ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.456
Committed: Fri Nov 21 07:29:04 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
Changes since 1.455: +2 -0 lines
Log Message:
*** empty log message ***

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