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

Comparing Coro/Changes (file contents):
Revision 1.202 by root, Sat Jan 20 19:52:15 2007 UTC vs.
Revision 1.625 by root, Tue Mar 4 05:07:44 2014 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: better (earlier) error messages on broken transfer's 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: destroy gets only called on cancel: fix(?) 4TODO: channel->maxsize(newsize)?
5TODO: fast reloads in cf cause segfaults (coro_state_free->destroy->crash) 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6TODO: swap_sv, maybe add scope_swap_sv, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
6 8
73.41 9Program terminated with signal 11, Segmentation fault.
10#0 coro_semaphore_adjust (av=0x0, adjust=0) at State.xs:2915
112915 State.xs: No such file or directory.
12(gdb) bt
13#0 coro_semaphore_adjust (av=0x0, adjust=0) at State.xs:2915
14#1 0x00007f638f0c0703 in slf_destroy (coro=0x5957180) at State.xs:2605
15#2 coro_state_destroy (coro=0x5957180) at State.xs:1712
16#3 0x00007f638f0c150c in coro_state_destroy (coro=<optimized out>) at State.xs:1709
17#4 XS_Coro__destroy (cv=<optimized out>) at State.xs:3853
18#5 0x00000000004b572d in Perl_pp_entersub () at pp_hot.c:2881
19#6 0x00000000004adfe3 in Perl_runops_standard () at run.c:42
20#7 0x000000000044335f in S_run_body (oldscope=<optimized out>) at perl.c:2462
21#8 perl_run (my_perl=<optimized out>) at perl.c:2383
22#9 0x00007f638f0ba3a1 in cctx_run (arg=<optimized out>) at State.xs:1478
23#10 0x00007f638f0b74d5 in coro_init () at libcoro/coro.c:96
24(gdb) up
25#1 0x00007f638f0c0703 in slf_destroy (coro=0x5957180) at State.xs:2605
262605 in State.xs
27(gdb) p coro->slf_frame
28$1 = {
29 prepare = 0x7f638f0baf60 <prepare_schedule>,
30 check = 0x7f638f0b6e70 <slf_check_repeat>,
31 data = 0x0,
32 destroy = 0x7f638f0bb820 <coro_semaphore_destroy>
33
34
35TODO: warnhook reports from wrong line number now, when arg is Math::GMP number.
36 - FATAL: $Coro::IDLE blocked itself will now use Carp::confess
37 to report a full stacktrace, which should help find out
38 where the actual call is.
39 - "atomically" destroy data in slf_destroy, because it is
40 the right thing to do, just in case.
41 - disable ecb.h libm dependency, to compile on ancient systems
42 or under adverse conditions.
43
446.33 Mon Nov 18 11:26:27 CET 2013
45 - do not crash when freeing padlists with holes (in 5.18).
46 - tentative SVt_BIND 5.19 port/fix.
47
486.32 Tue Nov 5 15:35:35 CET 2013
49 - use a new algorithm to derive padlists for perl 5.18. The old
50 one could lead to 0-pointer accesses inside perl (reported
51 by Darin McBride).
52
536.31 Thu May 9 07:39:48 CEST 2013
54 - Coro::AIO requests would crash if the thread was ready'd
55 while the request was ongoing.
56
576.29 Wed May 8 02:55:18 CEST 2013
58 - when an on_destroy handler destructs the coro currently being
59 destructed a perl scalar could be accessed after being freed,
60 likely causing a crash.
61
626.28 Wed Mar 6 06:58:02 CET 2013
63 - clean remnants of existing __DIE__ and __WARN__ handlers so
64 they lose their magic and will not cause segfaults later
65 (testcase by Andrey Sagulin).
66 - improved Coro::State documentation a bit.
67 - Coro::Debug::command now flushes the output.
68 - add hack detection code for x32 abi, because the braindead slugs
69 who designed that made it look exactly like x86_64 without
70 providing proper compile time symbols to test for it. as a result,
71 this detection cannot work reliably.
72 - valgrind stack registering was broken.
73 - do not rely on Time::HiRes anymore in Coro::Debug.
74
756.23 Fri Dec 7 23:36:37 CET 2012
76 - use experimental fiber implementation on native windows
77 perls.
78 - use sizeof (void *) as multiplication factor for stack sizes,
79 to accomodate the totally braindamaged microsoft 64 bit "os".
80 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
81 activeperl 5.16 crashes when PerlIO_define_layer is called due
82 to some bug in the perl dll, strawberry perl at least passes
83 the testsuite.
84 - implement Coro::Handle->peeraddr/host/port, for slightly
85 improved compatibility with LWP.
86 - implement 5.17 compatibility by almost blindly applying a
87 good-looking patch by Father Chrysostomos.
88 - move stack management functions into libcoro 3.
89 - libcoro version 3 "released".
90 - support magic values as timed_io_once args.
91 - recommend AnyEvent 7+ or EV 4+, also require EV
92 version 4 or newer for Coro::EV.
93
946.10 Tue Oct 9 01:14:27 CEST 2012
95 - updated ecb.h, it had a typo that caused it to not compile on many
96 big endian systems (reported by many people).
97 - disable memory fences in ecb.h to improve portability.
98
996.09 Sat Oct 6 23:25:02 CEST 2012
100 - Coro::EV I/O watchers were not interruptible by exceptions
101 (Coro::State::throw) (testcase by sten).
102 - ->throw now puts threads into the ready queue, as this seems to
103 be expected by existing code, and code that doesn't cope with spurious
104 wakeups needs fixing anyway.
105 - use fd -1 in mmap.
106 - cast I32 to int in error message printf.
107 - warn about broken so-called "hardened" kernels.
108
1096.08 Fri Apr 13 12:05:47 CEST 2012
110 - be more aggressive about exiting like perl does - formerly,
111 exiting from the non-main thread would not execute END blocks.
112
1136.07 Fri Nov 11 21:21:48 CET 2011
114 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
115 - update ecb.h.
116
1176.06 Mon Aug 8 23:59:48 CEST 2011
118 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
119 backend, which is an order of magnitude slower. unfortunately, cygwins
120 pthread implementation isn't very complete either, so allocate the stack
121 twice just to be sure.
122 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
123 in its ithreaded perl - best recompile cygwin and use the 'w'indows
124 backend for much better performance. also disable ithreads for
125 even better performance...).
126
1276.05 Thu Aug 4 21:36:36 CEST 2011
128 - blush, condvar values would not be propagated from send to recv anymore
129 (reported by Chip Salzenberg).
130 - use exponential increase for the readline buffer length in
131 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
132
1336.04 Wed Aug 3 17:15:45 CEST 2011
134 - use even more efficient and more compatible condvars for
135 compatibility to AnyEvent 6.x :)
136 - more inconsequential ecb.h updates.
137
1386.03 Wed Aug 3 11:41:30 CEST 2011
139 - change how Coro patches AnyEvent condvars for compatibility to
140 AnyEvent 6.x.
141 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
142 hasn't been included.
143
1446.02 Wed Jul 13 04:35:19 CEST 2011
145 - "improve portability to Gentoo" - gentoo manages to put perl variables
146 in memory areas that are farther than 2gb apart, which the jit couldn't
147 handle and barfed. now it's just a bit slower on gentoo and similar
148 systems.
149
1506.01 Sun Jul 3 12:31:14 CEST 2011
151 - workarounds are good, but the test for whether pthreads are used
152 was not good. thisone should be better.
153 - check differently whether gcc generates cfi instructions itself.
154
1556.0 Wed Jun 29 19:43:35 CEST 2011
156 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
157 destroyed and cleaned up automatically (e.g. async { schedule }).
158 - implement a JIT compiler for part of the thread switch code,
159 which gives a 50% speed improvement on threaded perls, and
160 about 4% on non-thraeded perls (so threaded perls now finally
161 reach about half the speed of non-threaded perls).
162 - slightly modernise Coro::Intro, add section about rouse functions.
163 - avoid DEFSV and ERRSV, giving another 10% improvement
164 in thread switching.
165 - Coro::State->is_destroyed is now called is_zombie.
166 - implement a Coro->safe_cancel method that might fail, but
167 cancels in a "safer" way if it succeeds.
168 - add preliminary support for DEBUGGING perls.
169 - get rid of two hash-accesses when initialising a new Coro - this
170 speeds up coro creation by almost a factor of two.
171 - croak when a coro that is being cancelled tries to block
172 (e.g. while executing a guard block), instead of crashing or
173 deadlocking.
174 - use a more robust and also faster method to identify Coro::State
175 objects - speeds up everything a bit.
176 - implement Coro->cancel in XS for a 20% speed improvement, and to
177 be able to implement mutual cancellation.
178 - speed up context switches by a percent or two by more efficiently
179 allocating context stack entries.
180 - implement Coro->join and Coro->on_destroy in XS for a speedup and
181 a reduction in memory use.
182 - cancelling a coro while it itself is cancelling another coro is
183 now supported and working, instead of triggering an assertion.
184 - be a bit more crash-resistant when calling (buggy) on_destroy
185 callbacks (best effort).
186 - move on_destroy into the slf_frame, to allow extension slf
187 functions to have destructors.
188 - get rid if coro refcounting - simply crash in other interpreter
189 threads by nulling the pointers on clone.
190 - simplify warn/die hook handling when loading Coro - the convoluted
191 logic seems to be no longer neccessary.
192 - use libecb instead of our own home-grown gcc hacks.
193 - document alternatives to Coro::LWP. Please use them :)
194 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
195 sigaltstack. Really. wine suffers from it, erlang suffers from it,
196 and it's known since at least 2006.
197
1985.372 Wed Feb 23 06:14:30 CET 2011
199 - apparently mingw doesn't provide a working gettimeofday, try to
200 work around that by relying on Time::HiRes (indirectly brought to
201 my attention by Max Maischein).
202 - fix some portability issues when Time::HiRes was used.
203
2045.371 Mon Feb 21 14:36:08 CET 2011
205 - backport to windows process emulation code again.
206
2075.37 Sat Feb 19 07:49:44 CET 2011
208 - add a big "Coro thread life cycle" section to "man Coro".
209 - try a tentative workaround against the breakage that 5.13 has
210 introduced without depreciation period. sigh.
211 - no longer use Time::HiRes if gettimeofday is available, which
212 saves quite a lot of memory.
213
2145.36 Sun Feb 13 05:33:41 CET 2011
215 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
216 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
217 method call.
218 - undocument Coro::Timer::sleep and obsolete whole module.
219 - optimise Coro::Timer::timeout memory and cpu usage.
220 - slightly updated Coro::Intro for recent changes.
221 - do not initialise PL_dirty anymore.
222
2235.25 Thu Nov 11 01:08:39 CET 2010
224 - try a different approach on netbsd - netbsd 5 finally has marginally
225 working pthreads, but still broken ucontext/sigaltstack.
226 - openbsd 4.8 finally got their act together, Coro works out of the box
227 with asm, setjmp and pthreads (no change, just informational).
228
2295.24 Sat Oct 23 11:27:12 CEST 2010
230 - port to the EV 4.0 API.
231 - work around bugs in mingw32, making strawberry perl work
232 out of the box.
233 - correctly modify Coro::AIO function prototypes
234 so that they reflect the "no optional parameters" rule.
235 - "ported" libcoro to C++.
236
2375.23 Mon May 17 18:50:42 CEST 2010
238 - be more resistant to ordering changes when initialising
239 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
240 Waldorf).
241 - document that perl 5.12 deliberately removed support for cloning.
242
2435.22 Wed Apr 14 03:55:35 CEST 2010
244 - correctly return udnef on errors in Coro::Handle::read/write
245 (testcase by Marc Mims).
246 - convert Coro::Util into a "perl compatibility wrapper" - the functions
247 are less useful now, but are drop-in replacements for existing
248 functions, listing better alternatives in the documentation. This also
249 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
250 with Coro::Util::inet_aton.
251 - do not override $Coro::idle unconditionally in Coro.pm, as other
252 modules could have provided their own idle coro already
253 (for exmaple, Coro::AnyEvent).
254 - fix Coro::Util::gethost* functions.
255 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
256
2575.21 Wed Dec 16 07:19:51 CET 2009
258 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
259 together.
260 - add some examples on how to combine other event loops with Coro in
261 Coro::AnyEvent, and how to run it (and not to block). Seems to be
262 the most common source of confusion.
263 - try to catch people naively blocking in an event callback.
264 - work around the perl filehandle bug issue in conjunction with
265 older common::sense (as indirectly pointed out by ZSystem).
266 - clarify the "not from signal handlers" section.
267
2685.2 Sun Oct 4 14:54:24 CEST 2009
269 - Coro::Storable destroyed the prototypes of the functions it wrapped.
270 - export rouse_cb and rouse_wait by default now.
271 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
272 - new method $state->swap_sv.
273 - added section on "windows process emulation" to the manpage,
274 after a not-so-fruitful (nor-friendly) "discussion" with chip
275 salzenberg (discussion implies arguments, but his only arguments
276 were ad-hominems, one wonders why he started it in the first
277 place). I hope this explains it well enough for him to understand,
278 and maybe well enough for others to understand.
279 - use common::sense everywhere now.
280 - idle callbacks are no longer supported, use idle coros instead.
281 - print a thread listing when a deadlock is detected.
282
2835.17 Sat Aug 22 23:09:31 CEST 2009
284 - work around a bug in the perl debugger causing crashes
285 when running under the debugger by marking _pool_handler as nodebug.
286 - speed up Coro::async considerably.
287 - try some hacks to get netbsd to work "more often" - their broken
288 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
289 - convert Coro to AE 5.0 API.
290
2915.162 Tue Jul 28 04:04:03 CEST 2009
292 - perl 5.8.2 is now minimum requirement.
293 - skip t/19_handle.t on broken windows perls.
294
2955.161 Wed Jul 22 04:47:38 CEST 2009
296 - Coro::AnyEvent::poll could have a different prototype when EV was
297 used as backend (analyzed by Tatsuhiko Miyagawa).
298 - Coro::AnyEvent errornously initialised the event loop when loaded,
299 not on demand.
300 - try to workaround rare */t/01_unblock.t failures.
301
3025.16 Tue Jul 21 01:44:37 CEST 2009
303 - Coro::AnyEvent failed to hook into the event loop
304 when no threads had been readied between detecting
305 the event loop and actually running it.
306 - considerably speed up Coro::Select by taking avdantage
307 of AnyEvent > 4.8 and some other optimisations.
308 - implement paragraph readline mode in Coro::Handle
309 (based on patches by Zsbán Ambrus).
310 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
311 and analyzed by Yasuhiro MATSUMOTO).
312 - clarified libcoro license and copyright.
313 - someone stole my EXTRA_META!!!
314 - implement Coro::Select::patch_pp_sselect and it's brother,
315 for hardcode select overriding.
316
3175.151 Mon Jul 6 05:41:57 CEST 2009
318 - backport to windows process emulation code again (patch by
319 Yasuhiro MATSUMOTO).
320 - slightly update Coro::MakeMaker.
321
3225.15 Tue Jun 30 10:28:06 CEST 2009
323 - deprecate Coro::Socket, document how to get ipv6 support via
324 AnyEvent::Socket instead.
325 - implement signal->wait ($cb) interface, similar to semaphores.
326 - work around SvOK not supporting getmagic, so we have to getmagic
327 to test for undef :/ (reported by Matthias Waldorf).
328 - load Coro::AnyEvent in all modules using AnyEvent.
329 - work around perl corrupting our internal data structures,
330 reported by Tokuhiro Matsuno.
331 - enable per-coroutine real and cpu time gathering
332 (Coro::State::enable_times).
333
3345.14 Wed Jun 24 01:37:48 CEST 2009
335 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
336 as this seems to have been a difficult concept (poll, sleep, idle,
337 idle_upto).
338 - add Coro::AnyEvent::readable/writable functions.
339 - clarify Coro::EV/Event/AnyEvent manpages.
340 - free per-thread global scalars in the thread calling ->cancel, to
341 avoid crashes when $_, $@ etc., are magical but some of those
342 had already been freed.
343 - "unexperimentalise" the callback interface for Coro::Semaphore.
344 - speed up ready queue management/context switching by using a linked
345 list instead of an array (~5-10%).
346 - implement "watch" command in Coro::Debug shells.
347 - for fun, implement time-slicing as an example in the manpage.
348 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
349 Coro::Event, use the normal AnyEvent handling instead of dieing
350 (the same is true for Coro::Handle).
351 - properly document Coro::EV::timed_io_once.
352 - avoid unneccessary ->cancel calls in Coro::Handle.
353 - maybe make it work on mingw32 with win32 backend
354 (based on patch by Yasuhiro Matsumoto).
355
3565.132 Fri May 29 09:00:39 CEST 2009
357 - do not keep a reference to the argument itself in
358 Coro::Semaphore::guard, as it could change later.
359 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
360 should support a prepare callback.
361
3625.131 Mon Mar 16 23:20:37 CET 2009
363 - implement and document Coro->suspend, Coro->resume.
364 - fix Coro::Select implementation to not (often) close
365 the passed file descriptors (testcase provided by pippijn).
366
3675.13 Mon Dec 15 21:51:42 CET 2008
368 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
369 - don't set diehook to C<undef> but instead to NULL, to avoid
370 spurious warnings.
371 - fix a lot of bugs in Coro::SemaphoreSet.
372 - Coro::SemaphoreSet will less often create a semaphore needlessly.
373 - add Coro::SemaphoreSet::count and wait methods.
374 - take advantage of the new Guard module.
375 - deprecate Coro::guard.
376 - try to fix the dreaded 01_unblock tests once more. I hate it when
377 testsuites need more fixing than the code they are supposed to test.
378 - croak in more cases when a required callback isn't resolvable.
379 - fix some minor issues in Coro::State->call/eval.
380 - use current coroutine context instead of a temporary one
381 when temporarily switching to another coroutine.
382 - do not call C-level on_destroy handlers during global destruction,
383 to avoid needless segfaults.
384
3855.12 Sun Dec 7 13:30:38 CET 2008
386 - add default config for MirOS, which seems to be a bug-to-bug
387 compatible fork of openbsd ("world domination by releasing
388 openbsd cvs before the openbsd folks do it" or so :).
389 - free_padlist did destroy the names pad, not good, but didn't
390 seem to bother perl - this could fix issues such as eval ""
391 inside a function called from multiple coroutines.
392 - use a different method to detect destruction time.
393 - be more careful when freeing padlists just before global
394 destruction.
395 - fixed and expanded the call/cc example.
396 - renamed _terminate to _coro_run.
397 - new method Coro::Channel->shutdown.
398 - try pthreads on openbsd <4.4 (broken sigaltstack, will
399 pthreads fare better?).
400 - be less picky about destroying "a" running coroutine.
401
4025.11 Tue Nov 25 21:49:05 CET 2008
403 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
404 working code with failed assertions, introducing more bugs than
405 it fixes, requiring elaborate workarounds :(
406
4075.1 Mon Nov 24 08:54:59 CET 2008
408 - wrote a small introductory tutorial - Coro::Intro.
409 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
410 - Coro::Select did errornously dup the file descriptors
411 and didn't work with all AnyEvent backends.
412 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
413 LWP data transfers.
414 - disassociate c contexts from coro objects - this is agruably more
415 correct, but mostly allows sharing of cctxs between coro and state
416 objects, for added memory savings and speed increases.
417 - bumped $Coro::POOL_RSS up to 32kb by default.
418 - no longer set the optype to OP_CUSTOM, as B::* understandably
419 doesn't like this very much (and we *are* a type of entersub).
420 - implement state cloning, just to prove that call/cc can be done.
421 - automatically load Coro::AnyEvent in Coro::Handle.
422 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
423 do not have this method (and don't need it either).
424 - speed up generic anyevent methods in Coro::Handle by using rouse
425 callbacks.
426 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
427 others. It also makes the debugger happier, as you can trace
428 through the idle threads now.
429 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
430 - eg/event was pretty broken.
431 - better 5.8.6 compatibility.
432
4335.0 Thu Nov 20 10:35:05 CET 2008
434 - NEW ARCHITECTURE: use the latest 4.x version if you experience
435 stability issues.
436 - bump API version to 7 - all dependents must be recompiled.
437 - removed timed_* functions - they were not being used anyways
438 and should be replaced by a more generic mechanism -
439 and were annoying to support anyways :)
440 - removed SemaphoreSet's waiter method - use sem method instead.
441 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
442 - async_pool did free a scalar value twice
443 ("Attempt to unreference...").
444 - fix a longstanding bug where calling terminate on a coro that
445 was waiting for a semaphore that was just becoming available
446 would cause a deadlock (semaphore would get into a state where
447 it was available but waiters were still blocked).
448 - calling throw on a coroutine that is waiting for a semaphore will
449 no longer make it acquire the semaphore (and thus leak a count).
450 - perl's process emulation is now not even theoretically supported
451 anymore.
452 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
453 conversion of callback-style to blocking-style.
454 - new methods $coro->schedule_to and ->cede_to, to specifically
455 schedule or cede to a specific coroutine.
456 - new function Coro::Semaphore::wait.
457 - use named constants in Coro::Channel (Richard Hundt).
458 - directly patch the entersub opcode calling SLF functions (cede,
459 transfer and so on). this does speed up context switching, but
460 more importanly, it frees us from the hardcoded behaviour of
461 entersub, so we might actually be able to return something from
462 those functions and atcually create new ones.
463 - take advantage of __builtin_frame_address on gcc.
464 - expose THX in coroapi (not sure whether this was a wise decision,
465 as "threaded" perls are running at half speed anyways).
466 - implement execute_slf (schedule-like-function) interface that makes
467 it possible to implement schedule-like-functions in XS.
468 - use new SLF interface to massively speed up Coro::EV by roughly a
469 factor of two.
470 - used new SLF interface to massively speed up Coro::Semaphore by a
471 factor of three.
472 - used new SLF interface to speed up Coro::AIO by roughly a factor of
473 four and reduce its memory usage considerably.
474 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
475 for a nice speedup and vastly more correct behaviour. Also implement
476 a new method "sem" to get at the underlying semaphore object.
477 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
478 (in comparison) 20-40% speedup.
479 - used new SLF interface to reimplement Coro::Signal gaining
480 some unknown (because I was too lazy), but certain, speedup, and also
481 making signals reliable for the first time.
482 - used new SLF interface and other optimisations to speed up async_pool
483 by a factor of two. It also doesn't rely on perl's exception mechanism
484 to exit anymore. The overhead for terminating an async_pool, coro over
485 a normal async is now very small.
486 - sped up coroutine creation/destruction by 40%.
487 - forgot to include Coro/libcoro/README in the dist for all these years.
488 - work around a freebsd pthreads bug (manual testcancel is required as
489 pthread_cond_wait isn't a cancellation point on freebsd).
490 - use new rouse functions to speed up and simplify Coro::BDB.
491 - make "prefer perl native functions" work with threaded perls.
492 - condense Coro::Debug ps output, hint at v and w flags.
493 - (libcoro) lots of minor cleanups and portability improvements.
494
4954.914 Wed Nov 19 12:54:18 CET 2008
496 - fix a disastrous bug in the readline optimisation
497 introduced in 4.801.
498
4994.913 Sat Nov 15 07:58:28 CET 2008
500 - async_pool did free a scalar value twice
501 ("Attempt to unreference...").
502
5034.912 Thu Nov 13 18:31:23 CET 2008
504 - minor cleanups.
505 - use much larger stacks on linux and perl < 5.8.8.
506 - Coro::Debug::new_unix_server did not unlink the socket
507 when destroyed.
508
5094.911 Tue Nov 11 04:26:17 CET 2008
510 - "port" to threaded perls.
511
5124.91 Mon Nov 10 05:36:38 CET 2008
513 - the ->throw exception object no longer leaks.
514 - creating a new cctx leaked a scopestack entry (memleak).
515 - new coroutines didn't get created with a zero flags field
516 (unknown impact).
517 - calling ->throw on a not-yet-started coroutine should now work
518 instead of being ignored.
519 - ->throw is now supported on Coro::State objects.
520 - clean up cctx creation code a bit.
521 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
522
5234.9 Sat Nov 8 17:45:27 CET 2008
524 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
525 - (libcoro) no longer leak threads in the experimental pthread backend,
526 also speed it up considerably.
527 - (libcoro) do not rely on makecontext passing void *'s unscathed.
528 - fix compiletime dependencies on libcoro in the Makefile.
529 - cctx_count wasn't always updated properly.
530 - Coro::State::cctx_stacksize wasn't applied correctly.
531 - new function Coro::State::cctx_max_idle.
532 - the default max number of idle C contexts is now 4.
533 - (libcoro) try harder to get _setjmp/_longjmp.
534 - (libcoro) cleanup and extend the libcoro API to officially
535 allow the creation of empty source contexts.
536 - very experimental workaround for the totally broken netbsd platform.
537 - tried to hack around openbsd bugs.
538
5394.804 Wed Nov 5 16:36:00 CET 2008
540 - Coro::Debug::new_unix_server would not create a non-blocking listening
541 socket, sometimes causing freezes.
542 - (libcoro) fix misaligned stack points for setjmp and assembly
543 methods, which can cause crashes on x86/x86_64 with a sufficiently
544 aggressive compiler.
545 - new function: Coro::Debug::new_tcp_server.
546 - move ->throw into the Coro class because it only works on coro objects.
547
5484.803 Mon Nov 3 17:16:12 CET 2008
549 - (libcoro) use a global asm statement to become independent of gcc
550 otpimisations for CORO_ASM (thanks to pippijn for the idea).
551 - try to workaround yet another broken bsd, this time dragonfly.
552
5534.802 Thu Oct 30 10:56:12 CET 2008
554 - support -fno-omit-frame-pointer on x86 with the assembly method.
555 - tune 01_unblock.t tests a bit.
556
5574.801 Wed Oct 22 18:33:37 CEST 2008
558 - improve readline speed for very long "lines".
559 - backport to 5.8.8.
560
5614.8 Thu Oct 2 13:34:40 CEST 2008
562 - new function Coro::AIO::aio_wait.
563 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
564 - greatly speed up and reduce memory usage of Coro::AIO requests.
565 - implement some other µ-optimisations.
566
5674.749 Mon Sep 29 14:40:12 CEST 2008
568 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
569
5704.748 Sat Sep 27 14:03:19 CEST 2008
571 - implement, but do not document, PerlIO::cede(granularity).
572 - Coro::Storable forgot to wrap Storable::pstore.
573 - work around the multitude of leaks and memory corruption
574 bugs in PerlIO::via by using our own C-level perliol. As a side
575 effect, Coro::Storable is now much, much, much faster.
576
5774.747 Tue Sep 23 01:59:41 CEST 2008
578 - fix a per-cv memleak (one empty array was leaked per
579 code reference).
580 - avoid a crash in coro->call|rss when the coroutine was already
581 destroyed (most noticably when using Coro::Debug::ps :)
582 - also protect *Storable::FILE.
583 - push up default storable granularity to 20ms.
584
5854.746 Sun Sep 21 03:22:20 CEST 2008
586 - be more insistent on locking Storable against reentrancy
587 in Coro::Storable.
588 - move swap_def?v and throw to Coro::State, as documented.
589
5904.745 Thu Jul 24 00:14:38 CEST 2008
591 - remove debugging code related to MgPV_nolen_const, also try to
592 make it compile with perl 5.8.6 (yes, apple apparently loves
593 outdated software). Reported by John S.
594
5954.744 Tue Jul 8 22:06:35 CEST 2008
596 - correctly provide default DIE/WARN handlers as documented.
597 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
598 implementations inside perl work fine for us.
599
6004.743 Mon Jun 16 00:21:57 CEST 2008
601 - when using Coro::EV without running EV::loop it could
602 result in busy-waiting for events, this has been fixed.
603 - reduce codesize and improve performance by using EV_DEFAULT_UC.
604
6054.742 Sat May 31 14:10:21 CEST 2008
606 - implement a workaround for (some) perl <5.8.8 versions.
607 - require EV 3.3+.
608
6094.741 Fri May 30 23:33:09 CEST 2008
610 - tell netbsd how utterly broken their imitation of an OS is
611 and refuse to build by default if pthreads are in use.
612 - switch to "s" method on all bsd's by default, as their ucontext
613 stuff seems just too broken.
614 - fix a bug in Coro::Select.
615
6164.74 Thu May 29 20:05:31 CEST 2008
617 - do not test Coro::LWP for lack of dependencies.
618 - do not test Coro::Select for lack of working perls.
619
6204.73 Thu May 29 2008
621 - fix a bug in Coro::EV which would cause it to block despite
622 there being runnable coroutines.
623 - sprinkle "no warnings" freely over everything, also suppress
624 warnings for some other modules.
625 - fix typo in WSAEWOULDBLOCK.
626
6274.72 Sun May 25 05:14:36 CEST 2008
628 - tweak META.yaml a bit, unfortunately, there is no documented way
629 to have optional dependencies with CPAN. doh :(
630 - avoid running some tests on windows because they would fail due to
631 perl bug (broken fork, broken pipes...).
632 - work around perl on windows bugs where perl returns undocumented
633 error codes for sysread, syswrite etc. by taking advantage
634 of AnyEvent's workaround for that problem.
635 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
636 a common perl implementation bug on windows.
637 - use unix domain sockets in testsuite to work around
638 common perl implementation bugs on widows (they are emulated by
639 tcp sockets on windows. ugh.)
640
6414.71 Sat May 24 20:01:27 CEST 2008
642 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
643 - take advantage of async name resolution of AnyEvent::Util.
644 - work around brutal inet_aton override in Coro::LWP.
645 - take advantage of the readyhook in Coro::EV, for smoother
646 scheduling.
647
6484.7 Sun May 11 00:32:19 CEST 2008
649 - completely reworked the Coro manpage.
650 - added Coro::AnyEvent, generic event loop integration.
651 - implement cancel, ready and kill commands in Coro::Debug.
652 - document find_coro in Coro::Debug.
653 - incompatible change: rename has_stack to has_cctx.
654 - Coro::AIO and Coro::BDB no longer force event model detection,
655 use AnyEvent::AIO and AnyEvent::BDB.
656
6574.6 Sat Apr 26 10:05:14 CEST 2008
658 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
659 by the Coro module. It was a mistake to have it in the first place.
660 - (experimental) support for activestate perl 5.10 (method "w").
661 - (experimental) support for strawberry perl (method "a").
662 - coro_sigelem_set did not return a value although it had to,
663 actual impact unknown.
664
6654.51 Mon Apr 14 13:28:27 CEST 2008
666 - make it compile again on 5.8.
667
6684.50 Thu Apr 10 09:43:17 CEST 2008
669 - I did it twice! (see 4.49).
670
6714.49 Sun Apr 6 21:23:31 CEST 2008
672 - grr, instead of compiling the recent changes on 5.10 only they
673 were compiled on 5.8 only.
674
6754.48 Sun Apr 6 20:36:46 CEST 2008
676 - allow coroutine switches during eval's under 5.10.x, as it seems
677 the parser is a per-interpreter option now, so this is safe
678 (this might fix the odd crashing bug).
679 - drop support for 5.9.x versions: they are dead, jim.
680
6814.47 Sun Apr 6 00:37:52 CEST 2008
682 - force cctx allocation on API calls: we know we need to force one
683 and gcc actually manages to confuse our heuristic nowadays,
684 leading to crashes and worse.
685 - document force_cctx.
686
6874.46 Fri Apr 4 22:05:43 CEST 2008
688 - upgrade libcoro, resulting in pthread-backend (which was only created
689 to fulfill the rules of the programming languages shootout).
690
6914.45 Thu Mar 13 11:55:36 CET 2008
692 - fix a file leaking bug in eg/mhyttpd that would allow
693 downloading of any file (reported by oesi).
694 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
695 (also add testcase).
696 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
697 apparently having loads of fun with that).
698 - implement $state->swap_defsv, ->swap_defav and document ->throw.
699
7004.4 Wed Feb 13 16:44:29 CET 2008
701 - only assume .cfi pseudo isns exist on GNU/Linux.
702 - add get_timed method to Coro::Channels.
703 - fixed Coro::Debug manpage.
704 - perl 5.11 compatbility improvement based on patch by
705 Andreas König.
706
7074.37 Sun Jan 20 11:25:23 CET 2008
708 - soften the check for an active parser for perl 5.10 (avoiding
709 "Coro::State::transfer called while parsing" in many cases).
710
7114.36 Sun Jan 13 10:53:56 CET 2008
712 - reset diehook when terminating from an async_pool as to not
713 trigger any __DIE__ handlers.
714
7154.35 Sun Jan 13 04:14:13 CET 2008
716 - "bt" debug command now displays any exceptions
717 from longmess and also skips the innermost
718 stackframes, giving more useufl output.
719 - allow backtraces in coroutines blocked in Coro::EV,
720 at a <1% speed hit.
721 - handle localising of $SIG{__WARN__} and __DIE__
722 properly (with a proper amount of dirty hacking).
723
7244.34 Sat Dec 22 17:49:53 CET 2007
725 - upgrade to EV version 2.0 API.
726
7274.33 Mon Dec 17 08:36:12 CET 2007
728 - make Coro::AIO etc. loadable in the absence of EV.
729
7304.32 Mon Dec 17 07:46:02 CET 2007
731 - majorly improved Coro::Handle's performance with Coro::EV.
732 - implemented the until now mythical Coro::BDB module.
733 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
734
7354.31 Wed Dec 5 12:32:39 CET 2007
736 - remove warn statement form Coro::Util.
737
7384.3 Tue Dec 4 20:33:14 CET 2007
739 - calls to the idle function could cause stack corruption
740 when the stack changed.
741 - do no longer rely on the presence of EV::DNS if EV is used
742 (because it is gone), but instead take avdantage of EV::ADNS
743 if available.
744 - add ($) prototypes to all functions in Coro::Storable.
745 - use a conventional (safer) idle callback in Coro::EV.
746 - do accept quickly in Coro::Debug to avoid endless loops.
747
7484.22 Fri Nov 30 16:04:04 CET 2007
749 - really use optimised versions for Event and EV in Coro::Util
750 and Coro::Handle.
751
7524.21 Sun Nov 25 10:48:59 CET 2007
753 - fix a spurious memory read.
754 - Coro::EV no longer keeps the eventloop "alive".
755
7564.2 Fri Nov 9 20:47:05 CET 2007
757 - enable/disable tracing from a new coroutine, not a pooled one.
758 - fix a memleak in Coro::Event.
759 - removed killall call from fork_eval.
760 - made sure store_fd is already loaded so that fork_eval
761 does not have to parse autoload in each subprocess.
762 - only use assembly method if -O switch is in $Config{optimize}.
763 - add (optional) Coro::EV module, so far the best event loop module
764 directly supported by Coro.
765 - if the event model is EV, use EV::DNS to resolve
766 stuff in Coro::Util.
767 - don't get confused by multiple event notifications in Coro::Handle.
768 - initial support for EV (libevent interface).
769 - require Event and EV using configure_requires, to force their existance.
770
7714.13 Wed Oct 24 07:26:45 CEST 2007
772 - add Coro::Storable::blocking_thaw.
773 - use a vastly more complicated technique to localise
774 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
775 - use a coroutine for the idle callback Coro::Event,
776 instead of running Event in the current coroutine context.
777 This also catches recursive invocations.
778 - actually report fork errors in gethostbyname and inet_aton.
779
7804.11 Thu Oct 11 02:40:24 CEST 2007
781 - port to threaded perls.
782
7834.1 Thu Oct 11 02:38:16 CEST 2007
784 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
785 be local to each coro (see Coro::State).
786 - incompatible change: for very deep reasons, cede and cede_notself
787 cannot return anything, so nothing will be returned.
788 - possibly bring back 5.10 compatibility (untested).
789 - work around stupid (and wrong) warning on 5.10 :(.
790 - overlay the saved state over the context stack. This saves
791 a few hundred bytes per coroutine on average and also
792 speeds up context switching a bit.
793 - further tune default stack sizes.
794 - (more) correctly calculate stack usage in coro_rss.
795 - Coro::Storable::blocking_* did not properly lock
796 resulting in races between coroutines.
797 - added Coro::Storable::guard.
798 - stopping to trace a coroutine could destroy the cctx of
799 an unrelated coroutine.
800 - explain the relationship between Perl and C coroutines in
801 more detail in Coro::State.
802 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
803 causing a fork per resolve. This also affected Coro::Socket.
804 - switch to a separate stack in $coro->call/eval to avoid
805 invalidating pointers.
806
8074.03 Sat Oct 6 21:24:00 CEST 2007
808 - added Coro::throw method.
809 - minor code cleanups.
810
8114.02 Sat Oct 6 02:36:47 CEST 2007
812 - fix a very minor per-coroutine memleak (a single codereference).
813 - fixed a bug where the currently in-use c context would be freed
814 prematurely (can happen only when programs change the stacksize
815 or use tracing).
816 - tracing can no longer keep a coro alive after it terminated.
817 - do static branch prediction in the common path for gcc. gives
818 about 2-5% speed improvement here.
819
8204.01 Fri Oct 5 22:10:49 CEST 2007
821 - instead of recreating *a* standard output handle we simply
822 use STDOUT, which is faster and hopefully more robust.
823
8244.0 Fri Oct 5 12:56:00 CEST 2007
825 - incompatibly changed Coro::Storable::freeze.
826 - major new feature: added Coro::Debug, for interactive coroutine
827 debugging, tracing and much more.
828 - major bug fix: unbelievable, but true: $_, $/ and many other
829 "saved" variables actually weren't being saved. This has been fixed,
830 of course, while increasing performance while losing all the save
831 flags.
832 - save flags are gone, and all the api functions dealing with them.
833 - added Coro::Semaphore::adjust.
834 - added Coro::Util::fork_eval.
835 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
836 - added Coro::killall.
837 - reduce initial stack sizes to allow for "micro-coroutines".
838 - better async_pool resource management, moved parts of async_pool
839 handling to XS (major speed improvement).
840 - actually croak before modifying important data structures.
841 - refuse to transfer while compiling.
842 - possibly support eval EXPR better now.
843 - enable assembly per default on linux+bsd x86+amd64.
844 - all internal members were renamed _something for easier subclassing.
845 - many minor tweaks.
846
8473.63 Wed May 16 14:10:06 CEST 2007
848 - implement handcoded assembly for x86/amd64 SVR ABI.
849
8503.62 Fri Apr 27 21:36:06 CEST 2007
851 - upgrade libcoro (which might set unwind info correctly).
852 - change default on linux to setjmp/longjmp.
853
8543.61 Thu Apr 19 12:36:18 CEST 2007
855 - Coro::Storable caused an endless loop when thawing invalid
856 pst's sometimes.
857 - use a Semaphore in Coro::Storable, as Storable doesn't
858 seem to be reentrant (although it is documented to
859 be threadsafe...).
860 - fix Coro::Signal to bring back the original unreliable
861 but stateful semantics.
862 - fixed a lot of typos in Coro.pm (patch submitted by David
863 Steinbrunner, which applied flawlessly).
864
8653.6 Sat Apr 14 17:13:31 CEST 2007
866 - added some bugfixes to get eg/myhttpd working again.
867 - added Coro::Storable for often-cede'ing freeze/thaw.
868 - try to do a clean exit when a coroutine calls exit
869 (EXPERIMENTAL).
870 - got rid of indirect call through _coro_init.
871 - updated the partly antique examples in eg/ to
872 work again and be a bit less magic, too.
873 - fixed Coro::Signal semantics to work as documented again.
874
8753.55 Sun Mar 25 01:20:47 CET 2007
876 - add SAVE_DEFFH to save the default filehandle and enable
877 it by default.
878 - finally move socket-operations from Coro::Socket to Coro::Handle
879 to be able to unblock foreign sockets.
880 - add Coro::State::save_also and guarded_save.
881 - add count accessor to Coro::Semaphore.
882 - add Coro::State::cctx_stacksize.
883 - just for the fun of it, do not rely on implicit context,
884 which can dramatically improve performance, but people
885 using windows-process-emulation perls do not care much about
886 performance.
887
8883.51 Sun Mar 4 14:18:04 CET 2007
889 - fixed a problem when you weakened references to Coro::State's
890 (patch by Yuval Kogman).
891
8923.501 Wed Feb 28 12:44:07 CET 2007
893 - rename some global symbols as macosx from hell redefines
894 them without asking.
895
8963.5 Tue Feb 13 20:22:53 CET 2007
897 - do AnyEvent model detection earlier, avoiding problems
898 caused by first using AnyEvent and later Coro::Event.
899 - implement and document Coro::Event event objects.
900 - fix a potential problem in Coro::Event causing crashes.
901 - initialise PL_comppad when creating a new coroutine,
902 avoids crashes on early coro destruction.
903
9043.41 Mon Jan 22 19:19:49 CET 2007
8 - readline on Coro::Handle did not support undefined $/, 905 - readline on Coro::Handle did not support undefined $/,
9 nor did it deliver partial lines on EOF or error. 906 nor did it deliver partial lines on EOF or error.
907 - implement malloc fallback for stack allocation because
908 stupid broken idiotic OSX has a stupid broken
909 idiotic fits-the-whole-os mmap "implementation" and
910 my dick feels longer if Coro is portable even to
911 obsolete platforms.
10 912
113.4 Fri Jan 19 21:52:54 CET 2007 9133.4 Fri Jan 19 21:52:54 CET 2007
12 - remove t/09_timer.t, as it isn't really testing much 914 - remove t/09_timer.t, as it isn't really testing much
13 but was rather flaky in practise. 915 but was rather flaky in practise.
14 - async_pool coro would keep arguments and callback alive until 916 - async_pool coro would keep arguments and callback alive until

Diff Legend

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