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

Comparing Coro/Changes (file contents):
Revision 1.354 by root, Sun May 25 05:00:03 2008 UTC vs.
Revision 1.647 by root, Sat Jul 11 01:59:05 2015 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, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
8TODO: croak when async_pool tries to run canceled thread?
7 9
106.47 Sat Jul 11 03:58:20 CEST 2015
11 - swap_svs - NVs can be in the SV head beginning with 5.20.
12
136.46 Tue Jun 30 14:36:30 CEST 2015
14 - restore portability to perls with windows fork emulation
15 (patch by Petr Písař).
16
176.45 Tue Jun 30 01:40:08 CEST 2015
18 - restore portability to perl 5.16 and below.
19
206.44 Tue Jun 30 00:41:54 CEST 2015
21 - avoid segfaulting (or worse) when tracing a canceled thread.
22 - the Event module silently broke it's public hook API, causing
23 failures on perls compiled with -Duselongdouble. Requiring
24 the latest version of Event because I don't know which release
25 changed this (it's not mentioned in the ChangeLog).
26 - allow xs level enterleave hooks via CoroAPI.
27
286.43 Thu Jun 4 15:38:14 CEST 2015
29 - use stability canary.
30 - port to stableperl-5.22.0-1.001.
31 - update libecb.
32
336.42 Wed Feb 11 20:29:52 CET 2015
34 - Coro::SemaphoreSet->try did not actually work (analyzed by
35 SATO Kentaro).
36 - upgrade libecb to be C11 compliant.
37
386.41 Sat Sep 6 22:08:46 CEST 2014
39 - restore portability to perl 5.8.x.
40 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
41 up some memory, but fixes "use feature" and similar modules
42 when used inside a Coro.
43 - allow zero as argument to Coro::Channel to mean the same thing
44 as no argument. this works with older versions as well,
45 but wasn't legal until now.
46 - slightly better c header file detection.
47
486.39 Mon Jun 2 00:00:08 CEST 2014
49 - work around more incompatible changes in 5.20.
50
516.38 Sun Jun 1 21:54:23 CEST 2014
52 - check that perl slots actually have enough space to hold
53 interpreter variables.
54 - untested port to perl 5.19 (64 bit tmps indices) (reported
55 by Andreas König).
56 - croak when cancel is called without a thread context.
57
586.37 Tue Mar 4 13:27:33 CET 2014
59 - *sigh*, removed leftover debugging code from debugging a
60 perl bug, of all things.
61
626.36 Tue Mar 4 07:11:59 CET 2014
63 - semaphores would not clear the destroy hook when interrupted
64 by ->throw, causing segfaults or worse.
65 - ->throw on a thread waiting for a semaphore did not acquire
66 the semaphore, but also didn't wake up other waiters,
67 possibly causing a deadlock.
68 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
69 to report a full stacktrace, which should help find out
70 where the actual call is.
71 - "atomically" destroy data in slf_destroy, because it is
72 the right thing to do, just in case.
73 - disable ecb.h libm dependency, to compile on ancient systems
74 or under adverse conditions.
75
766.33 Mon Nov 18 11:26:27 CET 2013
77 - do not crash when freeing padlists with holes (in 5.18).
78 - tentative SVt_BIND 5.19 port/fix.
79
806.32 Tue Nov 5 15:35:35 CET 2013
81 - use a new algorithm to derive padlists for perl 5.18. The old
82 one could lead to 0-pointer accesses inside perl (reported
83 by Darin McBride).
84
856.31 Thu May 9 07:39:48 CEST 2013
86 - Coro::AIO requests would crash if the thread was ready'd
87 while the request was ongoing.
88
896.29 Wed May 8 02:55:18 CEST 2013
90 - when an on_destroy handler destructs the coro currently being
91 destructed a perl scalar could be accessed after being freed,
92 likely causing a crash.
93
946.28 Wed Mar 6 06:58:02 CET 2013
95 - clean remnants of existing __DIE__ and __WARN__ handlers so
96 they lose their magic and will not cause segfaults later
97 (testcase by Andrey Sagulin).
98 - improved Coro::State documentation a bit.
99 - Coro::Debug::command now flushes the output.
100 - add hack detection code for x32 abi, because the braindead slugs
101 who designed that made it look exactly like x86_64 without
102 providing proper compile time symbols to test for it. as a result,
103 this detection cannot work reliably.
104 - valgrind stack registering was broken.
105 - do not rely on Time::HiRes anymore in Coro::Debug.
106
1076.23 Fri Dec 7 23:36:37 CET 2012
108 - use experimental fiber implementation on native windows
109 perls.
110 - use sizeof (void *) as multiplication factor for stack sizes,
111 to accomodate the totally braindamaged microsoft 64 bit "os".
112 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
113 activeperl 5.16 crashes when PerlIO_define_layer is called due
114 to some bug in the perl dll, strawberry perl at least passes
115 the testsuite.
116 - implement Coro::Handle->peeraddr/host/port, for slightly
117 improved compatibility with LWP.
118 - implement 5.17 compatibility by almost blindly applying a
119 good-looking patch by Father Chrysostomos.
120 - move stack management functions into libcoro 3.
121 - libcoro version 3 "released".
122 - support magic values as timed_io_once args.
123 - recommend AnyEvent 7+ or EV 4+, also require EV
124 version 4 or newer for Coro::EV.
125
1266.10 Tue Oct 9 01:14:27 CEST 2012
127 - updated ecb.h, it had a typo that caused it to not compile on many
128 big endian systems (reported by many people).
129 - disable memory fences in ecb.h to improve portability.
130
1316.09 Sat Oct 6 23:25:02 CEST 2012
132 - Coro::EV I/O watchers were not interruptible by exceptions
133 (Coro::State::throw) (testcase by sten).
134 - ->throw now puts threads into the ready queue, as this seems to
135 be expected by existing code, and code that doesn't cope with spurious
136 wakeups needs fixing anyway.
137 - use fd -1 in mmap.
138 - cast I32 to int in error message printf.
139 - warn about broken so-called "hardened" kernels.
140
1416.08 Fri Apr 13 12:05:47 CEST 2012
142 - be more aggressive about exiting like perl does - formerly,
143 exiting from the non-main thread would not execute END blocks.
144
1456.07 Fri Nov 11 21:21:48 CET 2011
146 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
147 - update ecb.h.
148
1496.06 Mon Aug 8 23:59:48 CEST 2011
150 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
151 backend, which is an order of magnitude slower. unfortunately, cygwins
152 pthread implementation isn't very complete either, so allocate the stack
153 twice just to be sure.
154 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
155 in its ithreaded perl - best recompile cygwin and use the 'w'indows
156 backend for much better performance. also disable ithreads for
157 even better performance...).
158
1596.05 Thu Aug 4 21:36:36 CEST 2011
160 - blush, condvar values would not be propagated from send to recv anymore
161 (reported by Chip Salzenberg).
162 - use exponential increase for the readline buffer length in
163 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
164
1656.04 Wed Aug 3 17:15:45 CEST 2011
166 - use even more efficient and more compatible condvars for
167 compatibility to AnyEvent 6.x :)
168 - more inconsequential ecb.h updates.
169
1706.03 Wed Aug 3 11:41:30 CEST 2011
171 - change how Coro patches AnyEvent condvars for compatibility to
172 AnyEvent 6.x.
173 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
174 hasn't been included.
175
1766.02 Wed Jul 13 04:35:19 CEST 2011
177 - "improve portability to Gentoo" - gentoo manages to put perl variables
178 in memory areas that are farther than 2gb apart, which the jit couldn't
179 handle and barfed. now it's just a bit slower on gentoo and similar
180 systems.
181
1826.01 Sun Jul 3 12:31:14 CEST 2011
183 - workarounds are good, but the test for whether pthreads are used
184 was not good. this one should be better.
185 - check differently whether gcc generates cfi instructions itself.
186
1876.0 Wed Jun 29 19:43:35 CEST 2011
188 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
189 destroyed and cleaned up automatically (e.g. async { schedule }).
190 - implement a JIT compiler for part of the thread switch code,
191 which gives a 50% speed improvement on threaded perls, and
192 about 4% on non-threaded perls (so threaded perls now finally
193 reach about half the speed of non-threaded perls).
194 - slightly modernise Coro::Intro, add section about rouse functions.
195 - avoid DEFSV and ERRSV, giving another 10% improvement
196 in thread switching.
197 - Coro::State->is_destroyed is now called is_zombie.
198 - implement a Coro->safe_cancel method that might fail, but
199 cancels in a "safer" way if it succeeds.
200 - add preliminary support for DEBUGGING perls.
201 - get rid of two hash-accesses when initialising a new Coro - this
202 speeds up coro creation by almost a factor of two.
203 - croak when a coro that is being cancelled tries to block
204 (e.g. while executing a guard block), instead of crashing or
205 deadlocking.
206 - use a more robust and also faster method to identify Coro::State
207 objects - speeds up everything a bit.
208 - implement Coro->cancel in XS for a 20% speed improvement, and to
209 be able to implement mutual cancellation.
210 - speed up context switches by a percent or two by more efficiently
211 allocating context stack entries.
212 - implement Coro->join and Coro->on_destroy in XS for a speedup and
213 a reduction in memory use.
214 - cancelling a coro while it itself is cancelling another coro is
215 now supported and working, instead of triggering an assertion.
216 - be a bit more crash-resistant when calling (buggy) on_destroy
217 callbacks (best effort).
218 - move on_destroy into the slf_frame, to allow extension slf
219 functions to have destructors.
220 - get rid if coro refcounting - simply crash in other interpreter
221 threads by nulling the pointers on clone.
222 - simplify warn/die hook handling when loading Coro - the convoluted
223 logic seems to be no longer neccessary.
224 - use libecb instead of our own home-grown gcc hacks.
225 - document alternatives to Coro::LWP. Please use them :)
226 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
227 sigaltstack. Really. wine suffers from it, erlang suffers from it,
228 and it's known since at least 2006.
229
2305.372 Wed Feb 23 06:14:30 CET 2011
231 - apparently mingw doesn't provide a working gettimeofday, try to
232 work around that by relying on Time::HiRes (indirectly brought to
233 my attention by Max Maischein).
234 - fix some portability issues when Time::HiRes was used.
235
2365.371 Mon Feb 21 14:36:08 CET 2011
237 - backport to windows process emulation code again.
238
2395.37 Sat Feb 19 07:49:44 CET 2011
240 - add a big "Coro thread life cycle" section to "man Coro".
241 - try a tentative workaround against the breakage that 5.13 has
242 introduced without depreciation period. sigh.
243 - no longer use Time::HiRes if gettimeofday is available, which
244 saves quite a lot of memory.
245
2465.36 Sun Feb 13 05:33:41 CET 2011
247 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
248 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
249 method call.
250 - undocument Coro::Timer::sleep and obsolete whole module.
251 - optimise Coro::Timer::timeout memory and cpu usage.
252 - slightly updated Coro::Intro for recent changes.
253 - do not initialise PL_dirty anymore.
254
2555.25 Thu Nov 11 01:08:39 CET 2010
256 - try a different approach on netbsd - netbsd 5 finally has marginally
257 working pthreads, but still broken ucontext/sigaltstack.
258 - openbsd 4.8 finally got their act together, Coro works out of the box
259 with asm, setjmp and pthreads (no change, just informational).
260
2615.24 Sat Oct 23 11:27:12 CEST 2010
262 - port to the EV 4.0 API.
263 - work around bugs in mingw32, making strawberry perl work
264 out of the box.
265 - correctly modify Coro::AIO function prototypes
266 so that they reflect the "no optional parameters" rule.
267 - "ported" libcoro to C++.
268
2695.23 Mon May 17 18:50:42 CEST 2010
270 - be more resistant to ordering changes when initialising
271 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
272 Waldorf).
273 - document that perl 5.12 deliberately removed support for cloning.
274
2755.22 Wed Apr 14 03:55:35 CEST 2010
276 - correctly return udnef on errors in Coro::Handle::read/write
277 (testcase by Marc Mims).
278 - convert Coro::Util into a "perl compatibility wrapper" - the functions
279 are less useful now, but are drop-in replacements for existing
280 functions, listing better alternatives in the documentation. This also
281 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
282 with Coro::Util::inet_aton.
283 - do not override $Coro::idle unconditionally in Coro.pm, as other
284 modules could have provided their own idle coro already
285 (for exmaple, Coro::AnyEvent).
286 - fix Coro::Util::gethost* functions.
287 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
288
2895.21 Wed Dec 16 07:19:51 CET 2009
290 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
291 together.
292 - add some examples on how to combine other event loops with Coro in
293 Coro::AnyEvent, and how to run it (and not to block). Seems to be
294 the most common source of confusion.
295 - try to catch people naively blocking in an event callback.
296 - work around the perl filehandle bug issue in conjunction with
297 older common::sense (as indirectly pointed out by ZSystem).
298 - clarify the "not from signal handlers" section.
299
3005.2 Sun Oct 4 14:54:24 CEST 2009
301 - Coro::Storable destroyed the prototypes of the functions it wrapped.
302 - export rouse_cb and rouse_wait by default now.
303 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
304 - new method $state->swap_sv.
305 - added section on "windows process emulation" to the manpage,
306 after a not-so-fruitful (nor-friendly) "discussion" with chip
307 salzenberg (discussion implies arguments, but his only arguments
308 were ad-hominems, one wonders why he started it in the first
309 place). I hope this explains it well enough for him to understand,
310 and maybe well enough for others to understand.
311 - use common::sense everywhere now.
312 - idle callbacks are no longer supported, use idle coros instead.
313 - print a thread listing when a deadlock is detected.
314
3155.17 Sat Aug 22 23:09:31 CEST 2009
316 - work around a bug in the perl debugger causing crashes
317 when running under the debugger by marking _pool_handler as nodebug.
318 - speed up Coro::async considerably.
319 - try some hacks to get netbsd to work "more often" - their broken
320 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
321 - convert Coro to AE 5.0 API.
322
3235.162 Tue Jul 28 04:04:03 CEST 2009
324 - perl 5.8.2 is now minimum requirement.
325 - skip t/19_handle.t on broken windows perls.
326
3275.161 Wed Jul 22 04:47:38 CEST 2009
328 - Coro::AnyEvent::poll could have a different prototype when EV was
329 used as backend (analyzed by Tatsuhiko Miyagawa).
330 - Coro::AnyEvent errornously initialised the event loop when loaded,
331 not on demand.
332 - try to workaround rare */t/01_unblock.t failures.
333
3345.16 Tue Jul 21 01:44:37 CEST 2009
335 - Coro::AnyEvent failed to hook into the event loop
336 when no threads had been readied between detecting
337 the event loop and actually running it.
338 - considerably speed up Coro::Select by taking avdantage
339 of AnyEvent > 4.8 and some other optimisations.
340 - implement paragraph readline mode in Coro::Handle
341 (based on patches by Zsbán Ambrus).
342 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
343 and analyzed by Yasuhiro MATSUMOTO).
344 - clarified libcoro license and copyright.
345 - someone stole my EXTRA_META!!!
346 - implement Coro::Select::patch_pp_sselect and it's brother,
347 for hardcode select overriding.
348
3495.151 Mon Jul 6 05:41:57 CEST 2009
350 - backport to windows process emulation code again (patch by
351 Yasuhiro MATSUMOTO).
352 - slightly update Coro::MakeMaker.
353
3545.15 Tue Jun 30 10:28:06 CEST 2009
355 - deprecate Coro::Socket, document how to get ipv6 support via
356 AnyEvent::Socket instead.
357 - implement signal->wait ($cb) interface, similar to semaphores.
358 - work around SvOK not supporting getmagic, so we have to getmagic
359 to test for undef :/ (reported by Matthias Waldorf).
360 - load Coro::AnyEvent in all modules using AnyEvent.
361 - work around perl corrupting our internal data structures,
362 reported by Tokuhiro Matsuno.
363 - enable per-coroutine real and cpu time gathering
364 (Coro::State::enable_times).
365
3665.14 Wed Jun 24 01:37:48 CEST 2009
367 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
368 as this seems to have been a difficult concept (poll, sleep, idle,
369 idle_upto).
370 - add Coro::AnyEvent::readable/writable functions.
371 - clarify Coro::EV/Event/AnyEvent manpages.
372 - free per-thread global scalars in the thread calling ->cancel, to
373 avoid crashes when $_, $@ etc., are magical but some of those
374 had already been freed.
375 - "unexperimentalise" the callback interface for Coro::Semaphore.
376 - speed up ready queue management/context switching by using a linked
377 list instead of an array (~5-10%).
378 - implement "watch" command in Coro::Debug shells.
379 - for fun, implement time-slicing as an example in the manpage.
380 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
381 Coro::Event, use the normal AnyEvent handling instead of dieing
382 (the same is true for Coro::Handle).
383 - properly document Coro::EV::timed_io_once.
384 - avoid unneccessary ->cancel calls in Coro::Handle.
385 - maybe make it work on mingw32 with win32 backend
386 (based on patch by Yasuhiro Matsumoto).
387
3885.132 Fri May 29 09:00:39 CEST 2009
389 - do not keep a reference to the argument itself in
390 Coro::Semaphore::guard, as it could change later.
391 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
392 should support a prepare callback.
393
3945.131 Mon Mar 16 23:20:37 CET 2009
395 - implement and document Coro->suspend, Coro->resume.
396 - fix Coro::Select implementation to not (often) close
397 the passed file descriptors (testcase provided by pippijn).
398
3995.13 Mon Dec 15 21:51:42 CET 2008
400 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
401 - don't set diehook to C<undef> but instead to NULL, to avoid
402 spurious warnings.
403 - fix a lot of bugs in Coro::SemaphoreSet.
404 - Coro::SemaphoreSet will less often create a semaphore needlessly.
405 - add Coro::SemaphoreSet::count and wait methods.
406 - take advantage of the new Guard module.
407 - deprecate Coro::guard.
408 - try to fix the dreaded 01_unblock tests once more. I hate it when
409 testsuites need more fixing than the code they are supposed to test.
410 - croak in more cases when a required callback isn't resolvable.
411 - fix some minor issues in Coro::State->call/eval.
412 - use current coroutine context instead of a temporary one
413 when temporarily switching to another coroutine.
414 - do not call C-level on_destroy handlers during global destruction,
415 to avoid needless segfaults.
416
4175.12 Sun Dec 7 13:30:38 CET 2008
418 - add default config for MirOS, which seems to be a bug-to-bug
419 compatible fork of openbsd ("world domination by releasing
420 openbsd cvs before the openbsd folks do it" or so :).
421 - free_padlist did destroy the names pad, not good, but didn't
422 seem to bother perl - this could fix issues such as eval ""
423 inside a function called from multiple coroutines.
424 - use a different method to detect destruction time.
425 - be more careful when freeing padlists just before global
426 destruction.
427 - fixed and expanded the call/cc example.
428 - renamed _terminate to _coro_run.
429 - new method Coro::Channel->shutdown.
430 - try pthreads on openbsd <4.4 (broken sigaltstack, will
431 pthreads fare better?).
432 - be less picky about destroying "a" running coroutine.
433
4345.11 Tue Nov 25 21:49:05 CET 2008
435 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
436 working code with failed assertions, introducing more bugs than
437 it fixes, requiring elaborate workarounds :(
438
4395.1 Mon Nov 24 08:54:59 CET 2008
440 - wrote a small introductory tutorial - Coro::Intro.
441 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
442 - Coro::Select did errornously dup the file descriptors
443 and didn't work with all AnyEvent backends.
444 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
445 LWP data transfers.
446 - disassociate c contexts from coro objects - this is agruably more
447 correct, but mostly allows sharing of cctxs between coro and state
448 objects, for added memory savings and speed increases.
449 - bumped $Coro::POOL_RSS up to 32kb by default.
450 - no longer set the optype to OP_CUSTOM, as B::* understandably
451 doesn't like this very much (and we *are* a type of entersub).
452 - implement state cloning, just to prove that call/cc can be done.
453 - automatically load Coro::AnyEvent in Coro::Handle.
454 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
455 do not have this method (and don't need it either).
456 - speed up generic anyevent methods in Coro::Handle by using rouse
457 callbacks.
458 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
459 others. It also makes the debugger happier, as you can trace
460 through the idle threads now.
461 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
462 - eg/event was pretty broken.
463 - better 5.8.6 compatibility.
464
4655.0 Thu Nov 20 10:35:05 CET 2008
466 - NEW ARCHITECTURE: use the latest 4.x version if you experience
467 stability issues.
468 - bump API version to 7 - all dependents must be recompiled.
469 - removed timed_* functions - they were not being used anyways
470 and should be replaced by a more generic mechanism -
471 and were annoying to support anyways :)
472 - removed SemaphoreSet's waiter method - use sem method instead.
473 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
474 - async_pool did free a scalar value twice
475 ("Attempt to unreference...").
476 - fix a longstanding bug where calling terminate on a coro that
477 was waiting for a semaphore that was just becoming available
478 would cause a deadlock (semaphore would get into a state where
479 it was available but waiters were still blocked).
480 - calling throw on a coroutine that is waiting for a semaphore will
481 no longer make it acquire the semaphore (and thus leak a count).
482 - perl's process emulation is now not even theoretically supported
483 anymore.
484 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
485 conversion of callback-style to blocking-style.
486 - new methods $coro->schedule_to and ->cede_to, to specifically
487 schedule or cede to a specific coroutine.
488 - new function Coro::Semaphore::wait.
489 - use named constants in Coro::Channel (Richard Hundt).
490 - directly patch the entersub opcode calling SLF functions (cede,
491 transfer and so on). this does speed up context switching, but
492 more importanly, it frees us from the hardcoded behaviour of
493 entersub, so we might actually be able to return something from
494 those functions and atcually create new ones.
495 - take advantage of __builtin_frame_address on gcc.
496 - expose THX in coroapi (not sure whether this was a wise decision,
497 as "threaded" perls are running at half speed anyways).
498 - implement execute_slf (schedule-like-function) interface that makes
499 it possible to implement schedule-like-functions in XS.
500 - use new SLF interface to massively speed up Coro::EV by roughly a
501 factor of two.
502 - used new SLF interface to massively speed up Coro::Semaphore by a
503 factor of three.
504 - used new SLF interface to speed up Coro::AIO by roughly a factor of
505 four and reduce its memory usage considerably.
506 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
507 for a nice speedup and vastly more correct behaviour. Also implement
508 a new method "sem" to get at the underlying semaphore object.
509 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
510 (in comparison) 20-40% speedup.
511 - used new SLF interface to reimplement Coro::Signal gaining
512 some unknown (because I was too lazy), but certain, speedup, and also
513 making signals reliable for the first time.
514 - used new SLF interface and other optimisations to speed up async_pool
515 by a factor of two. It also doesn't rely on perl's exception mechanism
516 to exit anymore. The overhead for terminating an async_pool, coro over
517 a normal async is now very small.
518 - sped up coroutine creation/destruction by 40%.
519 - forgot to include Coro/libcoro/README in the dist for all these years.
520 - work around a freebsd pthreads bug (manual testcancel is required as
521 pthread_cond_wait isn't a cancellation point on freebsd).
522 - use new rouse functions to speed up and simplify Coro::BDB.
523 - make "prefer perl native functions" work with threaded perls.
524 - condense Coro::Debug ps output, hint at v and w flags.
525 - (libcoro) lots of minor cleanups and portability improvements.
526
5274.914 Wed Nov 19 12:54:18 CET 2008
528 - fix a disastrous bug in the readline optimisation
529 introduced in 4.801.
530
5314.913 Sat Nov 15 07:58:28 CET 2008
532 - async_pool did free a scalar value twice
533 ("Attempt to unreference...").
534
5354.912 Thu Nov 13 18:31:23 CET 2008
536 - minor cleanups.
537 - use much larger stacks on linux and perl < 5.8.8.
538 - Coro::Debug::new_unix_server did not unlink the socket
539 when destroyed.
540
5414.911 Tue Nov 11 04:26:17 CET 2008
542 - "port" to threaded perls.
543
5444.91 Mon Nov 10 05:36:38 CET 2008
545 - the ->throw exception object no longer leaks.
546 - creating a new cctx leaked a scopestack entry (memleak).
547 - new coroutines didn't get created with a zero flags field
548 (unknown impact).
549 - calling ->throw on a not-yet-started coroutine should now work
550 instead of being ignored.
551 - ->throw is now supported on Coro::State objects.
552 - clean up cctx creation code a bit.
553 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
554
5554.9 Sat Nov 8 17:45:27 CET 2008
556 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
557 - (libcoro) no longer leak threads in the experimental pthread backend,
558 also speed it up considerably.
559 - (libcoro) do not rely on makecontext passing void *'s unscathed.
560 - fix compiletime dependencies on libcoro in the Makefile.
561 - cctx_count wasn't always updated properly.
562 - Coro::State::cctx_stacksize wasn't applied correctly.
563 - new function Coro::State::cctx_max_idle.
564 - the default max number of idle C contexts is now 4.
565 - (libcoro) try harder to get _setjmp/_longjmp.
566 - (libcoro) cleanup and extend the libcoro API to officially
567 allow the creation of empty source contexts.
568 - very experimental workaround for the totally broken netbsd platform.
569 - tried to hack around openbsd bugs.
570
5714.804 Wed Nov 5 16:36:00 CET 2008
572 - Coro::Debug::new_unix_server would not create a non-blocking listening
573 socket, sometimes causing freezes.
574 - (libcoro) fix misaligned stack points for setjmp and assembly
575 methods, which can cause crashes on x86/x86_64 with a sufficiently
576 aggressive compiler.
577 - new function: Coro::Debug::new_tcp_server.
578 - move ->throw into the Coro class because it only works on coro objects.
579
5804.803 Mon Nov 3 17:16:12 CET 2008
581 - (libcoro) use a global asm statement to become independent of gcc
582 otpimisations for CORO_ASM (thanks to pippijn for the idea).
583 - try to workaround yet another broken bsd, this time dragonfly.
584
5854.802 Thu Oct 30 10:56:12 CET 2008
586 - support -fno-omit-frame-pointer on x86 with the assembly method.
587 - tune 01_unblock.t tests a bit.
588
5894.801 Wed Oct 22 18:33:37 CEST 2008
590 - improve readline speed for very long "lines".
591 - backport to 5.8.8.
592
5934.8 Thu Oct 2 13:34:40 CEST 2008
594 - new function Coro::AIO::aio_wait.
595 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
596 - greatly speed up and reduce memory usage of Coro::AIO requests.
597 - implement some other µ-optimisations.
598
5994.749 Mon Sep 29 14:40:12 CEST 2008
600 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
601
6024.748 Sat Sep 27 14:03:19 CEST 2008
603 - implement, but do not document, PerlIO::cede(granularity).
604 - Coro::Storable forgot to wrap Storable::pstore.
605 - work around the multitude of leaks and memory corruption
606 bugs in PerlIO::via by using our own C-level perliol. As a side
607 effect, Coro::Storable is now much, much, much faster.
608
6094.747 Tue Sep 23 01:59:41 CEST 2008
610 - fix a per-cv memleak (one empty array was leaked per
611 code reference).
612 - avoid a crash in coro->call|rss when the coroutine was already
613 destroyed (most noticably when using Coro::Debug::ps :)
614 - also protect *Storable::FILE.
615 - push up default storable granularity to 20ms.
616
6174.746 Sun Sep 21 03:22:20 CEST 2008
618 - be more insistent on locking Storable against reentrancy
619 in Coro::Storable.
620 - move swap_def?v and throw to Coro::State, as documented.
621
6224.745 Thu Jul 24 00:14:38 CEST 2008
623 - remove debugging code related to MgPV_nolen_const, also try to
624 make it compile with perl 5.8.6 (yes, apple apparently loves
625 outdated software). Reported by John S.
626
6274.744 Tue Jul 8 22:06:35 CEST 2008
628 - correctly provide default DIE/WARN handlers as documented.
629 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
630 implementations inside perl work fine for us.
631
6324.743 Mon Jun 16 00:21:57 CEST 2008
633 - when using Coro::EV without running EV::loop it could
634 result in busy-waiting for events, this has been fixed.
635 - reduce codesize and improve performance by using EV_DEFAULT_UC.
636
6374.742 Sat May 31 14:10:21 CEST 2008
638 - implement a workaround for (some) perl <5.8.8 versions.
639 - require EV 3.3+.
640
6414.741 Fri May 30 23:33:09 CEST 2008
642 - tell netbsd how utterly broken their imitation of an OS is
643 and refuse to build by default if pthreads are in use.
644 - switch to "s" method on all bsd's by default, as their ucontext
645 stuff seems just too broken.
646 - fix a bug in Coro::Select.
647
6484.74 Thu May 29 20:05:31 CEST 2008
649 - do not test Coro::LWP for lack of dependencies.
650 - do not test Coro::Select for lack of working perls.
651
6524.73 Thu May 29 2008
653 - fix a bug in Coro::EV which would cause it to block despite
654 there being runnable coroutines.
8 - sprinkle "no warnings" freely over everything, also suppress 655 - sprinkle "no warnings" freely over everything, also suppress
9 warnings for some other modules. 656 warnings for some other modules.
657 - fix typo in WSAEWOULDBLOCK.
10 658
114.72 Sun May 25 05:14:36 CEST 2008 6594.72 Sun May 25 05:14:36 CEST 2008
12 - tweak META.yaml a bit, unfortunately, there is no documented way 660 - tweak META.yaml a bit, unfortunately, there is no documented way
13 to have optional dependencies with CPAN. doh :( 661 to have optional dependencies with CPAN. doh :(
14 - avoid running some tests on windows because they would fail due to 662 - avoid running some tests on windows because they would fail due to
15 perl bug (broken fork, broken pipes...). 663 perl bug (broken fork, broken pipes...).
16 - work around perl on windows bugs where perl returns undocumented 664 - work around perl on windows bugs where perl returns undocumented
17 error codes for sysread, syswrite etc. by taking advantage 665 error codes for sysread, syswrite etc. by taking advantage
… …
21 - use unix domain sockets in testsuite to work around 669 - use unix domain sockets in testsuite to work around
22 common perl implementation bugs on widows (they are emulated by 670 common perl implementation bugs on widows (they are emulated by
23 tcp sockets on windows. ugh.) 671 tcp sockets on windows. ugh.)
24 672
254.71 Sat May 24 20:01:27 CEST 2008 6734.71 Sat May 24 20:01:27 CEST 2008
26 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 674 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
27 - take advantage of async name resolution of AnyEvent::Util. 675 - take advantage of async name resolution of AnyEvent::Util.
28 - work around brutal inet_aton override in Coro::LWP. 676 - work around brutal inet_aton override in Coro::LWP.
29 - take advantage of the readyhook in Coro::EV, for smoother 677 - take advantage of the readyhook in Coro::EV, for smoother
30 scheduling. 678 scheduling.
31 679
324.7 Sun May 11 00:32:19 CEST 2008 6804.7 Sun May 11 00:32:19 CEST 2008
33 - completely reworked the Coro manpage. 681 - completely reworked the Coro manpage.
34 - added Coro::AnyEvent, generic event loop integration. 682 - added Coro::AnyEvent, generic event loop integration.
35 - implement cancel, ready and kill commands in Coro::Debug. 683 - implement cancel, ready and kill commands in Coro::Debug.
36 - document find_coro in Coro::Debug. 684 - document find_coro in Coro::Debug.
37 - incompatible change: rename has_stack to has_cctx. 685 - incompatible change: rename has_stack to has_cctx.
38 - Coro::AIO and Coro::BDB no longer force event model detection, 686 - Coro::AIO and Coro::BDB no longer force event model detection,
39 use AnyEvent::AIO and AnyEvent::BDB. 687 use AnyEvent::AIO and AnyEvent::BDB.
… …
163 811
1644.11 Thu Oct 11 02:40:24 CEST 2007 8124.11 Thu Oct 11 02:40:24 CEST 2007
165 - port to threaded perls. 813 - port to threaded perls.
166 814
1674.1 Thu Oct 11 02:38:16 CEST 2007 8154.1 Thu Oct 11 02:38:16 CEST 2007
168 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 816 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
169 be local to each coro (see Coro::State). 817 be local to each coro (see Coro::State).
170 - incompatible change: for very deep reasons, cede and cede_notself 818 - incompatible change: for very deep reasons, cede and cede_notself
171 cannot return anything, so nothing will be returned. 819 cannot return anything, so nothing will be returned.
172 - possibly bring back 5.10 compatibility (untested). 820 - possibly bring back 5.10 compatibility (untested).
173 - work around stupid (and wrong) warning on 5.10 :(. 821 - work around stupid (and wrong) warning on 5.10 :(.
… …
245 but stateful semantics. 893 but stateful semantics.
246 - fixed a lot of typos in Coro.pm (patch submitted by David 894 - fixed a lot of typos in Coro.pm (patch submitted by David
247 Steinbrunner, which applied flawlessly). 895 Steinbrunner, which applied flawlessly).
248 896
2493.6 Sat Apr 14 17:13:31 CEST 2007 8973.6 Sat Apr 14 17:13:31 CEST 2007
250 - added some bugfixes to get eg/myhttpd working again. 898 - added some bugfixes to get eg/myhttpd working again.
251 - added Coro::Storable for often-cede'ing freeze/thaw. 899 - added Coro::Storable for often-cede'ing freeze/thaw.
252 - try to do a clean exit when a coroutine calls exit 900 - try to do a clean exit when a coroutine calls exit
253 (EXPERIMENTAL). 901 (EXPERIMENTAL).
254 - got rid of indirect call through _coro_init. 902 - got rid of indirect call through _coro_init.
255 - updated the partly antique examples in eg/ to 903 - updated the partly antique examples in eg/ to

Diff Legend

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