 only they
792 were compiled on 5.8 only.
793
7944.48 Sun Apr 6 20:36:46 CEST 2008
795 - allow coroutine switches during eval's under 5.10.x, as it seems
796 the parser is a per-interpreter option now, so this is safe
797 (this might fix the odd crashing bug).
798 - drop support for 5.9.x versions: they are dead, jim.
799
8004.47 Sun Apr 6 00:37:52 CEST 2008
801 - force cctx allocation on API calls: we know we need to force one
802 and gcc actually manages to confuse our heuristic nowadays,
803 leading to crashes and worse.
804 - document force_cctx.
805
8064.46 Fri Apr 4 22:05:43 CEST 2008
807 - upgrade libcoro, resulting in pthread-backend (which was only created
808 to fulfill the rules of the programming languages shootout).
809
8104.45 Thu Mar 13 11:55:36 CET 2008
811 - fix a file leaking bug in eg/mhyttpd that would allow
812 downloading of any file (reported by oesi).
813 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
814 (also add testcase).
815 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
816 apparently having loads of fun with that).
817 - implement $state->swap_defsv, ->swap_defav and document ->throw.
818
8194.4 Wed Feb 13 16:44:29 CET 2008
820 - only assume .cfi pseudo isns exist on GNU/Linux.
821 - add get_timed method to Coro::Channels.
822 - fixed Coro::Debug manpage.
823 - perl 5.11 compatbility improvement based on patch by
824 Andreas König.
825
8264.37 Sun Jan 20 11:25:23 CET 2008
827 - soften the check for an active parser for perl 5.10 (avoiding
828 "Coro::State::transfer called while parsing" in many cases).
829
8304.36 Sun Jan 13 10:53:56 CET 2008
831 - reset diehook when terminating from an async_pool as to not
832 trigger any __DIE__ handlers.
833
8344.35 Sun Jan 13 04:14:13 CET 2008
835 - "bt" debug command now displays any exceptions
836 from longmess and also skips the innermost
837 stackframes, giving more useufl output.
838 - allow backtraces in coroutines blocked in Coro::EV,
839 at a <1% speed hit.
840 - handle localising of $SIG{__WARN__} and __DIE__
841 properly (with a proper amount of dirty hacking).
842
8434.34 Sat Dec 22 17:49:53 CET 2007
844 - upgrade to EV version 2.0 API.
845
8464.33 Mon Dec 17 08:36:12 CET 2007
847 - make Coro::AIO etc. loadable in the absence of EV.
848
8494.32 Mon Dec 17 07:46:02 CET 2007
850 - majorly improved Coro::Handle's performance with Coro::EV.
851 - implemented the until now mythical Coro::BDB module.
852 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
853
8544.31 Wed Dec 5 12:32:39 CET 2007
855 - remove warn statement form Coro::Util.
856
8574.3 Tue Dec 4 20:33:14 CET 2007
858 - calls to the idle function could cause stack corruption
859 when the stack changed.
860 - do no longer rely on the presence of EV::DNS if EV is used
861 (because it is gone), but instead take avdantage of EV::ADNS
862 if available.
863 - add ($) prototypes to all functions in Coro::Storable.
864 - use a conventional (safer) idle callback in Coro::EV.
865 - do accept quickly in Coro::Debug to avoid endless loops.
7 866
84.22 Fri Nov 30 16:04:04 CET 2007 8674.22 Fri Nov 30 16:04:04 CET 2007
9 - really use optimised versions for Event and EV in Coro::Util 868 - really use optimised versions for Event and EV in Coro::Util
10 and Coro::Handle. 869 and Coro::Handle.
11 870
… …
39 898
404.11 Thu Oct 11 02:40:24 CEST 2007 8994.11 Thu Oct 11 02:40:24 CEST 2007
41 - port to threaded perls. 900 - port to threaded perls.
42 901
434.1 Thu Oct 11 02:38:16 CEST 2007 9024.1 Thu Oct 11 02:38:16 CEST 2007
44 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 903 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
45 be local to each coro (see Coro::State). 904 be local to each coro (see Coro::State).
46 - incompatible change: for very deep reasons, cede and cede_notself 905 - incompatible change: for very deep reasons, cede and cede_notself
47 cannot return anything, so nothing will be returned. 906 cannot return anything, so nothing will be returned.
48 - possibly bring back 5.10 compatibility (untested). 907 - possibly bring back 5.10 compatibility (untested).
49 - work around stupid (and wrong) warning on 5.10 :(. 908 - work around stupid (and wrong) warning on 5.10 :(.
… …
121 but stateful semantics. 980 but stateful semantics.
122 - fixed a lot of typos in Coro.pm (patch submitted by David 981 - fixed a lot of typos in Coro.pm (patch submitted by David
123 Steinbrunner, which applied flawlessly). 982 Steinbrunner, which applied flawlessly).
124 983
1253.6 Sat Apr 14 17:13:31 CEST 2007 9843.6 Sat Apr 14 17:13:31 CEST 2007
126 - added some bugfixes to get eg/myhttpd working again. 985 - added some bugfixes to get eg/myhttpd working again.
127 - added Coro::Storable for often-cede'ing freeze/thaw. 986 - added Coro::Storable for often-cede'ing freeze/thaw.
128 - try to do a clean exit when a coroutine calls exit 987 - try to do a clean exit when a coroutine calls exit
129 (EXPERIMENTAL). 988 (EXPERIMENTAL).
130 - got rid of indirect call through _coro_init. 989 - got rid of indirect call through _coro_init.
131 - updated the partly antique examples in eg/ to 990 - updated the partly antique examples in eg/ to

Diff Legend

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