ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.511
Committed: Sat Jul 11 14:24:12 2009 UTC (14 years, 10 months ago) by root
Branch: MAIN
Changes since 1.510: +1 -0 lines
Log Message:
license change

File Contents

# Content
1 Revision history for Perl extension Coro.
2
3 TODO: should explore PerlIO::coroaio (perl leaks like hell).
4 TODO: 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)).
7 TODO: unready_all
8 TODO: myhttpd header parsing
9 TODO: channel->maxsize(newsize)?
10
11 - considerably speed up Coro::Select by taking avdantage
12 of AnyEvent > 4.8 and some other optimisations.
13 - clarified libcoro license and copyright.
14 - someone stole my EXTRA_META!!!
15
16 5.151 Mon Jul 6 05:41:57 CEST 2009
17 - backport to windows process emulation code again (patch by
18 Yasuhiro MATSUMOTO).
19 - slightly update Coro::MakeMaker.
20
21 5.15 Tue Jun 30 10:28:06 CEST 2009
22 - deprecate Coro::Socket, document how to get ipv6 support via
23 AnyEvent::Socket instead.
24 - implement signal->wait ($cb) interface, similar to semaphores.
25 - work around SvOK not supporting getmagic, so we have to getmagic
26 to test for undef :/ (reported by Matthias Waldorf).
27 - load Coro::AnyEvent in all modules using AnyEvent.
28 - work around perl corrupting our internal data structures,
29 reported by Tokuhiro Matsuno.
30 - enable per-coroutine real and cpu time gathering
31 (Coro::State::enable_times).
32
33 5.14 Wed Jun 24 01:37:48 CEST 2009
34 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
35 as this seems to have been a difficult concept (poll, sleep, idle,
36 idle_upto).
37 - add Coro::AnyEvent::readable/writable functions.
38 - clarify Coro::EV/Event/AnyEvent manpages.
39 - free per-thread global scalars in the thread calling ->cancel, to
40 avoid crashes when $_, $@ etc., are magical but some of those
41 had already been freed.
42 - "unexperimentalise" the callback interface for Coro::Semaphore.
43 - speed up ready queue management/context switching by using a linked
44 list instead of an array (~5-10%).
45 - implement "watch" command in Coro::Debug shells.
46 - for fun, implement time-slicing as an example in the manpage.
47 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
48 Coro::Event, use the normal AnyEvent handling instead of dieing
49 (the same is true for Coro::Handle).
50 - properly document Coro::EV::timed_io_once.
51 - avoid unneccessary ->cancel calls in Coro::Handle.
52 - maybe make it work on mingw32 with win32 backend
53 (based on patch by Yasuhiro Matsumoto).
54
55 5.132 Fri May 29 09:00:39 CEST 2009
56 - do not keep a reference to the argument itself in
57 Coro::Semaphore::guard, as it could change later.
58 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
59 should support a prepare callback.
60
61 5.131 Mon Mar 16 23:20:37 CET 2009
62 - implement and document Coro->suspend, Coro->resume.
63 - fix Coro::Select implementation to not (often) close
64 the passed file descriptors (testcase provided by pippijn).
65
66 5.13 Mon Dec 15 21:51:42 CET 2008
67 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
68 - don't set diehook to C<undef> but instead to NULL, to avoid
69 spurious warnings.
70 - fix a lot of bugs in Coro::SemaphoreSet.
71 - Coro::SemaphoreSet will less often create a semaphore needlessly.
72 - add Coro::SemaphoreSet::count and wait methods.
73 - take advantage of the new Guard module.
74 - deprecate Coro::guard.
75 - try to fix the dreaded 01_unblock tests once more. I hate it when
76 testsuites need more fixing than the code they are supposed to test.
77 - croak in more cases when a required callback isn't resolvable.
78 - fix some minor issues in Coro::State->call/eval.
79 - use current coroutine context instead of a temporary one
80 when temporarily switching to another coroutine.
81 - do not call C-level on_destroy handlers during global destruction,
82 to avoid needless segfaults.
83
84 5.12 Sun Dec 7 13:30:38 CET 2008
85 - add default config for MirOS, which seems to be a bug-to-bug
86 compatible fork of openbsd ("world domination by releasing
87 openbsd cvs before the openbsd folks do it" or so :).
88 - free_padlist did destroy the names pad, not good, but didn't
89 seem to bother perl - this could fix issues such as eval ""
90 inside a function called from multiple coroutines.
91 - use a different method to detect destruction time.
92 - be more careful when freeing padlists just before global
93 destruction.
94 - fixed and expanded the call/cc example.
95 - renamed _terminate to _coro_run.
96 - new method Coro::Channel->shutdown.
97 - try pthreads on openbsd <4.4 (broken sigaltstack, will
98 pthreads fare better?).
99 - be less picky about destroying "a" running coroutine.
100
101 5.11 Tue Nov 25 21:49:05 CET 2008
102 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
103 working code with failed assertions, introducing more bugs than
104 it fixes, requiring elaborate workarounds :(
105
106 5.1 Mon Nov 24 08:54:59 CET 2008
107 - wrote a small introductory tutorial - Coro::Intro.
108 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
109 - Coro::Select did errornously dup the file descriptors
110 and didn't work with all AnyEvent backends.
111 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
112 LWP data transfers.
113 - disassociate c contexts from coro objects - this is agruably more
114 correct, but mostly allows sharing of cctxs between coro and state
115 objects, for added memory savings and speed increases.
116 - bumped $Coro::POOL_RSS up to 32kb by default.
117 - no longer set the optype to OP_CUSTOM, as B::* understandably
118 doesn't like this very much (and we *are* a type of entersub).
119 - implement state cloning, just to prove that call/cc can be done.
120 - automatically load Coro::AnyEvent in Coro::Handle.
121 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
122 do not have this method (and don't need it either).
123 - speed up generic anyevent methods in Coro::Handle by using rouse
124 callbacks.
125 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
126 others. It also makes the debugger happier, as you can trace
127 through the idle threads now.
128 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
129 - eg/event was pretty broken.
130 - better 5.8.6 compatibility.
131
132 5.0 Thu Nov 20 10:35:05 CET 2008
133 - NEW ARCHITECTURE: use the latest 4.x version if you experience
134 stability issues.
135 - bump API version to 7 - all dependents must be recompiled.
136 - removed timed_* functions - they were not being used anyways
137 and should be replaced by a more generic mechanism -
138 and were annoying to support anyways :)
139 - removed SemaphoreSet's waiter method - use sem method instead.
140 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
141 - async_pool did free a scalar value twice
142 ("Attempt to unreference...").
143 - fix a longstanding bug where calling terminate on a coro that
144 was waiting for a semaphore that was just becoming available
145 would cause a deadlock (semaphore would get into a state where
146 it was available but waiters were still blocked).
147 - calling throw on a coroutine that is waiting for a semaphore will
148 no longer make it acquire the semaphore (and thus leak a count).
149 - perl's process emulation is now not even theoretically supported
150 anymore.
151 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
152 conversion of callback-style to blocking-style.
153 - new methods $coro->schedule_to and ->cede_to, to specifically
154 schedule or cede to a specific coroutine.
155 - new function Coro::Semaphore::wait.
156 - use named constants in Coro::Channel (Richard Hundt).
157 - directly patch the entersub opcode calling SLF functions (cede,
158 transfer and so on). this does speed up context switching, but
159 more importanly, it frees us from the hardcoded behaviour of
160 entersub, so we might actually be able to return something from
161 those functions and atcually create new ones.
162 - take advantage of __builtin_frame_address on gcc.
163 - expose THX in coroapi (not sure whether this was a wise decision,
164 as "threaded" perls are running at half speed anyways).
165 - implement execute_slf (schedule-like-function) interface that makes
166 it possible to implement schedule-like-functions in XS.
167 - use new SLF interface to massively speed up Coro::EV by roughly a
168 factor of two.
169 - used new SLF interface to massively speed up Coro::Semaphore by a
170 factor of three.
171 - used new SLF interface to speed up Coro::AIO by roughly a factor of
172 four and reduce its memory usage considerably.
173 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
174 for a nice speedup and vastly more correct behaviour. Also implement
175 a new method "sem" to get at the underlying semaphore object.
176 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
177 (in comparison) 20-40% speedup.
178 - used new SLF interface to reimplement Coro::Signal gaining
179 some unknown (because I was too lazy), but certain, speedup, and also
180 making signals reliable for the first time.
181 - used new SLF interface and other optimisations to speed up async_pool
182 by a factor of two. It also doesn't rely on perl's exception mechanism
183 to exit anymore. The overhead for terminating an async_pool, coro over
184 a normal async is now very small.
185 - sped up coroutine creation/destruction by 40%.
186 - forgot to include Coro/libcoro/README in the dist for all these years.
187 - work around a freebsd pthreads bug (manual testcancel is required as
188 pthread_cond_wait isn't a cancellation point on freebsd).
189 - use new rouse functions to speed up and simplify Coro::BDB.
190 - make "prefer perl native functions" work with threaded perls.
191 - condense Coro::Debug ps output, hint at v and w flags.
192 - (libcoro) lots of minor cleanups and portability improvements.
193
194 4.914 Wed Nov 19 12:54:18 CET 2008
195 - fix a disastrous bug in the readline optimisation
196 introduced in 4.801.
197
198 4.913 Sat Nov 15 07:58:28 CET 2008
199 - async_pool did free a scalar value twice
200 ("Attempt to unreference...").
201
202 4.912 Thu Nov 13 18:31:23 CET 2008
203 - minor cleanups.
204 - use much larger stacks on linux and perl < 5.8.8.
205 - Coro::Debug::new_unix_server did not unlink the socket
206 when destroyed.
207
208 4.911 Tue Nov 11 04:26:17 CET 2008
209 - "port" to threaded perls.
210
211 4.91 Mon Nov 10 05:36:38 CET 2008
212 - the ->throw exception object no longer leaks.
213 - creating a new cctx leaked a scopestack entry (memleak).
214 - new coroutines didn't get created with a zero flags field
215 (unknown impact).
216 - calling ->throw on a not-yet-started coroutine should now work
217 instead of being ignored.
218 - ->throw is now supported on Coro::State objects.
219 - clean up cctx creation code a bit.
220 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
221
222 4.9 Sat Nov 8 17:45:27 CET 2008
223 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
224 - (libcoro) no longer leak threads in the experimental pthread backend,
225 also speed it up considerably.
226 - (libcoro) do not rely on makecontext passing void *'s unscathed.
227 - fix compiletime dependencies on libcoro in the Makefile.
228 - cctx_count wasn't always updated properly.
229 - Coro::State::cctx_stacksize wasn't applied correctly.
230 - new function Coro::State::cctx_max_idle.
231 - the default max number of idle C contexts is now 4.
232 - (libcoro) try harder to get _setjmp/_longjmp.
233 - (libcoro) cleanup and extend the libcoro API to officially
234 allow the creation of empty source contexts.
235 - very experimental workaround for the totally broken netbsd platform.
236 - tried to hack around openbsd bugs.
237
238 4.804 Wed Nov 5 16:36:00 CET 2008
239 - Coro::Debug::new_unix_server would not create a non-blocking listening
240 socket, sometimes causing freezes.
241 - (libcoro) fix misaligned stack points for setjmp and assembly
242 methods, which can cause crashes on x86/x86_64 with a sufficiently
243 aggressive compiler.
244 - new function: Coro::Debug::new_tcp_server.
245 - move ->throw into the Coro class because it only works on coro objects.
246
247 4.803 Mon Nov 3 17:16:12 CET 2008
248 - (libcoro) use a global asm statement to become independent of gcc
249 otpimisations for CORO_ASM (thanks to pippijn for the idea).
250 - try to workaround yet another broken bsd, this time dragonfly.
251
252 4.802 Thu Oct 30 10:56:12 CET 2008
253 - support -fno-omit-frame-pointer on x86 with the assembly method.
254 - tune 01_unblock.t tests a bit.
255
256 4.801 Wed Oct 22 18:33:37 CEST 2008
257 - improve readline speed for very long "lines".
258 - backport to 5.8.8.
259
260 4.8 Thu Oct 2 13:34:40 CEST 2008
261 - new function Coro::AIO::aio_wait.
262 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
263 - greatly speed up and reduce memory usage of Coro::AIO requests.
264 - implement some other ยต-optimisations.
265
266 4.749 Mon Sep 29 14:40:12 CEST 2008
267 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
268
269 4.748 Sat Sep 27 14:03:19 CEST 2008
270 - implement, but do not document, PerlIO::cede(granularity).
271 - Coro::Storable forgot to wrap Storable::pstore.
272 - work around the multitude of leaks and memory corruption
273 bugs in PerlIO::via by using our own C-level perliol. As a side
274 effect, Coro::Storable is now much, much, much faster.
275
276 4.747 Tue Sep 23 01:59:41 CEST 2008
277 - fix a per-cv memleak (one empty array was leaked per
278 code reference).
279 - avoid a crash in coro->call|rss when the coroutine was already
280 destroyed (most noticably when using Coro::Debug::ps :)
281 - also protect *Storable::FILE.
282 - push up default storable granularity to 20ms.
283
284 4.746 Sun Sep 21 03:22:20 CEST 2008
285 - be more insistent on locking Storable against reentrancy
286 in Coro::Storable.
287 - move swap_def?v and throw to Coro::State, as documented.
288
289 4.745 Thu Jul 24 00:14:38 CEST 2008
290 - remove debugging code related to MgPV_nolen_const, also try to
291 make it compile with perl 5.8.6 (yes, apple apparently loves
292 outdated software). Reported by John S.
293
294 4.744 Tue Jul 8 22:06:35 CEST 2008
295 - correctly provide default DIE/WARN handlers as documented.
296 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
297 implementations inside perl work fine for us.
298
299 4.743 Mon Jun 16 00:21:57 CEST 2008
300 - when using Coro::EV without running EV::loop it could
301 result in busy-waiting for events, this has been fixed.
302 - reduce codesize and improve performance by using EV_DEFAULT_UC.
303
304 4.742 Sat May 31 14:10:21 CEST 2008
305 - implement a workaround for (some) perl <5.8.8 versions.
306 - require EV 3.3+.
307
308 4.741 Fri May 30 23:33:09 CEST 2008
309 - tell netbsd how utterly broken their imitation of an OS is
310 and refuse to build by default if pthreads are in use.
311 - switch to "s" method on all bsd's by default, as their ucontext
312 stuff seems just too broken.
313 - fix a bug in Coro::Select.
314
315 4.74 Thu May 29 20:05:31 CEST 2008
316 - do not test Coro::LWP for lack of dependencies.
317 - do not test Coro::Select for lack of working perls.
318
319 4.73 Thu May 29 2008
320 - fix a bug in Coro::EV which would cause it to block despite
321 there being runnable coroutines.
322 - sprinkle "no warnings" freely over everything, also suppress
323 warnings for some other modules.
324 - fix typo in WSAEWOULDBLOCK.
325
326 4.72 Sun May 25 05:14:36 CEST 2008
327 - tweak META.yaml a bit, unfortunately, there is no documented way
328 to have optional dependencies with CPAN. doh :(
329 - avoid running some tests on windows because they would fail due to
330 perl bug (broken fork, broken pipes...).
331 - work around perl on windows bugs where perl returns undocumented
332 error codes for sysread, syswrite etc. by taking advantage
333 of AnyEvent's workaround for that problem.
334 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
335 a common perl implementation bug on windows.
336 - use unix domain sockets in testsuite to work around
337 common perl implementation bugs on widows (they are emulated by
338 tcp sockets on windows. ugh.)
339
340 4.71 Sat May 24 20:01:27 CEST 2008
341 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
342 - take advantage of async name resolution of AnyEvent::Util.
343 - work around brutal inet_aton override in Coro::LWP.
344 - take advantage of the readyhook in Coro::EV, for smoother
345 scheduling.
346
347 4.7 Sun May 11 00:32:19 CEST 2008
348 - completely reworked the Coro manpage.
349 - added Coro::AnyEvent, generic event loop integration.
350 - implement cancel, ready and kill commands in Coro::Debug.
351 - document find_coro in Coro::Debug.
352 - incompatible change: rename has_stack to has_cctx.
353 - Coro::AIO and Coro::BDB no longer force event model detection,
354 use AnyEvent::AIO and AnyEvent::BDB.
355
356 4.6 Sat Apr 26 10:05:14 CEST 2008
357 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
358 by the Coro module. It was a mistake to have it in the first place.
359 - (experimental) support for activestate perl 5.10 (method "w").
360 - (experimental) support for strawberry perl (method "a").
361 - coro_sigelem_set did not return a value although it had to,
362 actual impact unknown.
363
364 4.51 Mon Apr 14 13:28:27 CEST 2008
365 - make it compile again on 5.8.
366
367 4.50 Thu Apr 10 09:43:17 CEST 2008
368 - I did it twice! (see 4.49).
369
370 4.49 Sun Apr 6 21:23:31 CEST 2008
371 - grr, instead of compiling the recent changes on 5.10 only they
372 were compiled on 5.8 only.
373
374 4.48 Sun Apr 6 20:36:46 CEST 2008
375 - allow coroutine switches during eval's under 5.10.x, as it seems
376 the parser is a per-interpreter option now, so this is safe
377 (this might fix the odd crashing bug).
378 - drop support for 5.9.x versions: they are dead, jim.
379
380 4.47 Sun Apr 6 00:37:52 CEST 2008
381 - force cctx allocation on API calls: we know we need to force one
382 and gcc actually manages to confuse our heuristic nowadays,
383 leading to crashes and worse.
384 - document force_cctx.
385
386 4.46 Fri Apr 4 22:05:43 CEST 2008
387 - upgrade libcoro, resulting in pthread-backend (which was only created
388 to fulfill the rules of the programming languages shootout).
389
390 4.45 Thu Mar 13 11:55:36 CET 2008
391 - fix a file leaking bug in eg/mhyttpd that would allow
392 downloading of any file (reported by oesi).
393 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
394 (also add testcase).
395 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
396 apparently having loads of fun with that).
397 - implement $state->swap_defsv, ->swap_defav and document ->throw.
398
399 4.4 Wed Feb 13 16:44:29 CET 2008
400 - only assume .cfi pseudo isns exist on GNU/Linux.
401 - add get_timed method to Coro::Channels.
402 - fixed Coro::Debug manpage.
403 - perl 5.11 compatbility improvement based on patch by
404 Andreas Kรถnig.
405
406 4.37 Sun Jan 20 11:25:23 CET 2008
407 - soften the check for an active parser for perl 5.10 (avoiding
408 "Coro::State::transfer called while parsing" in many cases).
409
410 4.36 Sun Jan 13 10:53:56 CET 2008
411 - reset diehook when terminating from an async_pool as to not
412 trigger any __DIE__ handlers.
413
414 4.35 Sun Jan 13 04:14:13 CET 2008
415 - "bt" debug command now displays any exceptions
416 from longmess and also skips the innermost
417 stackframes, giving more useufl output.
418 - allow backtraces in coroutines blocked in Coro::EV,
419 at a <1% speed hit.
420 - handle localising of $SIG{__WARN__} and __DIE__
421 properly (with a proper amount of dirty hacking).
422
423 4.34 Sat Dec 22 17:49:53 CET 2007
424 - upgrade to EV version 2.0 API.
425
426 4.33 Mon Dec 17 08:36:12 CET 2007
427 - make Coro::AIO etc. loadable in the absence of EV.
428
429 4.32 Mon Dec 17 07:46:02 CET 2007
430 - majorly improved Coro::Handle's performance with Coro::EV.
431 - implemented the until now mythical Coro::BDB module.
432 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
433
434 4.31 Wed Dec 5 12:32:39 CET 2007
435 - remove warn statement form Coro::Util.
436
437 4.3 Tue Dec 4 20:33:14 CET 2007
438 - calls to the idle function could cause stack corruption
439 when the stack changed.
440 - do no longer rely on the presence of EV::DNS if EV is used
441 (because it is gone), but instead take avdantage of EV::ADNS
442 if available.
443 - add ($) prototypes to all functions in Coro::Storable.
444 - use a conventional (safer) idle callback in Coro::EV.
445 - do accept quickly in Coro::Debug to avoid endless loops.
446
447 4.22 Fri Nov 30 16:04:04 CET 2007
448 - really use optimised versions for Event and EV in Coro::Util
449 and Coro::Handle.
450
451 4.21 Sun Nov 25 10:48:59 CET 2007
452 - fix a spurious memory read.
453 - Coro::EV no longer keeps the eventloop "alive".
454
455 4.2 Fri Nov 9 20:47:05 CET 2007
456 - enable/disable tracing from a new coroutine, not a pooled one.
457 - fix a memleak in Coro::Event.
458 - removed killall call from fork_eval.
459 - made sure store_fd is already loaded so that fork_eval
460 does not have to parse autoload in each subprocess.
461 - only use assembly method if -O switch is in $Config{optimize}.
462 - add (optional) Coro::EV module, so far the best event loop module
463 directly supported by Coro.
464 - if the event model is EV, use EV::DNS to resolve
465 stuff in Coro::Util.
466 - don't get confused by multiple event notifications in Coro::Handle.
467 - initial support for EV (libevent interface).
468 - require Event and EV using configure_requires, to force their existance.
469
470 4.13 Wed Oct 24 07:26:45 CEST 2007
471 - add Coro::Storable::blocking_thaw.
472 - use a vastly more complicated technique to localise
473 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
474 - use a coroutine for the idle callback Coro::Event,
475 instead of running Event in the current coroutine context.
476 This also catches recursive invocations.
477 - actually report fork errors in gethostbyname and inet_aton.
478
479 4.11 Thu Oct 11 02:40:24 CEST 2007
480 - port to threaded perls.
481
482 4.1 Thu Oct 11 02:38:16 CEST 2007
483 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
484 be local to each coro (see Coro::State).
485 - incompatible change: for very deep reasons, cede and cede_notself
486 cannot return anything, so nothing will be returned.
487 - possibly bring back 5.10 compatibility (untested).
488 - work around stupid (and wrong) warning on 5.10 :(.
489 - overlay the saved state over the context stack. This saves
490 a few hundred bytes per coroutine on average and also
491 speeds up context switching a bit.
492 - further tune default stack sizes.
493 - (more) correctly calculate stack usage in coro_rss.
494 - Coro::Storable::blocking_* did not properly lock
495 resulting in races between coroutines.
496 - added Coro::Storable::guard.
497 - stopping to trace a coroutine could destroy the cctx of
498 an unrelated coroutine.
499 - explain the relationship between Perl and C coroutines in
500 more detail in Coro::State.
501 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
502 causing a fork per resolve. This also affected Coro::Socket.
503 - switch to a separate stack in $coro->call/eval to avoid
504 invalidating pointers.
505
506 4.03 Sat Oct 6 21:24:00 CEST 2007
507 - added Coro::throw method.
508 - minor code cleanups.
509
510 4.02 Sat Oct 6 02:36:47 CEST 2007
511 - fix a very minor per-coroutine memleak (a single codereference).
512 - fixed a bug where the currently in-use c context would be freed
513 prematurely (can happen only when programs change the stacksize
514 or use tracing).
515 - tracing can no longer keep a coro alive after it terminated.
516 - do static branch prediction in the common path for gcc. gives
517 about 2-5% speed improvement here.
518
519 4.01 Fri Oct 5 22:10:49 CEST 2007
520 - instead of recreating *a* standard output handle we simply
521 use STDOUT, which is faster and hopefully more robust.
522
523 4.0 Fri Oct 5 12:56:00 CEST 2007
524 - incompatibly changed Coro::Storable::freeze.
525 - major new feature: added Coro::Debug, for interactive coroutine
526 debugging, tracing and much more.
527 - major bug fix: unbelievable, but true: $_, $/ and many other
528 "saved" variables actually weren't being saved. This has been fixed,
529 of course, while increasing performance while losing all the save
530 flags.
531 - save flags are gone, and all the api functions dealing with them.
532 - added Coro::Semaphore::adjust.
533 - added Coro::Util::fork_eval.
534 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
535 - added Coro::killall.
536 - reduce initial stack sizes to allow for "micro-coroutines".
537 - better async_pool resource management, moved parts of async_pool
538 handling to XS (major speed improvement).
539 - actually croak before modifying important data structures.
540 - refuse to transfer while compiling.
541 - possibly support eval EXPR better now.
542 - enable assembly per default on linux+bsd x86+amd64.
543 - all internal members were renamed _something for easier subclassing.
544 - many minor tweaks.
545
546 3.63 Wed May 16 14:10:06 CEST 2007
547 - implement handcoded assembly for x86/amd64 SVR ABI.
548
549 3.62 Fri Apr 27 21:36:06 CEST 2007
550 - upgrade libcoro (which might set unwind info correctly).
551 - change default on linux to setjmp/longjmp.
552
553 3.61 Thu Apr 19 12:36:18 CEST 2007
554 - Coro::Storable caused an endless loop when thawing invalid
555 pst's sometimes.
556 - use a Semaphore in Coro::Storable, as Storable doesn't
557 seem to be reentrant (although it is documented to
558 be threadsafe...).
559 - fix Coro::Signal to bring back the original unreliable
560 but stateful semantics.
561 - fixed a lot of typos in Coro.pm (patch submitted by David
562 Steinbrunner, which applied flawlessly).
563
564 3.6 Sat Apr 14 17:13:31 CEST 2007
565 - added some bugfixes to get eg/myhttpd working again.
566 - added Coro::Storable for often-cede'ing freeze/thaw.
567 - try to do a clean exit when a coroutine calls exit
568 (EXPERIMENTAL).
569 - got rid of indirect call through _coro_init.
570 - updated the partly antique examples in eg/ to
571 work again and be a bit less magic, too.
572 - fixed Coro::Signal semantics to work as documented again.
573
574 3.55 Sun Mar 25 01:20:47 CET 2007
575 - add SAVE_DEFFH to save the default filehandle and enable
576 it by default.
577 - finally move socket-operations from Coro::Socket to Coro::Handle
578 to be able to unblock foreign sockets.
579 - add Coro::State::save_also and guarded_save.
580 - add count accessor to Coro::Semaphore.
581 - add Coro::State::cctx_stacksize.
582 - just for the fun of it, do not rely on implicit context,
583 which can dramatically improve performance, but people
584 using windows-process-emulation perls do not care much about
585 performance.
586
587 3.51 Sun Mar 4 14:18:04 CET 2007
588 - fixed a problem when you weakened references to Coro::State's
589 (patch by Yuval Kogman).
590
591 3.501 Wed Feb 28 12:44:07 CET 2007
592 - rename some global symbols as macosx from hell redefines
593 them without asking.
594
595 3.5 Tue Feb 13 20:22:53 CET 2007
596 - do AnyEvent model detection earlier, avoiding problems
597 caused by first using AnyEvent and later Coro::Event.
598 - implement and document Coro::Event event objects.
599 - fix a potential problem in Coro::Event causing crashes.
600 - initialise PL_comppad when creating a new coroutine,
601 avoids crashes on early coro destruction.
602
603 3.41 Mon Jan 22 19:19:49 CET 2007
604 - readline on Coro::Handle did not support undefined $/,
605 nor did it deliver partial lines on EOF or error.
606 - implement malloc fallback for stack allocation because
607 stupid broken idiotic OSX has a stupid broken
608 idiotic fits-the-whole-os mmap "implementation" and
609 my dick feels longer if Coro is portable even to
610 obsolete platforms.
611
612 3.4 Fri Jan 19 21:52:54 CET 2007
613 - remove t/09_timer.t, as it isn't really testing much
614 but was rather flaky in practise.
615 - async_pool coro would keep arguments and callback alive until
616 it was reused.
617 - cancellation of a coroutine could cause spurious idle calls
618 in cede_notself.
619
620 3.3 Sat Jan 6 03:45:00 CET 2007
621 - implement $coro->on_destroy.
622 - Coro::Event blocking semantics have been changed,
623 documented and - hopefully - improved.
624 - fix nice adding, not subtracting, from priority.
625 - fix ->prio and api_is_ready (patch by Mark Hinds).
626 - fixed an assert ("... == prev__cctx->idle_te")
627 that could errronously trigger.
628 - fix various large and small memleaks.
629 - use a (hopefully) more stable cancel implementation
630 that immediately frees the coroutine data.
631 - cede/cede_notself return a status now.
632 - added Coro::guard function.
633 - added a global coroutine pool for jobs (on my machine,
634 I can create and execute 48k simple coros/s with async,
635 and 128k coros with async_pool).
636 - Coro::AIO now uses the coroutine priority as io priority.
637
638 3.2 Fri Dec 22 05:07:09 CET 2006
639 - improve portability to slightly older perls.
640 - use cleaner coroutine destruction.
641 - simplify configuration for users.
642 - optionally (unrecommended) prefer perl functions over
643 their coro replacements.
644
645 3.11 Tue Dec 5 13:11:24 CET 2006
646 - fixed some bogus assert's, but as perl.h disables assert even
647 without NDEBUG (thank you very much), not too many people should
648 notice (that did include myself). Andreas Kรถnig noticed, though :)
649 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
650 it. Also noticed by Andreas Kรถnig :)
651 - save/restore tainted status.
652 - verified to pass the testsuite on my 5.9.5.
653
654 3.1 Mon Dec 4 23:03:40 CET 2006
655 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
656 - incompatible change: transfer flags are now per-state.
657 - give a better error message on deadlock.
658 - document Coro::nready.
659 - enhanced testsuite.
660
661 3.01 Sun Dec 3 23:47:42 CET 2006
662 - forgot to include Coro::Timer.
663
664 3.0 Sun Dec 3 22:57:25 CET 2006
665 - the "FINALLY COMPLETELY STABLE" release (coming soon:
666 the "FAMOUS LAST WORDS" release).
667 - implement a new stack sharing algorithm, which uses a stack
668 pool (size currently hardcoded).
669 - make stack sharing mandatory (it no longer uses a heuristic).
670 - eval/die no longer cause weird problems under heavy use.
671 - Coro::Event could cause livelocks if it was used but
672 no Coro::Event watchers were used.
673 - Coro::Event now uses asynccheck as prepare does not
674 check for changed watchers.
675 - Coro::Event allows multiple waiting coros and will wake up one
676 per event.
677 - Coro::Event should be cleaner and more efficient now.
678 - new utility function Coro::unblock_sub.
679 - document the sad fact that Event is no longer reentrant.
680 - putting a coroutine into the ready queue twice could under
681 some circumstances lead to stack corruption.
682 - minor incompatibility: subclassing Coro::State is supported
683 directly now without going through a _coro_state member.
684 - state/coro switching is much faster now.
685 - very minor optimisations and code/documentation cleanup.
686 - avoid problems due to compiler inlining.
687 - removed timers from Coro::Timer -> use AnyEvent instead.
688 - replaced Coro::idle coroutine by (cleaner) idle handler.
689 - updated to newest libcoro.
690 - implement enhanced support for valgrind.
691 - implement is_ready and return value for ready.
692 - removed Coro::Cont, it was a misnomer (it's generators),
693 and it was somewhat annoying to get it right. Will come back
694 if somebody asks for it :)
695 - many ->wait methods and Coro::Event could return
696 spuriously without the event having happened.
697
698 2.5 Tue Nov 7 12:22:33 CET 2006
699 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
700 use AnyEvent, moved them to Coro/.
701 - added Coro::LWP which contains all the uglyness required to
702 make LWP non-blocking.
703 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
704 - fixed another bug in Coro::Select when the timeout was undef.
705 - reuse PL_start_env for all coros, saves some memory per coroutine.
706 - manage PL_top_env differently, hopefully to avoid panic: top_env.
707 - timeout argument was not properly used in Coro::Socket.
708 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
709 - emulate undocumented(!) functionality of IO::Socket required
710 by LWP(!!).
711 - updated eg/lwp to work with newer lwp's.
712 - remove "FATAL: uncaught exception" prefix. Coroutines that die
713 kill the whole process, just as exceptions in the main "coroutine"
714 did already.
715
716 2.1 Wed Nov 1 23:01:13 CET 2006
717 - fix a long-standing bug in Coro::Select where select with
718 zero timeout would instead change the current default filehandle.
719 - use a simpler and hopefully more robust way to clone padlists
720 (uses less memory and a perl function instead of our own).
721 - coro can now create a stack guard area on many architectures.
722 - Coro::AIO properly reexports additional functions from IO::AIO.
723 - updated libcoro with a workaround for OS X,
724 pach and testing by Michael Schwern.
725
726 2.0 Tue Oct 24 05:47:17 CEST 2006
727 - support additional aio requests in Coro::AIO.
728
729 1.9 (never properly released due to a glitch)
730
731 1.8 Thu Feb 2 00:59:06 CET 2006
732 - applied suggested patch by SAMV to avoid problems during stupid
733 mark & sweep gc run.
734 - applied patch by Scott Walters for 5.9.3 compatibility.
735
736 1.7 Tue Dec 27 01:41:58 CET 2005
737 - added Coro::AIO, a thin wrapper around IO::AIO.
738 - improved Makefile.PL explanations.
739
740 1.6 Mon Dec 12 21:30:05 CET 2005
741 - additionally save PM_curpm might fix as-of-yet
742 unseen problems with regex matches being attributed to the
743 wrong package.
744 - add t/10_bugs.t, which currently checks against imho broken
745 perls that use 0.26MB of stack space per Perl_magic_get
746 invocation (newer linuxes) as opposed to the 0.0002MB perl
747 normally uses.
748 - make stacksize configurable for the ultimate debian experience.
749
750 1.51 Mon Dec 12 18:48:36 CET 2005
751 - remove debugging warn() accidentally left in Coro::Select.
752
753 1.5 Tue Nov 29 13:32:44 CET 2005
754 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
755 - fix Coro::Select 'select'.
756 - strict'ify some modules.
757 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
758 - use XSLoader in selected modules.
759 - remove some 5.6 compatibility cruft.
760
761 1.4 Tue Sep 6 00:11:05 CEST 2005
762 - libcoro did not take into account the trampoline on amd64,
763 when 'l'inux method was used.
764
765 1.31 Tue Aug 30 23:31:33 CEST 2005
766 - some portability fixes/workarounds.
767
768 1.3 Sat Aug 20 03:08:56 CEST 2005
769 - no code changes. module seems to work fine.
770 - improve Coro::State docs, remove reference to nonexisting
771 Coro::State::flush method.
772 - no longer autodetect windows, present it as an option instead.
773
774 1.2 Mon May 16 02:00:55 CEST 2005
775 - included libcoro.c earlier in State.xs, might improve portability.
776 - use a faster and possibly more stable Coro::Cont implementation.
777 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
778 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
779 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
780
781 1.11 Thu Mar 3 18:00:52 CET 2005
782 - change of contact address.
783
784 1.1 Tue Feb 22 20:51:16 CET 2005
785 - support [l]inux method on amd64.
786 - allow some context switches while compiling/eval'ing.
787
788 1.0 Tue Aug 10 03:47:27 CEST 2004
789 - Event 0.89 fixes the bug again.
790 - fix the remaining(??) memleak. I didn't free the cache-AV for
791 padlists, so one AV per created anon-sub leaked.
792
793 0.97 Fri May 14 15:23:32 CEST 2004
794 - remove stupid gcc extension that was an experiment only anyway.
795
796 0.96 Thu May 13 18:09:29 CEST 2004
797 - tell people if they have yet another broken Event variant.
798 - made Coro to work _better_ with threads (it seems that, as long
799 as you use Coros in one thread only everything will work fine.
800 Looking at the perl source the problem might be missing locking
801 between Coro and some perl internal routines. That's not easily
802 fixable).
803 - (hopefully) got rid of the static xnext inside transfer.
804
805 0.95 Sun Feb 15 03:20:28 CET 2004
806 - removed Coro::State::flush, implemented a new and much
807 faster caching mechanism by attaching magic to the CV.
808 - a memleak with (real) closures remains.
809 - dramatically reduced size of stacks to 65536 (linux/x86), after
810 observing that even myhttpd never had more than a kilobyte
811 of stack in use (stack is still only physically allocated in
812 pagesize increments for systems suporting mmap).
813 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
814 - fix for threaded perl.
815
816 0.9 Sun Nov 23 04:32:39 CET 2003
817 - fix a few very hard to track down but (of course) stupid
818 memory leaks.
819 - stringify version number to avoid locale problems :(.
820 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
821 hopefully ignoring it "just works".
822
823 0.8 Wed Nov 5 19:38:40 CET 2003
824 - port to perl5.005_03, but only 5.8.x+ is supported!
825 - honor LocalAddr even without LocalPort in Coro::Socket.
826
827 0.7 Tue May 27 03:12:38 CEST 2003
828 - the version jump indicates some level of testing, not gobs
829 of new features.
830 - uh... I found the prompt function in ExtUtils::MakeMaker.
831 Highly correct stuff, that is...
832 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
833 the fix is not well understood by the author, so beware :(.
834
835 0.652 Thu May 8 02:54:46 CEST 2003
836 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
837 cause version 4 doesn't have ucontext.
838 - Benjamin Reed reported that setjmp works fine on darwin,
839 so preselect it.
840
841 0.651 Sat Mar 29 15:00:23 CET 2003
842 - fix a bug in Coro::Handle where some bytes could get lost
843 on reading, reported by jason@nichego.net.
844
845 0.65 Sun Mar 23 00:08:26 CET 2003
846 - added a README.linux-glibc.
847 - new module Coro::Select.
848 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
849 (I never claimed that I know what I am doing ;).
850 - set default to setjmp/longjmp on non-x86-linux.
851
852 0.6 Thu Nov 21 11:09:06 CET 2002
853 - made lazy context switching the default.
854 - fixed the bug where SemaphoreSet::guard
855 was the same as timed_guard and timed_guard was missing.
856 - fix a memleak in Coro::Timer.
857 - reclassified context sharing as not experimental.
858
859 0.534 Sun Apr 14 03:05:12 CEST 2002
860 - fixed backspace => tab in header parsing (myhttpd).
861 - added eg/readline, for elmex.
862 - Coro::Event now calls ->start on first call instead of
863 ->again, which created wrong timeouts for the first ->next.
864 - fixed a bug where Coro::Socket returned a closed socket instead
865 of nothing on ECONNREFUSED and similar error conditions.
866
867 0.533 Mon Feb 18 18:49:40 CET 2002
868 - adapt to finally working Event-0.85.
869 - fixed buggy event initialization (trapped under windows, but how
870 could it possibly have worked anywhere?)
871
872 0.532 Wed Jan 16 02:45:32 CET 2002
873 - removed the reference to pp_entersub, might work on
874 (some) win32 perls now (testsuite works under cygwin,
875 without Event, which freezes).
876
877 0.531 Mon Dec 10 22:18:44 CET 2001
878 - Fixed a bug in SemaphoreSet::timed_down.
879
880 0.53 Tue Nov 27 21:11:13 CET 2001
881 - some tests for Event.
882 - slightly different internal architecture to get refcnt into the object
883 and not the reference. might fix a few bugs, certainly did introduce
884 new ones.
885 - Coro::Timer (independent of Event).
886 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
887
888 0.52 Tue Nov 6 21:36:18 CET 2001
889 - ported to cygwin (trivial).
890 patch by Gerrit P. Haase <gp@familiehaase.de>.
891 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
892
893 0.51 Thu Nov 1 20:39:01 CET 2001
894 - terminate/cancel now work properly (otherwise termination
895 could cause a "next coroutine is not and contains not..."
896 error.
897 - added Coro::Socket::shutdown.
898 - Coro::Event::loop no is the same as Event::loop.
899 - implemented terminate with args + join.
900
901 0.5 Fri Sep 28 16:15:35 CEST 2001
902 - fixed "print" on a Coro::Handle. The print method worked.
903 - small tweaks (seem to reduce memory consumption a lot)
904 in various modules.
905 - splendid use of "no warnings" scattered throughout.
906 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
907
908 0.49 Sun Sep 16 02:42:45 CEST 2001
909 - changed some method calls to function calls for speed
910 inside Coro::Handle.
911 - make Coro::Handle use an array instead of a hash for
912 speed reasons.
913 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
914 sjlj and ucontext should now work.
915 - IRIX-specific port for libcoro.
916 - swapped order of accept results to match IO::Socket.
917 - changed getsock/peername to sock/peername to match IO::Socket.
918 - fixed a bug that caused segfault when returning to main task
919 under some circumstances.
920 - other bugfixes.
921
922 0.45 Sun Sep 2 02:54:01 CEST 2001
923 - new method Coro::Handle::timeout.
924 - corrected speling of set...name to get...name in Coro::Socket.
925 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
926 - Coro::Handle now supports fileno and FILENO.
927 - added eg/myhttpd, a web server.
928 - Coro::Socket now accepts numeric ports as well as "port(number)"
929 syntax.
930 - moved some scheduling primitives into xs code. more to come.
931 - new simple priority system.
932 - implemented Coro::Event using XS, almost four times faster!
933 - small memory corruption problem fixed (boy that was difficult).
934
935 0.13 Wed Aug 8 16:53:07 CEST 2001
936 - new method Coro::Handle::readline.
937 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
938 (see eg/lwp on how to make LWP non-blocking).
939 - renamed Coro::Socket::new_inet to new.
940 - added Coro::Util, some utility functions.
941 - die/eval now works better (still get annoying (but true ;)
942 "Callback called exit" messages).
943 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
944 - much better Coro::Socket support.
945
946 0.12 Fri Jul 27 04:19:01 CEST 2001
947 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
948 - very experimental stack sharing algorithm. not 100% safe but
949 should work well in practise ;)
950 - added Coro::RWLock.
951 - Coro::Specific now works.
952
953 0.11 Tue Jul 24 22:49:21 CEST 2001
954 - added specialized hack for newer and older linux versions (fast).
955 - renamed Coro::Event::IO to Coro::Handle.
956 - new module Coro::Socket.
957
958 0.1 Tue Jul 24 01:47:53 CEST 2001
959 - release candidate 3. A rather new internal structure :(
960 - the great renaming: Damian Conway gave me a suitable replacement
961 for yield.
962 - added Coro/libcoro, a portable coroutine implementation for C
963 (not even perl-dependent!!!), which can be used by Coro::State.
964 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
965 perl-5.6.
966 - new function Coro::Event::idle.
967 - the idle process is now overriden
968 by default in Coro::Event.
969 - Coro::Channel now enforces the size.
970 - canceling events now works in all cases.
971 - Coro state now include $_ and $@.
972 - yet another bug workaround that I do not really understand :(
973 - new module Coro::Event::IO (very undocumented) to do
974 non-blocking i/o.
975 - performance tuning.
976
977 0.08 Thu Jul 19 06:13:25 CEST 2001
978 - release candidate 2 (new functionality).
979 - add Coro::State::flush function.
980 - transfer now supports a flags value (mostly for speed ATM).
981 - might compile and work in the presence of threads now.
982 - continuations have a saner syntax.
983 - no more memleaks.
984
985 0.07 Tue Jul 17 17:40:18 CEST 2001
986 - release candidate 1 ;)
987 - slightly nicer code.
988 - fixed a scheduling bug in Coro::Event.
989
990 0.06 Tue Jul 17 04:23:24 CEST 2001
991 - ok, I found the showstopper - the same sub must not be
992 re-entered in two different coroutines, otherwise => crash. I
993 see no easy solution to this problem, except by walking the call
994 chaing and saving/restoring the cv's, which is what I do now.
995 - memory leaks still latent, especially at thread termination.
996 - Coro::Event now works (haha).
997
998 0.05 Sun Jul 15 17:32:20 CEST 2001
999 - fixed a few issues in Coro::Event.
1000 - I forgot to include Coro::Event in 0.04 :(:(:(
1001
1002 0.04 Sun Jul 15 05:24:59 CEST 2001
1003 - @_ is now properly localized.
1004 - Coro::State is now easier subclassable.
1005 - Coro::Cont now coroutine-aware.
1006 - Coro::Specific is a low-overhead module to create
1007 coroutine-specific vars.
1008 - Coro::Event provides a simple interface to Event.
1009
1010 0.03 Fri Jul 13 14:51:52 CEST 2001
1011 - transfer() now implemented in XS (beware).
1012 - new module Coro::Cont for really faked continuations.
1013 - big internal architecture changes: Coro::State is now
1014 really low-level and can thus be used to implement other
1015 interesting things, While "Coro::" implements a process-like
1016 model. Still crude and subject to change.
1017 - $_ and $@ are no longer being localized.
1018
1019 0.02 Tue Jul 10 01:38:17 CEST 2001
1020 - implemented "async" attribute.
1021 - $_ and $@ are now localized.
1022 - added Coro::Channel.
1023 - more testcases, still no docs.
1024
1025 0.01 Tue Jul 3 02:18:41 CEST 2001
1026 - original version; copied from Convert::Scalar.
1027