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

Comparing Coro/Changes (file contents):
Revision 1.301 by root, Thu Dec 13 05:15:27 2007 UTC vs.
Revision 1.689 by root, Sun Jul 26 10:30:56 2020 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?
7 10
11TODO: State.xs could short-circuit some macros such as CORO_EXECUTE_SLF
12TODO: use frame->destroy to improve Coro::EV
136.57
14 - override AnyEvent condvar methods "properly", as AnyEvent
15 has support for pluggable condvars since 2011. This also
16 avoids problems with recursive waits by avoiding to mix
17 impl methods with coro code.
18 - allow rouse_wait to be called in a different thread.
19 - slightly clarify rouse_cb/rouse_wait documentation.
20 - slightly improve rouse callback performance and memory usage.
21 - try to report correct calller when throwing exception in
22 on $condvar->recv, not a useless line in Coro::AnyEvent.
23
246.56 Thu Feb 20 19:40:07 CET 2020
25 - Coro::EV::time*once functions used ev_once, which is not cancelable.
26 This could cause problems due to I/O watchers not being stopped up
27 on exceptions/cancellation.
28
296.55 Fri Jul 19 17:28:09 CEST 2019
30 - Coro::RWLock woke up only one reader instead of all after a wrlock
31 (analyzed and patch by Maxim Skorobogatov).
32 - slightly improved documentation of Coro::RWLock and RWLock testcase.
33 - quickly (re)-ported to minix 3.3 before minix crashed again.
34
356.54 Sun Dec 16 11:06:07 CET 2018
36 - reinstate Coro::Select::select prototype - it breaks more than it fixes.
37
386.53 Sun Dec 16 10:33:15 CET 2018
39 - the autogenerated prototypes of Coro::AIO now add ";" at the end
40 to avoid generating named unary operators and messing up parsing
41 (mostly affects aio_mlockall 4 | 8 which was parsed as
42 (aio_mlockall 4) | 8).
43 - do not use a prototype for select in Coro::Select, as CORE::select also
44 doesn't have one.
45
466.52 Tue Aug 14 18:49:43 CEST 2018
47 - libcoro update: allow sharing of coro threads using the pthreads
48 backend among different pthreads, as is required by Coro::Multicore.
49 - always allow ->call and ->eval when the coro is the current coroutine. this
50 makes it possible, among other things, to call these on $Coro::main.
51 - minor code cleanups.
52
536.514 Thu Aug 31 18:25:31 CEST 2017
54 - safe_cancel no longer croaks when called on already-destroyed
55 thread, but instead "just works".
56
576.513 Sat Jul 15 01:18:51 CEST 2017
58 - experimental 5.22 comaptibility changes (PadlistNAMES).
59 - removed attributes example and made sure the other examples
60 still work.
61 - fix some statevar types for newer perls, should be of no consequence.
62
636.512 Fri Jul 14 05:03:17 CEST 2017
64 - seems to pass test suite down till 5.8.9, so make this the
65 minimum requirement in Makefile.PL.
66 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
67 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
68 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
69 fortified longjmp variants in some libcs.
70 - re-try JIT allocation without PROT_EXEC, to hopefully improve
71 portability (still doesn't seem to work around broken selinux
72 mmap everywhere).
73 - speling fixes (Xavier Guimard).
74
756.511 Sun Jun 26 23:44:50 CEST 2016
76 - make it compile with threaded perls again.
77 - simplify and speed up __DIE__ and __WARN__ handling.
78
796.51 Sat Jun 25 20:57:20 CEST 2016
80 - 6.5 release didn't compile on most older perls due to
81 a logic error enabling experimental 5.24 code for all versions.
82 - since various broken or deficient patches float around
83 to work around the vtbl API breakage in perl, let's provide
84 our own workaround, which seems to be much less invasive
85 and more compatible than the approaches seen so far. YMMV.
86 - work around assertion failure in perl_destruct on 5.24
87 (analyzed by Dave Mitchell).
88
896.5 Wed Jun 22 22:23:50 CEST 2016
90 - swap_sv swapping order was not symmetrical, causing
91 wrong swaps when swap_sv was used multiple times on the
92 same sv in the same thread.
93 - swap_sv calls can now be undone by calling it again
94 with the same variables.
95 - swap_sv calls will now be undone in async_pool threads.
96 - split Coro::Semaphore::up/adjust into separate xs functions
97 for better error reporting, at slight codesize increase.
98 - (libcoro) arm assembly support, please test and report.
99 - adjust to PL_savestack changes in perl 5.24 (adapted from
100 the debian patch, which unfortunately gets it wrong).
101
1026.49 Sat Oct 17 01:40:12 CEST 2015
103 - throwing an exception to a thread waiting in
104 Coro::Handle using Coro::EV did not stop the watchers,
105 causing the next call to fail (testcase by Martin Pritchard).
106 - bump minimum perl version to 5.10.
107
1086.48 Sun Oct 4 19:03:51 CEST 2015
109 - fix memory leak when Coro::Handle uses Coro::EV internally
110 (testcase by Sten Sten).
111 - update code to libev 4 API (internally, Coro still used the
112 version 3 API calls).
113 - remove unused hv_sig.
114
1156.47 Sat Jul 11 03:58:20 CEST 2015
116 - swap_svs - NVs can be in the SV head beginning with 5.20.
117
1186.46 Tue Jun 30 14:36:30 CEST 2015
119 - restore portability to perls with windows fork emulation
120 (patch by Petr Písař).
121
1226.45 Tue Jun 30 01:40:08 CEST 2015
123 - restore portability to perl 5.16 and below.
124
1256.44 Tue Jun 30 00:41:54 CEST 2015
126 - avoid segfaulting (or worse) when tracing a canceled thread.
127 - the Event module silently broke it's public hook API, causing
128 failures on perls compiled with -Duselongdouble. Requiring
129 the latest version of Event because I don't know which release
130 changed this (it's not mentioned in the ChangeLog).
131 - allow xs level enterleave hooks via CoroAPI.
132
1336.43 Thu Jun 4 15:38:14 CEST 2015
134 - use stability canary.
135 - port to stableperl-5.22.0-1.001.
136 - update libecb.
137
1386.42 Wed Feb 11 20:29:52 CET 2015
139 - Coro::SemaphoreSet->try did not actually work (analyzed by
140 SATO Kentaro).
141 - upgrade libecb to be C11 compliant.
142
1436.41 Sat Sep 6 22:08:46 CEST 2014
144 - restore portability to perl 5.8.x.
145 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
146 up some memory, but fixes "use feature" and similar modules
147 when used inside a Coro.
148 - allow zero as argument to Coro::Channel to mean the same thing
149 as no argument. this works with older versions as well,
150 but wasn't legal until now.
151 - slightly better c header file detection.
152
1536.39 Mon Jun 2 00:00:08 CEST 2014
154 - work around more incompatible changes in 5.20.
155
1566.38 Sun Jun 1 21:54:23 CEST 2014
157 - check that perl slots actually have enough space to hold
158 interpreter variables.
159 - untested port to perl 5.19 (64 bit tmps indices) (reported
160 by Andreas König).
161 - croak when cancel is called without a thread context.
162
1636.37 Tue Mar 4 13:27:33 CET 2014
164 - *sigh*, removed leftover debugging code from debugging a
165 perl bug, of all things.
166
1676.36 Tue Mar 4 07:11:59 CET 2014
168 - semaphores would not clear the destroy hook when interrupted
169 by ->throw, causing segfaults or worse.
170 - ->throw on a thread waiting for a semaphore did not acquire
171 the semaphore, but also didn't wake up other waiters,
172 possibly causing a deadlock.
173 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
174 to report a full stacktrace, which should help find out
175 where the actual call is.
176 - "atomically" destroy data in slf_destroy, because it is
177 the right thing to do, just in case.
178 - disable ecb.h libm dependency, to compile on ancient systems
179 or under adverse conditions.
180
1816.33 Mon Nov 18 11:26:27 CET 2013
182 - do not crash when freeing padlists with holes (in 5.18).
183 - tentative SVt_BIND 5.19 port/fix.
184
1856.32 Tue Nov 5 15:35:35 CET 2013
186 - use a new algorithm to derive padlists for perl 5.18. The old
187 one could lead to 0-pointer accesses inside perl (reported
188 by Darin McBride).
189
1906.31 Thu May 9 07:39:48 CEST 2013
191 - Coro::AIO requests would crash if the thread was ready'd
192 while the request was ongoing.
193
1946.29 Wed May 8 02:55:18 CEST 2013
195 - when an on_destroy handler destructs the coro currently being
196 destructed a perl scalar could be accessed after being freed,
197 likely causing a crash.
198
1996.28 Wed Mar 6 06:58:02 CET 2013
200 - clean remnants of existing __DIE__ and __WARN__ handlers so
201 they lose their magic and will not cause segfaults later
202 (testcase by Andrey Sagulin).
203 - improved Coro::State documentation a bit.
204 - Coro::Debug::command now flushes the output.
205 - add hack detection code for x32 abi, because the braindead slugs
206 who designed that made it look exactly like x86_64 without
207 providing proper compile time symbols to test for it. as a result,
208 this detection cannot work reliably.
209 - valgrind stack registering was broken.
210 - do not rely on Time::HiRes anymore in Coro::Debug.
211
2126.23 Fri Dec 7 23:36:37 CET 2012
213 - use experimental fiber implementation on native windows
214 perls.
215 - use sizeof (void *) as multiplication factor for stack sizes,
216 to accomodate the totally braindamaged microsoft 64 bit "os".
217 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
218 activeperl 5.16 crashes when PerlIO_define_layer is called due
219 to some bug in the perl dll, strawberry perl at least passes
220 the testsuite.
221 - implement Coro::Handle->peeraddr/host/port, for slightly
222 improved compatibility with LWP.
223 - implement 5.17 compatibility by almost blindly applying a
224 good-looking patch by Father Chrysostomos.
225 - move stack management functions into libcoro 3.
226 - libcoro version 3 "released".
227 - support magic values as timed_io_once args.
228 - recommend AnyEvent 7+ or EV 4+, also require EV
229 version 4 or newer for Coro::EV.
230
2316.10 Tue Oct 9 01:14:27 CEST 2012
232 - updated ecb.h, it had a typo that caused it to not compile on many
233 big endian systems (reported by many people).
234 - disable memory fences in ecb.h to improve portability.
235
2366.09 Sat Oct 6 23:25:02 CEST 2012
237 - Coro::EV I/O watchers were not interruptible by exceptions
238 (Coro::State::throw) (testcase by sten).
239 - ->throw now puts threads into the ready queue, as this seems to
240 be expected by existing code, and code that doesn't cope with spurious
241 wakeups needs fixing anyway.
242 - use fd -1 in mmap.
243 - cast I32 to int in error message printf.
244 - warn about broken so-called "hardened" kernels.
245
2466.08 Fri Apr 13 12:05:47 CEST 2012
247 - be more aggressive about exiting like perl does - formerly,
248 exiting from the non-main thread would not execute END blocks.
249
2506.07 Fri Nov 11 21:21:48 CET 2011
251 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
252 - update ecb.h.
253
2546.06 Mon Aug 8 23:59:48 CEST 2011
255 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
256 backend, which is an order of magnitude slower. unfortunately, cygwins
257 pthread implementation isn't very complete either, so allocate the stack
258 twice just to be sure.
259 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
260 in its ithreaded perl - best recompile cygwin and use the 'w'indows
261 backend for much better performance. also disable ithreads for
262 even better performance...).
263
2646.05 Thu Aug 4 21:36:36 CEST 2011
265 - blush, condvar values would not be propagated from send to recv anymore
266 (reported by Chip Salzenberg).
267 - use exponential increase for the readline buffer length in
268 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
269
2706.04 Wed Aug 3 17:15:45 CEST 2011
271 - use even more efficient and more compatible condvars for
272 compatibility to AnyEvent 6.x :)
273 - more inconsequential ecb.h updates.
274
2756.03 Wed Aug 3 11:41:30 CEST 2011
276 - change how Coro patches AnyEvent condvars for compatibility to
277 AnyEvent 6.x.
278 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
279 hasn't been included.
280
2816.02 Wed Jul 13 04:35:19 CEST 2011
282 - "improve portability to Gentoo" - gentoo manages to put perl variables
283 in memory areas that are farther than 2gb apart, which the jit couldn't
284 handle and barfed. now it's just a bit slower on gentoo and similar
285 systems.
286
2876.01 Sun Jul 3 12:31:14 CEST 2011
288 - workarounds are good, but the test for whether pthreads are used
289 was not good. this one should be better.
290 - check differently whether gcc generates cfi instructions itself.
291
2926.0 Wed Jun 29 19:43:35 CEST 2011
293 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
294 destroyed and cleaned up automatically (e.g. async { schedule }).
295 - implement a JIT compiler for part of the thread switch code,
296 which gives a 50% speed improvement on threaded perls, and
297 about 4% on non-threaded perls (so threaded perls now finally
298 reach about half the speed of non-threaded perls).
299 - slightly modernise Coro::Intro, add section about rouse functions.
300 - avoid DEFSV and ERRSV, giving another 10% improvement
301 in thread switching.
302 - Coro::State->is_destroyed is now called is_zombie.
303 - implement a Coro->safe_cancel method that might fail, but
304 cancels in a "safer" way if it succeeds.
305 - add preliminary support for DEBUGGING perls.
306 - get rid of two hash-accesses when initialising a new Coro - this
307 speeds up coro creation by almost a factor of two.
308 - croak when a coro that is being cancelled tries to block
309 (e.g. while executing a guard block), instead of crashing or
310 deadlocking.
311 - use a more robust and also faster method to identify Coro::State
312 objects - speeds up everything a bit.
313 - implement Coro->cancel in XS for a 20% speed improvement, and to
314 be able to implement mutual cancellation.
315 - speed up context switches by a percent or two by more efficiently
316 allocating context stack entries.
317 - implement Coro->join and Coro->on_destroy in XS for a speedup and
318 a reduction in memory use.
319 - cancelling a coro while it itself is cancelling another coro is
320 now supported and working, instead of triggering an assertion.
321 - be a bit more crash-resistant when calling (buggy) on_destroy
322 callbacks (best effort).
323 - move on_destroy into the slf_frame, to allow extension slf
324 functions to have destructors.
325 - get rid if coro refcounting - simply crash in other interpreter
326 threads by nulling the pointers on clone.
327 - simplify warn/die hook handling when loading Coro - the convoluted
328 logic seems to be no longer neccessary.
329 - use libecb instead of our own home-grown gcc hacks.
330 - document alternatives to Coro::LWP. Please use them :)
331 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
332 sigaltstack. Really. wine suffers from it, erlang suffers from it,
333 and it's known since at least 2006.
334
3355.372 Wed Feb 23 06:14:30 CET 2011
336 - apparently mingw doesn't provide a working gettimeofday, try to
337 work around that by relying on Time::HiRes (indirectly brought to
338 my attention by Max Maischein).
339 - fix some portability issues when Time::HiRes was used.
340
3415.371 Mon Feb 21 14:36:08 CET 2011
342 - backport to windows process emulation code again.
343
3445.37 Sat Feb 19 07:49:44 CET 2011
345 - add a big "Coro thread life cycle" section to "man Coro".
346 - try a tentative workaround against the breakage that 5.13 has
347 introduced without depreciation period. sigh.
348 - no longer use Time::HiRes if gettimeofday is available, which
349 saves quite a lot of memory.
350
3515.36 Sun Feb 13 05:33:41 CET 2011
352 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
353 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
354 method call.
355 - undocument Coro::Timer::sleep and obsolete whole module.
356 - optimise Coro::Timer::timeout memory and cpu usage.
357 - slightly updated Coro::Intro for recent changes.
358 - do not initialise PL_dirty anymore.
359
3605.25 Thu Nov 11 01:08:39 CET 2010
361 - try a different approach on netbsd - netbsd 5 finally has marginally
362 working pthreads, but still broken ucontext/sigaltstack.
363 - openbsd 4.8 finally got their act together, Coro works out of the box
364 with asm, setjmp and pthreads (no change, just informational).
365
3665.24 Sat Oct 23 11:27:12 CEST 2010
367 - port to the EV 4.0 API.
368 - work around bugs in mingw32, making strawberry perl work
369 out of the box.
370 - correctly modify Coro::AIO function prototypes
371 so that they reflect the "no optional parameters" rule.
372 - "ported" libcoro to C++.
373
3745.23 Mon May 17 18:50:42 CEST 2010
375 - be more resistant to ordering changes when initialising
376 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
377 Waldorf).
378 - document that perl 5.12 deliberately removed support for cloning.
379
3805.22 Wed Apr 14 03:55:35 CEST 2010
381 - correctly return udnef on errors in Coro::Handle::read/write
382 (testcase by Marc Mims).
383 - convert Coro::Util into a "perl compatibility wrapper" - the functions
384 are less useful now, but are drop-in replacements for existing
385 functions, listing better alternatives in the documentation. This also
386 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
387 with Coro::Util::inet_aton.
388 - do not override $Coro::idle unconditionally in Coro.pm, as other
389 modules could have provided their own idle coro already
390 (for exmaple, Coro::AnyEvent).
391 - fix Coro::Util::gethost* functions.
392 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
393
3945.21 Wed Dec 16 07:19:51 CET 2009
395 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
396 together.
397 - add some examples on how to combine other event loops with Coro in
398 Coro::AnyEvent, and how to run it (and not to block). Seems to be
399 the most common source of confusion.
400 - try to catch people naively blocking in an event callback.
401 - work around the perl filehandle bug issue in conjunction with
402 older common::sense (as indirectly pointed out by ZSystem).
403 - clarify the "not from signal handlers" section.
404
4055.2 Sun Oct 4 14:54:24 CEST 2009
406 - Coro::Storable destroyed the prototypes of the functions it wrapped.
407 - export rouse_cb and rouse_wait by default now.
408 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
409 - new method $state->swap_sv.
410 - added section on "windows process emulation" to the manpage,
411 after a not-so-fruitful (nor-friendly) "discussion" with chip
412 salzenberg (discussion implies arguments, but his only arguments
413 were ad-hominems, one wonders why he started it in the first
414 place). I hope this explains it well enough for him to understand,
415 and maybe well enough for others to understand.
416 - use common::sense everywhere now.
417 - idle callbacks are no longer supported, use idle coros instead.
418 - print a thread listing when a deadlock is detected.
419
4205.17 Sat Aug 22 23:09:31 CEST 2009
421 - work around a bug in the perl debugger causing crashes
422 when running under the debugger by marking _pool_handler as nodebug.
423 - speed up Coro::async considerably.
424 - try some hacks to get netbsd to work "more often" - their broken
425 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
426 - convert Coro to AE 5.0 API.
427
4285.162 Tue Jul 28 04:04:03 CEST 2009
429 - perl 5.8.2 is now minimum requirement.
430 - skip t/19_handle.t on broken windows perls.
431
4325.161 Wed Jul 22 04:47:38 CEST 2009
433 - Coro::AnyEvent::poll could have a different prototype when EV was
434 used as backend (analyzed by Tatsuhiko Miyagawa).
435 - Coro::AnyEvent errornously initialised the event loop when loaded,
436 not on demand.
437 - try to workaround rare */t/01_unblock.t failures.
438
4395.16 Tue Jul 21 01:44:37 CEST 2009
440 - Coro::AnyEvent failed to hook into the event loop
441 when no threads had been readied between detecting
442 the event loop and actually running it.
443 - considerably speed up Coro::Select by taking avdantage
444 of AnyEvent > 4.8 and some other optimisations.
445 - implement paragraph readline mode in Coro::Handle
446 (based on patches by Zsbán Ambrus).
447 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
448 and analyzed by Yasuhiro MATSUMOTO).
449 - clarified libcoro license and copyright.
450 - someone stole my EXTRA_META!!!
451 - implement Coro::Select::patch_pp_sselect and it's brother,
452 for hardcode select overriding.
453
4545.151 Mon Jul 6 05:41:57 CEST 2009
455 - backport to windows process emulation code again (patch by
456 Yasuhiro MATSUMOTO).
457 - slightly update Coro::MakeMaker.
458
4595.15 Tue Jun 30 10:28:06 CEST 2009
460 - deprecate Coro::Socket, document how to get ipv6 support via
461 AnyEvent::Socket instead.
462 - implement signal->wait ($cb) interface, similar to semaphores.
463 - work around SvOK not supporting getmagic, so we have to getmagic
464 to test for undef :/ (reported by Matthias Waldorf).
465 - load Coro::AnyEvent in all modules using AnyEvent.
466 - work around perl corrupting our internal data structures,
467 reported by Tokuhiro Matsuno.
468 - enable per-coroutine real and cpu time gathering
469 (Coro::State::enable_times).
470
4715.14 Wed Jun 24 01:37:48 CEST 2009
472 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
473 as this seems to have been a difficult concept (poll, sleep, idle,
474 idle_upto).
475 - add Coro::AnyEvent::readable/writable functions.
476 - clarify Coro::EV/Event/AnyEvent manpages.
477 - free per-thread global scalars in the thread calling ->cancel, to
478 avoid crashes when $_, $@ etc., are magical but some of those
479 had already been freed.
480 - "unexperimentalise" the callback interface for Coro::Semaphore.
481 - speed up ready queue management/context switching by using a linked
482 list instead of an array (~5-10%).
483 - implement "watch" command in Coro::Debug shells.
484 - for fun, implement time-slicing as an example in the manpage.
485 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
486 Coro::Event, use the normal AnyEvent handling instead of dieing
487 (the same is true for Coro::Handle).
488 - properly document Coro::EV::timed_io_once.
489 - avoid unneccessary ->cancel calls in Coro::Handle.
490 - maybe make it work on mingw32 with win32 backend
491 (based on patch by Yasuhiro Matsumoto).
492
4935.132 Fri May 29 09:00:39 CEST 2009
494 - do not keep a reference to the argument itself in
495 Coro::Semaphore::guard, as it could change later.
496 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
497 should support a prepare callback.
498
4995.131 Mon Mar 16 23:20:37 CET 2009
500 - implement and document Coro->suspend, Coro->resume.
501 - fix Coro::Select implementation to not (often) close
502 the passed file descriptors (testcase provided by pippijn).
503
5045.13 Mon Dec 15 21:51:42 CET 2008
505 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
506 - don't set diehook to C<undef> but instead to NULL, to avoid
507 spurious warnings.
508 - fix a lot of bugs in Coro::SemaphoreSet.
509 - Coro::SemaphoreSet will less often create a semaphore needlessly.
510 - add Coro::SemaphoreSet::count and wait methods.
511 - take advantage of the new Guard module.
512 - deprecate Coro::guard.
513 - try to fix the dreaded 01_unblock tests once more. I hate it when
514 testsuites need more fixing than the code they are supposed to test.
515 - croak in more cases when a required callback isn't resolvable.
516 - fix some minor issues in Coro::State->call/eval.
517 - use current coroutine context instead of a temporary one
518 when temporarily switching to another coroutine.
519 - do not call C-level on_destroy handlers during global destruction,
520 to avoid needless segfaults.
521
5225.12 Sun Dec 7 13:30:38 CET 2008
523 - add default config for MirOS, which seems to be a bug-to-bug
524 compatible fork of openbsd ("world domination by releasing
525 openbsd cvs before the openbsd folks do it" or so :).
526 - free_padlist did destroy the names pad, not good, but didn't
527 seem to bother perl - this could fix issues such as eval ""
528 inside a function called from multiple coroutines.
529 - use a different method to detect destruction time.
530 - be more careful when freeing padlists just before global
531 destruction.
532 - fixed and expanded the call/cc example.
533 - renamed _terminate to _coro_run.
534 - new method Coro::Channel->shutdown.
535 - try pthreads on openbsd <4.4 (broken sigaltstack, will
536 pthreads fare better?).
537 - be less picky about destroying "a" running coroutine.
538
5395.11 Tue Nov 25 21:49:05 CET 2008
540 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
541 working code with failed assertions, introducing more bugs than
542 it fixes, requiring elaborate workarounds :(
543
5445.1 Mon Nov 24 08:54:59 CET 2008
545 - wrote a small introductory tutorial - Coro::Intro.
546 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
547 - Coro::Select did errornously dup the file descriptors
548 and didn't work with all AnyEvent backends.
549 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
550 LWP data transfers.
551 - disassociate c contexts from coro objects - this is agruably more
552 correct, but mostly allows sharing of cctxs between coro and state
553 objects, for added memory savings and speed increases.
554 - bumped $Coro::POOL_RSS up to 32kb by default.
555 - no longer set the optype to OP_CUSTOM, as B::* understandably
556 doesn't like this very much (and we *are* a type of entersub).
557 - implement state cloning, just to prove that call/cc can be done.
558 - automatically load Coro::AnyEvent in Coro::Handle.
559 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
560 do not have this method (and don't need it either).
561 - speed up generic anyevent methods in Coro::Handle by using rouse
562 callbacks.
563 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
564 others. It also makes the debugger happier, as you can trace
565 through the idle threads now.
566 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
567 - eg/event was pretty broken.
568 - better 5.8.6 compatibility.
569
5705.0 Thu Nov 20 10:35:05 CET 2008
571 - NEW ARCHITECTURE: use the latest 4.x version if you experience
572 stability issues.
573 - bump API version to 7 - all dependents must be recompiled.
574 - removed timed_* functions - they were not being used anyways
575 and should be replaced by a more generic mechanism -
576 and were annoying to support anyways :)
577 - removed SemaphoreSet's waiter method - use sem method instead.
578 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
579 - async_pool did free a scalar value twice
580 ("Attempt to unreference...").
581 - fix a longstanding bug where calling terminate on a coro that
582 was waiting for a semaphore that was just becoming available
583 would cause a deadlock (semaphore would get into a state where
584 it was available but waiters were still blocked).
585 - calling throw on a coroutine that is waiting for a semaphore will
586 no longer make it acquire the semaphore (and thus leak a count).
587 - perl's process emulation is now not even theoretically supported
588 anymore.
589 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
590 conversion of callback-style to blocking-style.
591 - new methods $coro->schedule_to and ->cede_to, to specifically
592 schedule or cede to a specific coroutine.
593 - new function Coro::Semaphore::wait.
594 - use named constants in Coro::Channel (Richard Hundt).
595 - directly patch the entersub opcode calling SLF functions (cede,
596 transfer and so on). this does speed up context switching, but
597 more importanly, it frees us from the hardcoded behaviour of
598 entersub, so we might actually be able to return something from
599 those functions and atcually create new ones.
600 - take advantage of __builtin_frame_address on gcc.
601 - expose THX in coroapi (not sure whether this was a wise decision,
602 as "threaded" perls are running at half speed anyways).
603 - implement execute_slf (schedule-like-function) interface that makes
604 it possible to implement schedule-like-functions in XS.
605 - use new SLF interface to massively speed up Coro::EV by roughly a
606 factor of two.
607 - used new SLF interface to massively speed up Coro::Semaphore by a
608 factor of three.
609 - used new SLF interface to speed up Coro::AIO by roughly a factor of
610 four and reduce its memory usage considerably.
611 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
612 for a nice speedup and vastly more correct behaviour. Also implement
613 a new method "sem" to get at the underlying semaphore object.
614 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
615 (in comparison) 20-40% speedup.
616 - used new SLF interface to reimplement Coro::Signal gaining
617 some unknown (because I was too lazy), but certain, speedup, and also
618 making signals reliable for the first time.
619 - used new SLF interface and other optimisations to speed up async_pool
620 by a factor of two. It also doesn't rely on perl's exception mechanism
621 to exit anymore. The overhead for terminating an async_pool, coro over
622 a normal async is now very small.
623 - sped up coroutine creation/destruction by 40%.
624 - forgot to include Coro/libcoro/README in the dist for all these years.
625 - work around a freebsd pthreads bug (manual testcancel is required as
626 pthread_cond_wait isn't a cancellation point on freebsd).
627 - use new rouse functions to speed up and simplify Coro::BDB.
628 - make "prefer perl native functions" work with threaded perls.
629 - condense Coro::Debug ps output, hint at v and w flags.
630 - (libcoro) lots of minor cleanups and portability improvements.
631
6324.914 Wed Nov 19 12:54:18 CET 2008
633 - fix a disastrous bug in the readline optimisation
634 introduced in 4.801.
635
6364.913 Sat Nov 15 07:58:28 CET 2008
637 - async_pool did free a scalar value twice
638 ("Attempt to unreference...").
639
6404.912 Thu Nov 13 18:31:23 CET 2008
641 - minor cleanups.
642 - use much larger stacks on linux and perl < 5.8.8.
643 - Coro::Debug::new_unix_server did not unlink the socket
644 when destroyed.
645
6464.911 Tue Nov 11 04:26:17 CET 2008
647 - "port" to threaded perls.
648
6494.91 Mon Nov 10 05:36:38 CET 2008
650 - the ->throw exception object no longer leaks.
651 - creating a new cctx leaked a scopestack entry (memleak).
652 - new coroutines didn't get created with a zero flags field
653 (unknown impact).
654 - calling ->throw on a not-yet-started coroutine should now work
655 instead of being ignored.
656 - ->throw is now supported on Coro::State objects.
657 - clean up cctx creation code a bit.
658 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
659
6604.9 Sat Nov 8 17:45:27 CET 2008
661 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
662 - (libcoro) no longer leak threads in the experimental pthread backend,
663 also speed it up considerably.
664 - (libcoro) do not rely on makecontext passing void *'s unscathed.
665 - fix compiletime dependencies on libcoro in the Makefile.
666 - cctx_count wasn't always updated properly.
667 - Coro::State::cctx_stacksize wasn't applied correctly.
668 - new function Coro::State::cctx_max_idle.
669 - the default max number of idle C contexts is now 4.
670 - (libcoro) try harder to get _setjmp/_longjmp.
671 - (libcoro) cleanup and extend the libcoro API to officially
672 allow the creation of empty source contexts.
673 - very experimental workaround for the totally broken netbsd platform.
674 - tried to hack around openbsd bugs.
675
6764.804 Wed Nov 5 16:36:00 CET 2008
677 - Coro::Debug::new_unix_server would not create a non-blocking listening
678 socket, sometimes causing freezes.
679 - (libcoro) fix misaligned stack points for setjmp and assembly
680 methods, which can cause crashes on x86/x86_64 with a sufficiently
681 aggressive compiler.
682 - new function: Coro::Debug::new_tcp_server.
683 - move ->throw into the Coro class because it only works on coro objects.
684
6854.803 Mon Nov 3 17:16:12 CET 2008
686 - (libcoro) use a global asm statement to become independent of gcc
687 otpimisations for CORO_ASM (thanks to pippijn for the idea).
688 - try to workaround yet another broken bsd, this time dragonfly.
689
6904.802 Thu Oct 30 10:56:12 CET 2008
691 - support -fno-omit-frame-pointer on x86 with the assembly method.
692 - tune 01_unblock.t tests a bit.
693
6944.801 Wed Oct 22 18:33:37 CEST 2008
695 - improve readline speed for very long "lines".
696 - backport to 5.8.8.
697
6984.8 Thu Oct 2 13:34:40 CEST 2008
699 - new function Coro::AIO::aio_wait.
700 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
701 - greatly speed up and reduce memory usage of Coro::AIO requests.
702 - implement some other µ-optimisations.
703
7044.749 Mon Sep 29 14:40:12 CEST 2008
705 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
706
7074.748 Sat Sep 27 14:03:19 CEST 2008
708 - implement, but do not document, PerlIO::cede(granularity).
709 - Coro::Storable forgot to wrap Storable::pstore.
710 - work around the multitude of leaks and memory corruption
711 bugs in PerlIO::via by using our own C-level perliol. As a side
712 effect, Coro::Storable is now much, much, much faster.
713
7144.747 Tue Sep 23 01:59:41 CEST 2008
715 - fix a per-cv memleak (one empty array was leaked per
716 code reference).
717 - avoid a crash in coro->call|rss when the coroutine was already
718 destroyed (most noticably when using Coro::Debug::ps :)
719 - also protect *Storable::FILE.
720 - push up default storable granularity to 20ms.
721
7224.746 Sun Sep 21 03:22:20 CEST 2008
723 - be more insistent on locking Storable against reentrancy
724 in Coro::Storable.
725 - move swap_def?v and throw to Coro::State, as documented.
726
7274.745 Thu Jul 24 00:14:38 CEST 2008
728 - remove debugging code related to MgPV_nolen_const, also try to
729 make it compile with perl 5.8.6 (yes, apple apparently loves
730 outdated software). Reported by John S.
731
7324.744 Tue Jul 8 22:06:35 CEST 2008
733 - correctly provide default DIE/WARN handlers as documented.
734 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
735 implementations inside perl work fine for us.
736
7374.743 Mon Jun 16 00:21:57 CEST 2008
738 - when using Coro::EV without running EV::loop it could
739 result in busy-waiting for events, this has been fixed.
740 - reduce codesize and improve performance by using EV_DEFAULT_UC.
741
7424.742 Sat May 31 14:10:21 CEST 2008
743 - implement a workaround for (some) perl <5.8.8 versions.
744 - require EV 3.3+.
745
7464.741 Fri May 30 23:33:09 CEST 2008
747 - tell netbsd how utterly broken their imitation of an OS is
748 and refuse to build by default if pthreads are in use.
749 - switch to "s" method on all bsd's by default, as their ucontext
750 stuff seems just too broken.
751 - fix a bug in Coro::Select.
752
7534.74 Thu May 29 20:05:31 CEST 2008
754 - do not test Coro::LWP for lack of dependencies.
755 - do not test Coro::Select for lack of working perls.
756
7574.73 Thu May 29 2008
758 - fix a bug in Coro::EV which would cause it to block despite
759 there being runnable coroutines.
760 - sprinkle "no warnings" freely over everything, also suppress
761 warnings for some other modules.
762 - fix typo in WSAEWOULDBLOCK.
763
7644.72 Sun May 25 05:14:36 CEST 2008
765 - tweak META.yaml a bit, unfortunately, there is no documented way
766 to have optional dependencies with CPAN. doh :(
767 - avoid running some tests on windows because they would fail due to
768 perl bug (broken fork, broken pipes...).
769 - work around perl on windows bugs where perl returns undocumented
770 error codes for sysread, syswrite etc. by taking advantage
771 of AnyEvent's workaround for that problem.
772 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
773 a common perl implementation bug on windows.
774 - use unix domain sockets in testsuite to work around
775 common perl implementation bugs on widows (they are emulated by
776 tcp sockets on windows. ugh.)
777
7784.71 Sat May 24 20:01:27 CEST 2008
779 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
780 - take advantage of async name resolution of AnyEvent::Util.
781 - work around brutal inet_aton override in Coro::LWP.
782 - take advantage of the readyhook in Coro::EV, for smoother
783 scheduling.
784
7854.7 Sun May 11 00:32:19 CEST 2008
786 - completely reworked the Coro manpage.
787 - added Coro::AnyEvent, generic event loop integration.
788 - implement cancel, ready and kill commands in Coro::Debug.
789 - document find_coro in Coro::Debug.
790 - incompatible change: rename has_stack to has_cctx.
791 - Coro::AIO and Coro::BDB no longer force event model detection,
792 use AnyEvent::AIO and AnyEvent::BDB.
793
7944.6 Sat Apr 26 10:05:14 CEST 2008
795 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
796 by the Coro module. It was a mistake to have it in the first place.
797 - (experimental) support for activestate perl 5.10 (method "w").
798 - (experimental) support for strawberry perl (method "a").
799 - coro_sigelem_set did not return a value although it had to,
800 actual impact unknown.
801
8024.51 Mon Apr 14 13:28:27 CEST 2008
803 - make it compile again on 5.8.
804
8054.50 Thu Apr 10 09:43:17 CEST 2008
806 - I did it twice! (see 4.49).
807
8084.49 Sun Apr 6 21:23:31 CEST 2008
809 - grr, instead of compiling the recent changes on 5.10 only they
810 were compiled on 5.8 only.
811
8124.48 Sun Apr 6 20:36:46 CEST 2008
813 - allow coroutine switches during eval's under 5.10.x, as it seems
814 the parser is a per-interpreter option now, so this is safe
815 (this might fix the odd crashing bug).
816 - drop support for 5.9.x versions: they are dead, jim.
817
8184.47 Sun Apr 6 00:37:52 CEST 2008
819 - force cctx allocation on API calls: we know we need to force one
820 and gcc actually manages to confuse our heuristic nowadays,
821 leading to crashes and worse.
822 - document force_cctx.
823
8244.46 Fri Apr 4 22:05:43 CEST 2008
825 - upgrade libcoro, resulting in pthread-backend (which was only created
826 to fulfill the rules of the programming languages shootout).
827
8284.45 Thu Mar 13 11:55:36 CET 2008
829 - fix a file leaking bug in eg/mhyttpd that would allow
830 downloading of any file (reported by oesi).
831 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
832 (also add testcase).
833 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
834 apparently having loads of fun with that).
835 - implement $state->swap_defsv, ->swap_defav and document ->throw.
836
8374.4 Wed Feb 13 16:44:29 CET 2008
838 - only assume .cfi pseudo isns exist on GNU/Linux.
839 - add get_timed method to Coro::Channels.
840 - fixed Coro::Debug manpage.
841 - perl 5.11 compatbility improvement based on patch by
842 Andreas König.
843
8444.37 Sun Jan 20 11:25:23 CET 2008
845 - soften the check for an active parser for perl 5.10 (avoiding
846 "Coro::State::transfer called while parsing" in many cases).
847
8484.36 Sun Jan 13 10:53:56 CET 2008
849 - reset diehook when terminating from an async_pool as to not
850 trigger any __DIE__ handlers.
851
8524.35 Sun Jan 13 04:14:13 CET 2008
853 - "bt" debug command now displays any exceptions
854 from longmess and also skips the innermost
855 stackframes, giving more useufl output.
856 - allow backtraces in coroutines blocked in Coro::EV,
857 at a <1% speed hit.
858 - handle localising of $SIG{__WARN__} and __DIE__
859 properly (with a proper amount of dirty hacking).
860
8614.34 Sat Dec 22 17:49:53 CET 2007
862 - upgrade to EV version 2.0 API.
863
8644.33 Mon Dec 17 08:36:12 CET 2007
865 - make Coro::AIO etc. loadable in the absence of EV.
866
8674.32 Mon Dec 17 07:46:02 CET 2007
8 - majorly improved Coro::Handle's performance with Coro::EV. 868 - majorly improved Coro::Handle's performance with Coro::EV.
869 - implemented the until now mythical Coro::BDB module.
870 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
9 871
104.31 Wed Dec 5 12:32:39 CET 2007 8724.31 Wed Dec 5 12:32:39 CET 2007
11 - remove warn statement form Coro::Util. 873 - remove warn statement form Coro::Util.
12 874
134.3 Tue Dec 4 20:33:14 CET 2007 8754.3 Tue Dec 4 20:33:14 CET 2007
… …
54 916
554.11 Thu Oct 11 02:40:24 CEST 2007 9174.11 Thu Oct 11 02:40:24 CEST 2007
56 - port to threaded perls. 918 - port to threaded perls.
57 919
584.1 Thu Oct 11 02:38:16 CEST 2007 9204.1 Thu Oct 11 02:38:16 CEST 2007
59 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 921 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
60 be local to each coro (see Coro::State). 922 be local to each coro (see Coro::State).
61 - incompatible change: for very deep reasons, cede and cede_notself 923 - incompatible change: for very deep reasons, cede and cede_notself
62 cannot return anything, so nothing will be returned. 924 cannot return anything, so nothing will be returned.
63 - possibly bring back 5.10 compatibility (untested). 925 - possibly bring back 5.10 compatibility (untested).
64 - work around stupid (and wrong) warning on 5.10 :(. 926 - work around stupid (and wrong) warning on 5.10 :(.
… …
136 but stateful semantics. 998 but stateful semantics.
137 - fixed a lot of typos in Coro.pm (patch submitted by David 999 - fixed a lot of typos in Coro.pm (patch submitted by David
138 Steinbrunner, which applied flawlessly). 1000 Steinbrunner, which applied flawlessly).
139 1001
1403.6 Sat Apr 14 17:13:31 CEST 2007 10023.6 Sat Apr 14 17:13:31 CEST 2007
141 - added some bugfixes to get eg/myhttpd working again. 1003 - added some bugfixes to get eg/myhttpd working again.
142 - added Coro::Storable for often-cede'ing freeze/thaw. 1004 - added Coro::Storable for often-cede'ing freeze/thaw.
143 - try to do a clean exit when a coroutine calls exit 1005 - try to do a clean exit when a coroutine calls exit
144 (EXPERIMENTAL). 1006 (EXPERIMENTAL).
145 - got rid of indirect call through _coro_init. 1007 - got rid of indirect call through _coro_init.
146 - updated the partly antique examples in eg/ to 1008 - updated the partly antique examples in eg/ to

Diff Legend

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