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

Comparing Coro/Changes (file contents):
Revision 1.173 by root, Mon Dec 4 03:48:16 2006 UTC vs.
Revision 1.408 by root, Fri Nov 14 02:29:09 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: transfer flags must be per state 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: better (earlier) error messages on broken transfer's 4TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
5 7
63. 85.0
9 - ARCHITECTURAL REDESIGN, use 4.912 for the stable version.
10 - directly patch the entersub opcode calling ssl functions (cede,
11 transfer and so on). this does speed up context switching by
12 up to 10%. More importanly, it frees us from the hardcoded
13 behaviour of entersub, so we might actually be able to return
14 something from those functions.
15
164.912 Thu Nov 13 18:31:23 CET 2008
17 - minor cleanups.
18 - use much larger stacks on linux and perl < 5.8.8.
19 - Coro::Debug::new_unix_server did not unlink the socket
20 when destroyed.
21
224.911 Tue Nov 11 04:26:17 CET 2008
23 - "port" to threaded perls.
24
254.91 Mon Nov 10 05:36:38 CET 2008
26 - the ->throw exception object no longer leaks.
27 - creating a new cctx leaked a scopestack entry (memleak).
28 - new coroutines didn't get created with a zero flags field
29 (unknown impact).
30 - calling ->throw on a not-yet-started coroutine should now work
31 instead of being ignored.
32 - ->throw is now supported on Coro::State objects.
33 - clean up cctx creation code a bit.
34 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
35
364.9 Sat Nov 8 17:45:27 CET 2008
37 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
38 - (libcoro) no longer leak threads in the experimental pthread backend,
39 also speed it up considerably.
40 - (libcoro) do not rely on makecontext passing void *'s unscathed.
41 - fix compiletime dependencies on libcoro in the Makefile.
42 - cctx_count wasn't always updated properly.
43 - Coro::State::cctx_stacksize wasn't applied correctly.
44 - new function Coro::State::cctx_max_idle.
45 - the default max number of idle C contexts is now 4.
46 - (libcoro) try harder to get _setjmp/_longjmp.
47 - (libcoro) cleanup and extend the libcoro API to officially
48 allow the creation of empty source contexts.
49 - very experimental workaround for the totally broken netbsd platform.
50 - tried to hack around openbsd bugs.
51
524.804 Wed Nov 5 16:36:00 CET 2008
53 - Coro::Debug::new_unix_server would not create a non-blocking listening
54 socket, sometimes causing freezes.
55 - (libcoro) fix misaligned stack points for setjmp and assembly
56 methods, which can cause crashes on x86/x86_64 with a sufficiently
57 aggressive compiler.
58 - new function: Coro::Debug::new_tcp_server.
59 - move ->throw into the Coro class because it only works on coro objects.
60
614.803 Mon Nov 3 17:16:12 CET 2008
62 - (libcoro) use a global asm statement to become independent of gcc
63 otpimisations for CORO_ASM (thanks to pippijn for the idea).
64 - try to workaround yet another broken bsd, this time dragonfly.
65
664.802 Thu Oct 30 10:56:12 CET 2008
67 - support -fno-omit-frame-pointer on x86 with the assembly method.
68 - tune 01_unblock.t tests a bit.
69
704.801 Wed Oct 22 18:33:37 CEST 2008
71 - improve readline speed for very long "lines".
72 - backport to 5.8.8.
73
744.8 Thu Oct 2 13:34:40 CEST 2008
75 - new function Coro::AIO::aio_wait.
76 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
77 - greatly speed up and reduce memory usage of Coro::AIO requests.
78 - implement some other µ-optimisations.
79
804.749 Mon Sep 29 14:40:12 CEST 2008
81 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
82
834.748 Sat Sep 27 14:03:19 CEST 2008
84 - implement, but do not document, PerlIO::cede(granularity).
85 - Coro::Storable forgot to wrap Storable::pstore.
86 - work around the multitude of leaks and memory corruption
87 bugs in PerlIO::via by using our own C-level perliol. As a side
88 effect, Coro::Storable is now much, much, much faster.
89
904.747 Tue Sep 23 01:59:41 CEST 2008
91 - fix a per-cv memleak (one empty array was leaked per
92 code reference).
93 - avoid a crash in coro->call|rss when the coroutine was already
94 destroyed (most noticably when using Coro::Debug::ps :)
95 - also protect *Storable::FILE.
96 - push up default storable granularity to 20ms.
97
984.746 Sun Sep 21 03:22:20 CEST 2008
99 - be more insistent on locking Storable against reentrancy
100 in Coro::Storable.
101 - move swap_def?v and throw to Coro::State, as documented.
102
1034.745 Thu Jul 24 00:14:38 CEST 2008
104 - remove debugging code related to MgPV_nolen_const, also try to
105 make it compile with perl 5.8.6 (yes, apple apparently loves
106 outdated software). Reported by John S.
107
1084.744 Tue Jul 8 22:06:35 CEST 2008
109 - correctly provide default DIE/WARN handlers as documented.
110 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
111 implementations inside perl work fine for us.
112
1134.743 Mon Jun 16 00:21:57 CEST 2008
114 - when using Coro::EV without running EV::loop it could
115 result in busy-waiting for events, this has been fixed.
116 - reduce codesize and improve performance by using EV_DEFAULT_UC.
117
1184.742 Sat May 31 14:10:21 CEST 2008
119 - implement a workaround for (some) perl <5.8.8 versions.
120 - require EV 3.3+.
121
1224.741 Fri May 30 23:33:09 CEST 2008
123 - tell netbsd how utterly broken their imitation of an OS is
124 and refuse to build by default if pthreads are in use.
125 - switch to "s" method on all bsd's by default, as their ucontext
126 stuff seems just too broken.
127 - fix a bug in Coro::Select.
128
1294.74 Thu May 29 20:05:31 CEST 2008
130 - do not test Coro::LWP for lack of dependencies.
131 - do not test Coro::Select for lack of working perls.
132
1334.73 Thu May 29 2008
134 - fix a bug in Coro::EV which would cause it to block despite
135 there being runnable coroutines.
136 - sprinkle "no warnings" freely over everything, also suppress
137 warnings for some other modules.
138 - fix typo in WSAEWOULDBLOCK.
139
1404.72 Sun May 25 05:14:36 CEST 2008
141 - tweak META.yaml a bit, unfortunately, there is no documented way
142 to have optional dependencies with CPAN. doh :(
143 - avoid running some tests on windows because they would fail due to
144 perl bug (broken fork, broken pipes...).
145 - work around perl on windows bugs where perl returns undocumented
146 error codes for sysread, syswrite etc. by taking advantage
147 of AnyEvent's workaround for that problem.
148 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
149 a common perl implementation bug on windows.
150 - use unix domain sockets in testsuite to work around
151 common perl implementation bugs on widows (they are emulated by
152 tcp sockets on windows. ugh.)
153
1544.71 Sat May 24 20:01:27 CEST 2008
155 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
156 - take advantage of async name resolution of AnyEvent::Util.
157 - work around brutal inet_aton override in Coro::LWP.
158 - take advantage of the readyhook in Coro::EV, for smoother
159 scheduling.
160
1614.7 Sun May 11 00:32:19 CEST 2008
162 - completely reworked the Coro manpage.
163 - added Coro::AnyEvent, generic event loop integration.
164 - implement cancel, ready and kill commands in Coro::Debug.
165 - document find_coro in Coro::Debug.
166 - incompatible change: rename has_stack to has_cctx.
167 - Coro::AIO and Coro::BDB no longer force event model detection,
168 use AnyEvent::AIO and AnyEvent::BDB.
169
1704.6 Sat Apr 26 10:05:14 CEST 2008
171 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
172 by the Coro module. It was a mistake to have it in the first place.
173 - (experimental) support for activestate perl 5.10 (method "w").
174 - (experimental) support for strawberry perl (method "a").
175 - coro_sigelem_set did not return a value although it had to,
176 actual impact unknown.
177
1784.51 Mon Apr 14 13:28:27 CEST 2008
179 - make it compile again on 5.8.
180
1814.50 Thu Apr 10 09:43:17 CEST 2008
182 - I did it twice! (see 4.49).
183
1844.49 Sun Apr 6 21:23:31 CEST 2008
185 - grr, instead of compiling the recent changes on 5.10 only they
186 were compiled on 5.8 only.
187
1884.48 Sun Apr 6 20:36:46 CEST 2008
189 - allow coroutine switches during eval's under 5.10.x, as it seems
190 the parser is a per-interpreter option now, so this is safe
191 (this might fix the odd crashing bug).
192 - drop support for 5.9.x versions: they are dead, jim.
193
1944.47 Sun Apr 6 00:37:52 CEST 2008
195 - force cctx allocation on API calls: we know we need to force one
196 and gcc actually manages to confuse our heuristic nowadays,
197 leading to crashes and worse.
198 - document force_cctx.
199
2004.46 Fri Apr 4 22:05:43 CEST 2008
201 - upgrade libcoro, resulting in pthread-backend (which was only created
202 to fulfill the rules of the programming languages shootout).
203
2044.45 Thu Mar 13 11:55:36 CET 2008
205 - fix a file leaking bug in eg/mhyttpd that would allow
206 downloading of any file (reported by oesi).
207 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
208 (also add testcase).
209 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
210 apparently having loads of fun with that).
211 - implement $state->swap_defsv, ->swap_defav and document ->throw.
212
2134.4 Wed Feb 13 16:44:29 CET 2008
214 - only assume .cfi pseudo isns exist on GNU/Linux.
215 - add get_timed method to Coro::Channels.
216 - fixed Coro::Debug manpage.
217 - perl 5.11 compatbility improvement based on patch by
218 Andreas König.
219
2204.37 Sun Jan 20 11:25:23 CET 2008
221 - soften the check for an active parser for perl 5.10 (avoiding
222 "Coro::State::transfer called while parsing" in many cases).
223
2244.36 Sun Jan 13 10:53:56 CET 2008
225 - reset diehook when terminating from an async_pool as to not
226 trigger any __DIE__ handlers.
227
2284.35 Sun Jan 13 04:14:13 CET 2008
229 - "bt" debug command now displays any exceptions
230 from longmess and also skips the innermost
231 stackframes, giving more useufl output.
232 - allow backtraces in coroutines blocked in Coro::EV,
233 at a <1% speed hit.
234 - handle localising of $SIG{__WARN__} and __DIE__
235 properly (with a proper amount of dirty hacking).
236
2374.34 Sat Dec 22 17:49:53 CET 2007
238 - upgrade to EV version 2.0 API.
239
2404.33 Mon Dec 17 08:36:12 CET 2007
241 - make Coro::AIO etc. loadable in the absence of EV.
242
2434.32 Mon Dec 17 07:46:02 CET 2007
244 - majorly improved Coro::Handle's performance with Coro::EV.
245 - implemented the until now mythical Coro::BDB module.
246 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
247
2484.31 Wed Dec 5 12:32:39 CET 2007
249 - remove warn statement form Coro::Util.
250
2514.3 Tue Dec 4 20:33:14 CET 2007
252 - calls to the idle function could cause stack corruption
253 when the stack changed.
254 - do no longer rely on the presence of EV::DNS if EV is used
255 (because it is gone), but instead take avdantage of EV::ADNS
256 if available.
257 - add ($) prototypes to all functions in Coro::Storable.
258 - use a conventional (safer) idle callback in Coro::EV.
259 - do accept quickly in Coro::Debug to avoid endless loops.
260
2614.22 Fri Nov 30 16:04:04 CET 2007
262 - really use optimised versions for Event and EV in Coro::Util
263 and Coro::Handle.
264
2654.21 Sun Nov 25 10:48:59 CET 2007
266 - fix a spurious memory read.
267 - Coro::EV no longer keeps the eventloop "alive".
268
2694.2 Fri Nov 9 20:47:05 CET 2007
270 - enable/disable tracing from a new coroutine, not a pooled one.
271 - fix a memleak in Coro::Event.
272 - removed killall call from fork_eval.
273 - made sure store_fd is already loaded so that fork_eval
274 does not have to parse autoload in each subprocess.
275 - only use assembly method if -O switch is in $Config{optimize}.
276 - add (optional) Coro::EV module, so far the best event loop module
277 directly supported by Coro.
278 - if the event model is EV, use EV::DNS to resolve
279 stuff in Coro::Util.
280 - don't get confused by multiple event notifications in Coro::Handle.
281 - initial support for EV (libevent interface).
282 - require Event and EV using configure_requires, to force their existance.
283
2844.13 Wed Oct 24 07:26:45 CEST 2007
285 - add Coro::Storable::blocking_thaw.
286 - use a vastly more complicated technique to localise
287 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
288 - use a coroutine for the idle callback Coro::Event,
289 instead of running Event in the current coroutine context.
290 This also catches recursive invocations.
291 - actually report fork errors in gethostbyname and inet_aton.
292
2934.11 Thu Oct 11 02:40:24 CEST 2007
294 - port to threaded perls.
295
2964.1 Thu Oct 11 02:38:16 CEST 2007
297 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
298 be local to each coro (see Coro::State).
299 - incompatible change: for very deep reasons, cede and cede_notself
300 cannot return anything, so nothing will be returned.
301 - possibly bring back 5.10 compatibility (untested).
302 - work around stupid (and wrong) warning on 5.10 :(.
303 - overlay the saved state over the context stack. This saves
304 a few hundred bytes per coroutine on average and also
305 speeds up context switching a bit.
306 - further tune default stack sizes.
307 - (more) correctly calculate stack usage in coro_rss.
308 - Coro::Storable::blocking_* did not properly lock
309 resulting in races between coroutines.
310 - added Coro::Storable::guard.
311 - stopping to trace a coroutine could destroy the cctx of
312 an unrelated coroutine.
313 - explain the relationship between Perl and C coroutines in
314 more detail in Coro::State.
315 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
316 causing a fork per resolve. This also affected Coro::Socket.
317 - switch to a separate stack in $coro->call/eval to avoid
318 invalidating pointers.
319
3204.03 Sat Oct 6 21:24:00 CEST 2007
321 - added Coro::throw method.
322 - minor code cleanups.
323
3244.02 Sat Oct 6 02:36:47 CEST 2007
325 - fix a very minor per-coroutine memleak (a single codereference).
326 - fixed a bug where the currently in-use c context would be freed
327 prematurely (can happen only when programs change the stacksize
328 or use tracing).
329 - tracing can no longer keep a coro alive after it terminated.
330 - do static branch prediction in the common path for gcc. gives
331 about 2-5% speed improvement here.
332
3334.01 Fri Oct 5 22:10:49 CEST 2007
334 - instead of recreating *a* standard output handle we simply
335 use STDOUT, which is faster and hopefully more robust.
336
3374.0 Fri Oct 5 12:56:00 CEST 2007
338 - incompatibly changed Coro::Storable::freeze.
339 - major new feature: added Coro::Debug, for interactive coroutine
340 debugging, tracing and much more.
341 - major bug fix: unbelievable, but true: $_, $/ and many other
342 "saved" variables actually weren't being saved. This has been fixed,
343 of course, while increasing performance while losing all the save
344 flags.
345 - save flags are gone, and all the api functions dealing with them.
346 - added Coro::Semaphore::adjust.
347 - added Coro::Util::fork_eval.
348 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
349 - added Coro::killall.
350 - reduce initial stack sizes to allow for "micro-coroutines".
351 - better async_pool resource management, moved parts of async_pool
352 handling to XS (major speed improvement).
353 - actually croak before modifying important data structures.
354 - refuse to transfer while compiling.
355 - possibly support eval EXPR better now.
356 - enable assembly per default on linux+bsd x86+amd64.
357 - all internal members were renamed _something for easier subclassing.
358 - many minor tweaks.
359
3603.63 Wed May 16 14:10:06 CEST 2007
361 - implement handcoded assembly for x86/amd64 SVR ABI.
362
3633.62 Fri Apr 27 21:36:06 CEST 2007
364 - upgrade libcoro (which might set unwind info correctly).
365 - change default on linux to setjmp/longjmp.
366
3673.61 Thu Apr 19 12:36:18 CEST 2007
368 - Coro::Storable caused an endless loop when thawing invalid
369 pst's sometimes.
370 - use a Semaphore in Coro::Storable, as Storable doesn't
371 seem to be reentrant (although it is documented to
372 be threadsafe...).
373 - fix Coro::Signal to bring back the original unreliable
374 but stateful semantics.
375 - fixed a lot of typos in Coro.pm (patch submitted by David
376 Steinbrunner, which applied flawlessly).
377
3783.6 Sat Apr 14 17:13:31 CEST 2007
379 - added some bugfixes to get eg/myhttpd working again.
380 - added Coro::Storable for often-cede'ing freeze/thaw.
381 - try to do a clean exit when a coroutine calls exit
382 (EXPERIMENTAL).
383 - got rid of indirect call through _coro_init.
384 - updated the partly antique examples in eg/ to
385 work again and be a bit less magic, too.
386 - fixed Coro::Signal semantics to work as documented again.
387
3883.55 Sun Mar 25 01:20:47 CET 2007
389 - add SAVE_DEFFH to save the default filehandle and enable
390 it by default.
391 - finally move socket-operations from Coro::Socket to Coro::Handle
392 to be able to unblock foreign sockets.
393 - add Coro::State::save_also and guarded_save.
394 - add count accessor to Coro::Semaphore.
395 - add Coro::State::cctx_stacksize.
396 - just for the fun of it, do not rely on implicit context,
397 which can dramatically improve performance, but people
398 using windows-process-emulation perls do not care much about
399 performance.
400
4013.51 Sun Mar 4 14:18:04 CET 2007
402 - fixed a problem when you weakened references to Coro::State's
403 (patch by Yuval Kogman).
404
4053.501 Wed Feb 28 12:44:07 CET 2007
406 - rename some global symbols as macosx from hell redefines
407 them without asking.
408
4093.5 Tue Feb 13 20:22:53 CET 2007
410 - do AnyEvent model detection earlier, avoiding problems
411 caused by first using AnyEvent and later Coro::Event.
412 - implement and document Coro::Event event objects.
413 - fix a potential problem in Coro::Event causing crashes.
414 - initialise PL_comppad when creating a new coroutine,
415 avoids crashes on early coro destruction.
416
4173.41 Mon Jan 22 19:19:49 CET 2007
418 - readline on Coro::Handle did not support undefined $/,
419 nor did it deliver partial lines on EOF or error.
420 - implement malloc fallback for stack allocation because
421 stupid broken idiotic OSX has a stupid broken
422 idiotic fits-the-whole-os mmap "implementation" and
423 my dick feels longer if Coro is portable even to
424 obsolete platforms.
425
4263.4 Fri Jan 19 21:52:54 CET 2007
427 - remove t/09_timer.t, as it isn't really testing much
428 but was rather flaky in practise.
429 - async_pool coro would keep arguments and callback alive until
430 it was reused.
431 - cancellation of a coroutine could cause spurious idle calls
432 in cede_notself.
433
4343.3 Sat Jan 6 03:45:00 CET 2007
435 - implement $coro->on_destroy.
436 - Coro::Event blocking semantics have been changed,
437 documented and - hopefully - improved.
438 - fix nice adding, not subtracting, from priority.
439 - fix ->prio and api_is_ready (patch by Mark Hinds).
440 - fixed an assert ("... == prev__cctx->idle_te")
441 that could errronously trigger.
442 - fix various large and small memleaks.
443 - use a (hopefully) more stable cancel implementation
444 that immediately frees the coroutine data.
445 - cede/cede_notself return a status now.
446 - added Coro::guard function.
447 - added a global coroutine pool for jobs (on my machine,
448 I can create and execute 48k simple coros/s with async,
449 and 128k coros with async_pool).
450 - Coro::AIO now uses the coroutine priority as io priority.
451
4523.2 Fri Dec 22 05:07:09 CET 2006
453 - improve portability to slightly older perls.
454 - use cleaner coroutine destruction.
455 - simplify configuration for users.
456 - optionally (unrecommended) prefer perl functions over
457 their coro replacements.
458
4593.11 Tue Dec 5 13:11:24 CET 2006
460 - fixed some bogus assert's, but as perl.h disables assert even
461 without NDEBUG (thank you very much), not too many people should
462 notice (that did include myself). Andreas König noticed, though :)
463 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
464 it. Also noticed by Andreas König :)
465 - save/restore tainted status.
466 - verified to pass the testsuite on my 5.9.5.
467
4683.1 Mon Dec 4 23:03:40 CET 2006
469 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
470 - incompatible change: transfer flags are now per-state.
7 - give a better error message on deadlock. 471 - give a better error message on deadlock.
472 - document Coro::nready.
473 - enhanced testsuite.
8 474
93.01 Sun Dec 3 23:47:42 CET 2006 4753.01 Sun Dec 3 23:47:42 CET 2006
10 - forgot to include Coro::Timer. 476 - forgot to include Coro::Timer.
11 477
123.0 Sun Dec 3 22:57:25 CET 2006 4783.0 Sun Dec 3 22:57:25 CET 2006

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines