ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.449
Committed: Thu Nov 20 07:12:41 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
Changes since 1.448: +4 -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 5.0
9 - NEW ARCHITECTURE: use the latest 4.x version if you experience
10 stability issues.
11 - bump API version to 7 - all dependents must be recompiled.
12 - removed timed_* functions - they were not being used anyways
13 and should be replaced by a more generic mechanism -
14 and were annoying to support anyways :)
15 - removed SemaphoreSet's waiter method - use sem instead.
16 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
17 - async_pool did free a scalar value twice
18 ("Attempt to unreference...").
19 - fix a disastrous bug in the readline optimisation
20 introduced in 4.801.
21 - fix a longstanding bug where calling terminate on a coro that
22 was waiting for a semaphore that was just becoming available
23 would cause a deadlock (semaphore would get into a state where
24 it was available but waiters were still blocked).
25 - calling throw on a coroutine that is waiting for a semaphore will
26 no longer make it acquire the semaphore (and thus leak a count).
27 - perl's process emulation is now not even theoretically supported
28 anymore.
29 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
30 conversion of callback-style to blocking-style.
31 - new methods $coro->schedule_to and ->cede_to, to specifically
32 schedule or cede to a specific coroutine.
33 - new function Coro::Semaphore::wait.
34 - use named constants in Coro::Channel (Richard Hundt).
35 - directly patch the entersub opcode calling SLF functions (cede,
36 transfer and so on). this does speed up context switching, but
37 more importanly, it frees us from the hardcoded behaviour of
38 entersub, so we might actually be able to return something from
39 those functions and atcually create new ones.
40 - take advantage of __builtin_frame_address on gcc.
41 - expose THX in coroapi (not sure whether this was a wise decision,
42 as "threaded" perls are running at half speed anyways).
43 - implement execute_slf (schedule-like-function) interface that makes
44 it possible to implement schedule-like-functions in XS.
45 - use new SLF interface to massively speed up Coro::EV by roughly a
46 factor of two.
47 - used new SLF interface to massively speed up Coro::Semaphore by a
48 factor of three.
49 - used new SLF interface to speed up Coro::AIO by roughly a factor of
50 four and reduce its memory usage considerably.
51 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
52 for a nice speedup and vastly more correct behaviour. Also implement
53 a new method "sem" to get at the underlying semaphore object.
54 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
55 (in comparison) 20-40% speedup.
56 - used new SLF interface to reimplement Coro::Signal gaining
57 some unknown (because I was too lazy), but certain, speedup, and also
58 making signals reliable for the first time.
59 - used new SLF interface and other optimisations to speed up async_pool
60 by a factor of two. It also doesn't rely on perl's exception mechanism
61 to exit anymore. The overhead for terminating an async_pool, coro over
62 a normal async is now very small.
63 - sped up coroutine creation/destruction by 40%.
64 - forgot to include Coro/libcoro/README in the dist for all these years.
65 - work around a freebsd pthreads bug (manual testcancel is required as
66 pthread_cond_wait isn't a cancellation point on freebsd).
67 - use new rouse functions to speed up and simplify Coro::BDB.
68 - make "prefer perl native functions" work with threaded perls.
69 - condense Coro::Debug ps output, hint at v and w flags.
70 - (libcoro) lots of minor cleanups and portability improvements.
71
72 4.914 Wed Nov 19 12:54:18 CET 2008
73 - fix a disastrous bug in the readline optimisation
74 introduced in 4.801.
75
76 4.913 Sat Nov 15 07:58:28 CET 2008
77 - async_pool did free a scalar value twice
78 ("Attempt to unreference...").
79
80 4.912 Thu Nov 13 18:31:23 CET 2008
81 - minor cleanups.
82 - use much larger stacks on linux and perl < 5.8.8.
83 - Coro::Debug::new_unix_server did not unlink the socket
84 when destroyed.
85
86 4.911 Tue Nov 11 04:26:17 CET 2008
87 - "port" to threaded perls.
88
89 4.91 Mon Nov 10 05:36:38 CET 2008
90 - the ->throw exception object no longer leaks.
91 - creating a new cctx leaked a scopestack entry (memleak).
92 - new coroutines didn't get created with a zero flags field
93 (unknown impact).
94 - calling ->throw on a not-yet-started coroutine should now work
95 instead of being ignored.
96 - ->throw is now supported on Coro::State objects.
97 - clean up cctx creation code a bit.
98 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
99
100 4.9 Sat Nov 8 17:45:27 CET 2008
101 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
102 - (libcoro) no longer leak threads in the experimental pthread backend,
103 also speed it up considerably.
104 - (libcoro) do not rely on makecontext passing void *'s unscathed.
105 - fix compiletime dependencies on libcoro in the Makefile.
106 - cctx_count wasn't always updated properly.
107 - Coro::State::cctx_stacksize wasn't applied correctly.
108 - new function Coro::State::cctx_max_idle.
109 - the default max number of idle C contexts is now 4.
110 - (libcoro) try harder to get _setjmp/_longjmp.
111 - (libcoro) cleanup and extend the libcoro API to officially
112 allow the creation of empty source contexts.
113 - very experimental workaround for the totally broken netbsd platform.
114 - tried to hack around openbsd bugs.
115
116 4.804 Wed Nov 5 16:36:00 CET 2008
117 - Coro::Debug::new_unix_server would not create a non-blocking listening
118 socket, sometimes causing freezes.
119 - (libcoro) fix misaligned stack points for setjmp and assembly
120 methods, which can cause crashes on x86/x86_64 with a sufficiently
121 aggressive compiler.
122 - new function: Coro::Debug::new_tcp_server.
123 - move ->throw into the Coro class because it only works on coro objects.
124
125 4.803 Mon Nov 3 17:16:12 CET 2008
126 - (libcoro) use a global asm statement to become independent of gcc
127 otpimisations for CORO_ASM (thanks to pippijn for the idea).
128 - try to workaround yet another broken bsd, this time dragonfly.
129
130 4.802 Thu Oct 30 10:56:12 CET 2008
131 - support -fno-omit-frame-pointer on x86 with the assembly method.
132 - tune 01_unblock.t tests a bit.
133
134 4.801 Wed Oct 22 18:33:37 CEST 2008
135 - improve readline speed for very long "lines".
136 - backport to 5.8.8.
137
138 4.8 Thu Oct 2 13:34:40 CEST 2008
139 - new function Coro::AIO::aio_wait.
140 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
141 - greatly speed up and reduce memory usage of Coro::AIO requests.
142 - implement some other µ-optimisations.
143
144 4.749 Mon Sep 29 14:40:12 CEST 2008
145 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
146
147 4.748 Sat Sep 27 14:03:19 CEST 2008
148 - implement, but do not document, PerlIO::cede(granularity).
149 - Coro::Storable forgot to wrap Storable::pstore.
150 - work around the multitude of leaks and memory corruption
151 bugs in PerlIO::via by using our own C-level perliol. As a side
152 effect, Coro::Storable is now much, much, much faster.
153
154 4.747 Tue Sep 23 01:59:41 CEST 2008
155 - fix a per-cv memleak (one empty array was leaked per
156 code reference).
157 - avoid a crash in coro->call|rss when the coroutine was already
158 destroyed (most noticably when using Coro::Debug::ps :)
159 - also protect *Storable::FILE.
160 - push up default storable granularity to 20ms.
161
162 4.746 Sun Sep 21 03:22:20 CEST 2008
163 - be more insistent on locking Storable against reentrancy
164 in Coro::Storable.
165 - move swap_def?v and throw to Coro::State, as documented.
166
167 4.745 Thu Jul 24 00:14:38 CEST 2008
168 - remove debugging code related to MgPV_nolen_const, also try to
169 make it compile with perl 5.8.6 (yes, apple apparently loves
170 outdated software). Reported by John S.
171
172 4.744 Tue Jul 8 22:06:35 CEST 2008
173 - correctly provide default DIE/WARN handlers as documented.
174 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
175 implementations inside perl work fine for us.
176
177 4.743 Mon Jun 16 00:21:57 CEST 2008
178 - when using Coro::EV without running EV::loop it could
179 result in busy-waiting for events, this has been fixed.
180 - reduce codesize and improve performance by using EV_DEFAULT_UC.
181
182 4.742 Sat May 31 14:10:21 CEST 2008
183 - implement a workaround for (some) perl <5.8.8 versions.
184 - require EV 3.3+.
185
186 4.741 Fri May 30 23:33:09 CEST 2008
187 - tell netbsd how utterly broken their imitation of an OS is
188 and refuse to build by default if pthreads are in use.
189 - switch to "s" method on all bsd's by default, as their ucontext
190 stuff seems just too broken.
191 - fix a bug in Coro::Select.
192
193 4.74 Thu May 29 20:05:31 CEST 2008
194 - do not test Coro::LWP for lack of dependencies.
195 - do not test Coro::Select for lack of working perls.
196
197 4.73 Thu May 29 2008
198 - fix a bug in Coro::EV which would cause it to block despite
199 there being runnable coroutines.
200 - sprinkle "no warnings" freely over everything, also suppress
201 warnings for some other modules.
202 - fix typo in WSAEWOULDBLOCK.
203
204 4.72 Sun May 25 05:14:36 CEST 2008
205 - tweak META.yaml a bit, unfortunately, there is no documented way
206 to have optional dependencies with CPAN. doh :(
207 - avoid running some tests on windows because they would fail due to
208 perl bug (broken fork, broken pipes...).
209 - work around perl on windows bugs where perl returns undocumented
210 error codes for sysread, syswrite etc. by taking advantage
211 of AnyEvent's workaround for that problem.
212 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
213 a common perl implementation bug on windows.
214 - use unix domain sockets in testsuite to work around
215 common perl implementation bugs on widows (they are emulated by
216 tcp sockets on windows. ugh.)
217
218 4.71 Sat May 24 20:01:27 CEST 2008
219 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
220 - take advantage of async name resolution of AnyEvent::Util.
221 - work around brutal inet_aton override in Coro::LWP.
222 - take advantage of the readyhook in Coro::EV, for smoother
223 scheduling.
224
225 4.7 Sun May 11 00:32:19 CEST 2008
226 - completely reworked the Coro manpage.
227 - added Coro::AnyEvent, generic event loop integration.
228 - implement cancel, ready and kill commands in Coro::Debug.
229 - document find_coro in Coro::Debug.
230 - incompatible change: rename has_stack to has_cctx.
231 - Coro::AIO and Coro::BDB no longer force event model detection,
232 use AnyEvent::AIO and AnyEvent::BDB.
233
234 4.6 Sat Apr 26 10:05:14 CEST 2008
235 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
236 by the Coro module. It was a mistake to have it in the first place.
237 - (experimental) support for activestate perl 5.10 (method "w").
238 - (experimental) support for strawberry perl (method "a").
239 - coro_sigelem_set did not return a value although it had to,
240 actual impact unknown.
241
242 4.51 Mon Apr 14 13:28:27 CEST 2008
243 - make it compile again on 5.8.
244
245 4.50 Thu Apr 10 09:43:17 CEST 2008
246 - I did it twice! (see 4.49).
247
248 4.49 Sun Apr 6 21:23:31 CEST 2008
249 - grr, instead of compiling the recent changes on 5.10 only they
250 were compiled on 5.8 only.
251
252 4.48 Sun Apr 6 20:36:46 CEST 2008
253 - allow coroutine switches during eval's under 5.10.x, as it seems
254 the parser is a per-interpreter option now, so this is safe
255 (this might fix the odd crashing bug).
256 - drop support for 5.9.x versions: they are dead, jim.
257
258 4.47 Sun Apr 6 00:37:52 CEST 2008
259 - force cctx allocation on API calls: we know we need to force one
260 and gcc actually manages to confuse our heuristic nowadays,
261 leading to crashes and worse.
262 - document force_cctx.
263
264 4.46 Fri Apr 4 22:05:43 CEST 2008
265 - upgrade libcoro, resulting in pthread-backend (which was only created
266 to fulfill the rules of the programming languages shootout).
267
268 4.45 Thu Mar 13 11:55:36 CET 2008
269 - fix a file leaking bug in eg/mhyttpd that would allow
270 downloading of any file (reported by oesi).
271 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
272 (also add testcase).
273 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
274 apparently having loads of fun with that).
275 - implement $state->swap_defsv, ->swap_defav and document ->throw.
276
277 4.4 Wed Feb 13 16:44:29 CET 2008
278 - only assume .cfi pseudo isns exist on GNU/Linux.
279 - add get_timed method to Coro::Channels.
280 - fixed Coro::Debug manpage.
281 - perl 5.11 compatbility improvement based on patch by
282 Andreas König.
283
284 4.37 Sun Jan 20 11:25:23 CET 2008
285 - soften the check for an active parser for perl 5.10 (avoiding
286 "Coro::State::transfer called while parsing" in many cases).
287
288 4.36 Sun Jan 13 10:53:56 CET 2008
289 - reset diehook when terminating from an async_pool as to not
290 trigger any __DIE__ handlers.
291
292 4.35 Sun Jan 13 04:14:13 CET 2008
293 - "bt" debug command now displays any exceptions
294 from longmess and also skips the innermost
295 stackframes, giving more useufl output.
296 - allow backtraces in coroutines blocked in Coro::EV,
297 at a <1% speed hit.
298 - handle localising of $SIG{__WARN__} and __DIE__
299 properly (with a proper amount of dirty hacking).
300
301 4.34 Sat Dec 22 17:49:53 CET 2007
302 - upgrade to EV version 2.0 API.
303
304 4.33 Mon Dec 17 08:36:12 CET 2007
305 - make Coro::AIO etc. loadable in the absence of EV.
306
307 4.32 Mon Dec 17 07:46:02 CET 2007
308 - majorly improved Coro::Handle's performance with Coro::EV.
309 - implemented the until now mythical Coro::BDB module.
310 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
311
312 4.31 Wed Dec 5 12:32:39 CET 2007
313 - remove warn statement form Coro::Util.
314
315 4.3 Tue Dec 4 20:33:14 CET 2007
316 - calls to the idle function could cause stack corruption
317 when the stack changed.
318 - do no longer rely on the presence of EV::DNS if EV is used
319 (because it is gone), but instead take avdantage of EV::ADNS
320 if available.
321 - add ($) prototypes to all functions in Coro::Storable.
322 - use a conventional (safer) idle callback in Coro::EV.
323 - do accept quickly in Coro::Debug to avoid endless loops.
324
325 4.22 Fri Nov 30 16:04:04 CET 2007
326 - really use optimised versions for Event and EV in Coro::Util
327 and Coro::Handle.
328
329 4.21 Sun Nov 25 10:48:59 CET 2007
330 - fix a spurious memory read.
331 - Coro::EV no longer keeps the eventloop "alive".
332
333 4.2 Fri Nov 9 20:47:05 CET 2007
334 - enable/disable tracing from a new coroutine, not a pooled one.
335 - fix a memleak in Coro::Event.
336 - removed killall call from fork_eval.
337 - made sure store_fd is already loaded so that fork_eval
338 does not have to parse autoload in each subprocess.
339 - only use assembly method if -O switch is in $Config{optimize}.
340 - add (optional) Coro::EV module, so far the best event loop module
341 directly supported by Coro.
342 - if the event model is EV, use EV::DNS to resolve
343 stuff in Coro::Util.
344 - don't get confused by multiple event notifications in Coro::Handle.
345 - initial support for EV (libevent interface).
346 - require Event and EV using configure_requires, to force their existance.
347
348 4.13 Wed Oct 24 07:26:45 CEST 2007
349 - add Coro::Storable::blocking_thaw.
350 - use a vastly more complicated technique to localise
351 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
352 - use a coroutine for the idle callback Coro::Event,
353 instead of running Event in the current coroutine context.
354 This also catches recursive invocations.
355 - actually report fork errors in gethostbyname and inet_aton.
356
357 4.11 Thu Oct 11 02:40:24 CEST 2007
358 - port to threaded perls.
359
360 4.1 Thu Oct 11 02:38:16 CEST 2007
361 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
362 be local to each coro (see Coro::State).
363 - incompatible change: for very deep reasons, cede and cede_notself
364 cannot return anything, so nothing will be returned.
365 - possibly bring back 5.10 compatibility (untested).
366 - work around stupid (and wrong) warning on 5.10 :(.
367 - overlay the saved state over the context stack. This saves
368 a few hundred bytes per coroutine on average and also
369 speeds up context switching a bit.
370 - further tune default stack sizes.
371 - (more) correctly calculate stack usage in coro_rss.
372 - Coro::Storable::blocking_* did not properly lock
373 resulting in races between coroutines.
374 - added Coro::Storable::guard.
375 - stopping to trace a coroutine could destroy the cctx of
376 an unrelated coroutine.
377 - explain the relationship between Perl and C coroutines in
378 more detail in Coro::State.
379 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
380 causing a fork per resolve. This also affected Coro::Socket.
381 - switch to a separate stack in $coro->call/eval to avoid
382 invalidating pointers.
383
384 4.03 Sat Oct 6 21:24:00 CEST 2007
385 - added Coro::throw method.
386 - minor code cleanups.
387
388 4.02 Sat Oct 6 02:36:47 CEST 2007
389 - fix a very minor per-coroutine memleak (a single codereference).
390 - fixed a bug where the currently in-use c context would be freed
391 prematurely (can happen only when programs change the stacksize
392 or use tracing).
393 - tracing can no longer keep a coro alive after it terminated.
394 - do static branch prediction in the common path for gcc. gives
395 about 2-5% speed improvement here.
396
397 4.01 Fri Oct 5 22:10:49 CEST 2007
398 - instead of recreating *a* standard output handle we simply
399 use STDOUT, which is faster and hopefully more robust.
400
401 4.0 Fri Oct 5 12:56:00 CEST 2007
402 - incompatibly changed Coro::Storable::freeze.
403 - major new feature: added Coro::Debug, for interactive coroutine
404 debugging, tracing and much more.
405 - major bug fix: unbelievable, but true: $_, $/ and many other
406 "saved" variables actually weren't being saved. This has been fixed,
407 of course, while increasing performance while losing all the save
408 flags.
409 - save flags are gone, and all the api functions dealing with them.
410 - added Coro::Semaphore::adjust.
411 - added Coro::Util::fork_eval.
412 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
413 - added Coro::killall.
414 - reduce initial stack sizes to allow for "micro-coroutines".
415 - better async_pool resource management, moved parts of async_pool
416 handling to XS (major speed improvement).
417 - actually croak before modifying important data structures.
418 - refuse to transfer while compiling.
419 - possibly support eval EXPR better now.
420 - enable assembly per default on linux+bsd x86+amd64.
421 - all internal members were renamed _something for easier subclassing.
422 - many minor tweaks.
423
424 3.63 Wed May 16 14:10:06 CEST 2007
425 - implement handcoded assembly for x86/amd64 SVR ABI.
426
427 3.62 Fri Apr 27 21:36:06 CEST 2007
428 - upgrade libcoro (which might set unwind info correctly).
429 - change default on linux to setjmp/longjmp.
430
431 3.61 Thu Apr 19 12:36:18 CEST 2007
432 - Coro::Storable caused an endless loop when thawing invalid
433 pst's sometimes.
434 - use a Semaphore in Coro::Storable, as Storable doesn't
435 seem to be reentrant (although it is documented to
436 be threadsafe...).
437 - fix Coro::Signal to bring back the original unreliable
438 but stateful semantics.
439 - fixed a lot of typos in Coro.pm (patch submitted by David
440 Steinbrunner, which applied flawlessly).
441
442 3.6 Sat Apr 14 17:13:31 CEST 2007
443 - added some bugfixes to get eg/myhttpd working again.
444 - added Coro::Storable for often-cede'ing freeze/thaw.
445 - try to do a clean exit when a coroutine calls exit
446 (EXPERIMENTAL).
447 - got rid of indirect call through _coro_init.
448 - updated the partly antique examples in eg/ to
449 work again and be a bit less magic, too.
450 - fixed Coro::Signal semantics to work as documented again.
451
452 3.55 Sun Mar 25 01:20:47 CET 2007
453 - add SAVE_DEFFH to save the default filehandle and enable
454 it by default.
455 - finally move socket-operations from Coro::Socket to Coro::Handle
456 to be able to unblock foreign sockets.
457 - add Coro::State::save_also and guarded_save.
458 - add count accessor to Coro::Semaphore.
459 - add Coro::State::cctx_stacksize.
460 - just for the fun of it, do not rely on implicit context,
461 which can dramatically improve performance, but people
462 using windows-process-emulation perls do not care much about
463 performance.
464
465 3.51 Sun Mar 4 14:18:04 CET 2007
466 - fixed a problem when you weakened references to Coro::State's
467 (patch by Yuval Kogman).
468
469 3.501 Wed Feb 28 12:44:07 CET 2007
470 - rename some global symbols as macosx from hell redefines
471 them without asking.
472
473 3.5 Tue Feb 13 20:22:53 CET 2007
474 - do AnyEvent model detection earlier, avoiding problems
475 caused by first using AnyEvent and later Coro::Event.
476 - implement and document Coro::Event event objects.
477 - fix a potential problem in Coro::Event causing crashes.
478 - initialise PL_comppad when creating a new coroutine,
479 avoids crashes on early coro destruction.
480
481 3.41 Mon Jan 22 19:19:49 CET 2007
482 - readline on Coro::Handle did not support undefined $/,
483 nor did it deliver partial lines on EOF or error.
484 - implement malloc fallback for stack allocation because
485 stupid broken idiotic OSX has a stupid broken
486 idiotic fits-the-whole-os mmap "implementation" and
487 my dick feels longer if Coro is portable even to
488 obsolete platforms.
489
490 3.4 Fri Jan 19 21:52:54 CET 2007
491 - remove t/09_timer.t, as it isn't really testing much
492 but was rather flaky in practise.
493 - async_pool coro would keep arguments and callback alive until
494 it was reused.
495 - cancellation of a coroutine could cause spurious idle calls
496 in cede_notself.
497
498 3.3 Sat Jan 6 03:45:00 CET 2007
499 - implement $coro->on_destroy.
500 - Coro::Event blocking semantics have been changed,
501 documented and - hopefully - improved.
502 - fix nice adding, not subtracting, from priority.
503 - fix ->prio and api_is_ready (patch by Mark Hinds).
504 - fixed an assert ("... == prev__cctx->idle_te")
505 that could errronously trigger.
506 - fix various large and small memleaks.
507 - use a (hopefully) more stable cancel implementation
508 that immediately frees the coroutine data.
509 - cede/cede_notself return a status now.
510 - added Coro::guard function.
511 - added a global coroutine pool for jobs (on my machine,
512 I can create and execute 48k simple coros/s with async,
513 and 128k coros with async_pool).
514 - Coro::AIO now uses the coroutine priority as io priority.
515
516 3.2 Fri Dec 22 05:07:09 CET 2006
517 - improve portability to slightly older perls.
518 - use cleaner coroutine destruction.
519 - simplify configuration for users.
520 - optionally (unrecommended) prefer perl functions over
521 their coro replacements.
522
523 3.11 Tue Dec 5 13:11:24 CET 2006
524 - fixed some bogus assert's, but as perl.h disables assert even
525 without NDEBUG (thank you very much), not too many people should
526 notice (that did include myself). Andreas König noticed, though :)
527 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
528 it. Also noticed by Andreas König :)
529 - save/restore tainted status.
530 - verified to pass the testsuite on my 5.9.5.
531
532 3.1 Mon Dec 4 23:03:40 CET 2006
533 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
534 - incompatible change: transfer flags are now per-state.
535 - give a better error message on deadlock.
536 - document Coro::nready.
537 - enhanced testsuite.
538
539 3.01 Sun Dec 3 23:47:42 CET 2006
540 - forgot to include Coro::Timer.
541
542 3.0 Sun Dec 3 22:57:25 CET 2006
543 - the "FINALLY COMPLETELY STABLE" release (coming soon:
544 the "FAMOUS LAST WORDS" release).
545 - implement a new stack sharing algorithm, which uses a stack
546 pool (size currently hardcoded).
547 - make stack sharing mandatory (it no longer uses a heuristic).
548 - eval/die no longer cause weird problems under heavy use.
549 - Coro::Event could cause livelocks if it was used but
550 no Coro::Event watchers were used.
551 - Coro::Event now uses asynccheck as prepare does not
552 check for changed watchers.
553 - Coro::Event allows multiple waiting coros and will wake up one
554 per event.
555 - Coro::Event should be cleaner and more efficient now.
556 - new utility function Coro::unblock_sub.
557 - document the sad fact that Event is no longer reentrant.
558 - putting a coroutine into the ready queue twice could under
559 some circumstances lead to stack corruption.
560 - minor incompatibility: subclassing Coro::State is supported
561 directly now without going through a _coro_state member.
562 - state/coro switching is much faster now.
563 - very minor optimisations and code/documentation cleanup.
564 - avoid problems due to compiler inlining.
565 - removed timers from Coro::Timer -> use AnyEvent instead.
566 - replaced Coro::idle coroutine by (cleaner) idle handler.
567 - updated to newest libcoro.
568 - implement enhanced support for valgrind.
569 - implement is_ready and return value for ready.
570 - removed Coro::Cont, it was a misnomer (it's generators),
571 and it was somewhat annoying to get it right. Will come back
572 if somebody asks for it :)
573 - many ->wait methods and Coro::Event could return
574 spuriously without the event having happened.
575
576 2.5 Tue Nov 7 12:22:33 CET 2006
577 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
578 use AnyEvent, moved them to Coro/.
579 - added Coro::LWP which contains all the uglyness required to
580 make LWP non-blocking.
581 - should work with perl 5.9.x now (Andreas König made me do it).
582 - fixed another bug in Coro::Select when the timeout was undef.
583 - reuse PL_start_env for all coros, saves some memory per coroutine.
584 - manage PL_top_env differently, hopefully to avoid panic: top_env.
585 - timeout argument was not properly used in Coro::Socket.
586 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
587 - emulate undocumented(!) functionality of IO::Socket required
588 by LWP(!!).
589 - updated eg/lwp to work with newer lwp's.
590 - remove "FATAL: uncaught exception" prefix. Coroutines that die
591 kill the whole process, just as exceptions in the main "coroutine"
592 did already.
593
594 2.1 Wed Nov 1 23:01:13 CET 2006
595 - fix a long-standing bug in Coro::Select where select with
596 zero timeout would instead change the current default filehandle.
597 - use a simpler and hopefully more robust way to clone padlists
598 (uses less memory and a perl function instead of our own).
599 - coro can now create a stack guard area on many architectures.
600 - Coro::AIO properly reexports additional functions from IO::AIO.
601 - updated libcoro with a workaround for OS X,
602 pach and testing by Michael Schwern.
603
604 2.0 Tue Oct 24 05:47:17 CEST 2006
605 - support additional aio requests in Coro::AIO.
606
607 1.9 (never properly released due to a glitch)
608
609 1.8 Thu Feb 2 00:59:06 CET 2006
610 - applied suggested patch by SAMV to avoid problems during stupid
611 mark & sweep gc run.
612 - applied patch by Scott Walters for 5.9.3 compatibility.
613
614 1.7 Tue Dec 27 01:41:58 CET 2005
615 - added Coro::AIO, a thin wrapper around IO::AIO.
616 - improved Makefile.PL explanations.
617
618 1.6 Mon Dec 12 21:30:05 CET 2005
619 - additionally save PM_curpm might fix as-of-yet
620 unseen problems with regex matches being attributed to the
621 wrong package.
622 - add t/10_bugs.t, which currently checks against imho broken
623 perls that use 0.26MB of stack space per Perl_magic_get
624 invocation (newer linuxes) as opposed to the 0.0002MB perl
625 normally uses.
626 - make stacksize configurable for the ultimate debian experience.
627
628 1.51 Mon Dec 12 18:48:36 CET 2005
629 - remove debugging warn() accidentally left in Coro::Select.
630
631 1.5 Tue Nov 29 13:32:44 CET 2005
632 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
633 - fix Coro::Select 'select'.
634 - strict'ify some modules.
635 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
636 - use XSLoader in selected modules.
637 - remove some 5.6 compatibility cruft.
638
639 1.4 Tue Sep 6 00:11:05 CEST 2005
640 - libcoro did not take into account the trampoline on amd64,
641 when 'l'inux method was used.
642
643 1.31 Tue Aug 30 23:31:33 CEST 2005
644 - some portability fixes/workarounds.
645
646 1.3 Sat Aug 20 03:08:56 CEST 2005
647 - no code changes. module seems to work fine.
648 - improve Coro::State docs, remove reference to nonexisting
649 Coro::State::flush method.
650 - no longer autodetect windows, present it as an option instead.
651
652 1.2 Mon May 16 02:00:55 CEST 2005
653 - included libcoro.c earlier in State.xs, might improve portability.
654 - use a faster and possibly more stable Coro::Cont implementation.
655 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
656 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
657 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
658
659 1.11 Thu Mar 3 18:00:52 CET 2005
660 - change of contact address.
661
662 1.1 Tue Feb 22 20:51:16 CET 2005
663 - support [l]inux method on amd64.
664 - allow some context switches while compiling/eval'ing.
665
666 1.0 Tue Aug 10 03:47:27 CEST 2004
667 - Event 0.89 fixes the bug again.
668 - fix the remaining(??) memleak. I didn't free the cache-AV for
669 padlists, so one AV per created anon-sub leaked.
670
671 0.97 Fri May 14 15:23:32 CEST 2004
672 - remove stupid gcc extension that was an experiment only anyway.
673
674 0.96 Thu May 13 18:09:29 CEST 2004
675 - tell people if they have yet another broken Event variant.
676 - made Coro to work _better_ with threads (it seems that, as long
677 as you use Coros in one thread only everything will work fine.
678 Looking at the perl source the problem might be missing locking
679 between Coro and some perl internal routines. That's not easily
680 fixable).
681 - (hopefully) got rid of the static xnext inside transfer.
682
683 0.95 Sun Feb 15 03:20:28 CET 2004
684 - removed Coro::State::flush, implemented a new and much
685 faster caching mechanism by attaching magic to the CV.
686 - a memleak with (real) closures remains.
687 - dramatically reduced size of stacks to 65536 (linux/x86), after
688 observing that even myhttpd never had more than a kilobyte
689 of stack in use (stack is still only physically allocated in
690 pagesize increments for systems suporting mmap).
691 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
692 - fix for threaded perl.
693
694 0.9 Sun Nov 23 04:32:39 CET 2003
695 - fix a few very hard to track down but (of course) stupid
696 memory leaks.
697 - stringify version number to avoid locale problems :(.
698 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
699 hopefully ignoring it "just works".
700
701 0.8 Wed Nov 5 19:38:40 CET 2003
702 - port to perl5.005_03, but only 5.8.x+ is supported!
703 - honor LocalAddr even without LocalPort in Coro::Socket.
704
705 0.7 Tue May 27 03:12:38 CEST 2003
706 - the version jump indicates some level of testing, not gobs
707 of new features.
708 - uh... I found the prompt function in ExtUtils::MakeMaker.
709 Highly correct stuff, that is...
710 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
711 the fix is not well understood by the author, so beware :(.
712
713 0.652 Thu May 8 02:54:46 CEST 2003
714 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
715 cause version 4 doesn't have ucontext.
716 - Benjamin Reed reported that setjmp works fine on darwin,
717 so preselect it.
718
719 0.651 Sat Mar 29 15:00:23 CET 2003
720 - fix a bug in Coro::Handle where some bytes could get lost
721 on reading, reported by jason@nichego.net.
722
723 0.65 Sun Mar 23 00:08:26 CET 2003
724 - added a README.linux-glibc.
725 - new module Coro::Select.
726 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
727 (I never claimed that I know what I am doing ;).
728 - set default to setjmp/longjmp on non-x86-linux.
729
730 0.6 Thu Nov 21 11:09:06 CET 2002
731 - made lazy context switching the default.
732 - fixed the bug where SemaphoreSet::guard
733 was the same as timed_guard and timed_guard was missing.
734 - fix a memleak in Coro::Timer.
735 - reclassified context sharing as not experimental.
736
737 0.534 Sun Apr 14 03:05:12 CEST 2002
738 - fixed backspace => tab in header parsing (myhttpd).
739 - added eg/readline, for elmex.
740 - Coro::Event now calls ->start on first call instead of
741 ->again, which created wrong timeouts for the first ->next.
742 - fixed a bug where Coro::Socket returned a closed socket instead
743 of nothing on ECONNREFUSED and similar error conditions.
744
745 0.533 Mon Feb 18 18:49:40 CET 2002
746 - adapt to finally working Event-0.85.
747 - fixed buggy event initialization (trapped under windows, but how
748 could it possibly have worked anywhere?)
749
750 0.532 Wed Jan 16 02:45:32 CET 2002
751 - removed the reference to pp_entersub, might work on
752 (some) win32 perls now (testsuite works under cygwin,
753 without Event, which freezes).
754
755 0.531 Mon Dec 10 22:18:44 CET 2001
756 - Fixed a bug in SemaphoreSet::timed_down.
757
758 0.53 Tue Nov 27 21:11:13 CET 2001
759 - some tests for Event.
760 - slightly different internal architecture to get refcnt into the object
761 and not the reference. might fix a few bugs, certainly did introduce
762 new ones.
763 - Coro::Timer (independent of Event).
764 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
765
766 0.52 Tue Nov 6 21:36:18 CET 2001
767 - ported to cygwin (trivial).
768 patch by Gerrit P. Haase <gp@familiehaase.de>.
769 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
770
771 0.51 Thu Nov 1 20:39:01 CET 2001
772 - terminate/cancel now work properly (otherwise termination
773 could cause a "next coroutine is not and contains not..."
774 error.
775 - added Coro::Socket::shutdown.
776 - Coro::Event::loop no is the same as Event::loop.
777 - implemented terminate with args + join.
778
779 0.5 Fri Sep 28 16:15:35 CEST 2001
780 - fixed "print" on a Coro::Handle. The print method worked.
781 - small tweaks (seem to reduce memory consumption a lot)
782 in various modules.
783 - splendid use of "no warnings" scattered throughout.
784 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
785
786 0.49 Sun Sep 16 02:42:45 CEST 2001
787 - changed some method calls to function calls for speed
788 inside Coro::Handle.
789 - make Coro::Handle use an array instead of a hash for
790 speed reasons.
791 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
792 sjlj and ucontext should now work.
793 - IRIX-specific port for libcoro.
794 - swapped order of accept results to match IO::Socket.
795 - changed getsock/peername to sock/peername to match IO::Socket.
796 - fixed a bug that caused segfault when returning to main task
797 under some circumstances.
798 - other bugfixes.
799
800 0.45 Sun Sep 2 02:54:01 CEST 2001
801 - new method Coro::Handle::timeout.
802 - corrected speling of set...name to get...name in Coro::Socket.
803 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
804 - Coro::Handle now supports fileno and FILENO.
805 - added eg/myhttpd, a web server.
806 - Coro::Socket now accepts numeric ports as well as "port(number)"
807 syntax.
808 - moved some scheduling primitives into xs code. more to come.
809 - new simple priority system.
810 - implemented Coro::Event using XS, almost four times faster!
811 - small memory corruption problem fixed (boy that was difficult).
812
813 0.13 Wed Aug 8 16:53:07 CEST 2001
814 - new method Coro::Handle::readline.
815 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
816 (see eg/lwp on how to make LWP non-blocking).
817 - renamed Coro::Socket::new_inet to new.
818 - added Coro::Util, some utility functions.
819 - die/eval now works better (still get annoying (but true ;)
820 "Callback called exit" messages).
821 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
822 - much better Coro::Socket support.
823
824 0.12 Fri Jul 27 04:19:01 CEST 2001
825 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
826 - very experimental stack sharing algorithm. not 100% safe but
827 should work well in practise ;)
828 - added Coro::RWLock.
829 - Coro::Specific now works.
830
831 0.11 Tue Jul 24 22:49:21 CEST 2001
832 - added specialized hack for newer and older linux versions (fast).
833 - renamed Coro::Event::IO to Coro::Handle.
834 - new module Coro::Socket.
835
836 0.1 Tue Jul 24 01:47:53 CEST 2001
837 - release candidate 3. A rather new internal structure :(
838 - the great renaming: Damian Conway gave me a suitable replacement
839 for yield.
840 - added Coro/libcoro, a portable coroutine implementation for C
841 (not even perl-dependent!!!), which can be used by Coro::State.
842 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
843 perl-5.6.
844 - new function Coro::Event::idle.
845 - the idle process is now overriden
846 by default in Coro::Event.
847 - Coro::Channel now enforces the size.
848 - canceling events now works in all cases.
849 - Coro state now include $_ and $@.
850 - yet another bug workaround that I do not really understand :(
851 - new module Coro::Event::IO (very undocumented) to do
852 non-blocking i/o.
853 - performance tuning.
854
855 0.08 Thu Jul 19 06:13:25 CEST 2001
856 - release candidate 2 (new functionality).
857 - add Coro::State::flush function.
858 - transfer now supports a flags value (mostly for speed ATM).
859 - might compile and work in the presence of threads now.
860 - continuations have a saner syntax.
861 - no more memleaks.
862
863 0.07 Tue Jul 17 17:40:18 CEST 2001
864 - release candidate 1 ;)
865 - slightly nicer code.
866 - fixed a scheduling bug in Coro::Event.
867
868 0.06 Tue Jul 17 04:23:24 CEST 2001
869 - ok, I found the showstopper - the same sub must not be
870 re-entered in two different coroutines, otherwise => crash. I
871 see no easy solution to this problem, except by walking the call
872 chaing and saving/restoring the cv's, which is what I do now.
873 - memory leaks still latent, especially at thread termination.
874 - Coro::Event now works (haha).
875
876 0.05 Sun Jul 15 17:32:20 CEST 2001
877 - fixed a few issues in Coro::Event.
878 - I forgot to include Coro::Event in 0.04 :(:(:(
879
880 0.04 Sun Jul 15 05:24:59 CEST 2001
881 - @_ is now properly localized.
882 - Coro::State is now easier subclassable.
883 - Coro::Cont now coroutine-aware.
884 - Coro::Specific is a low-overhead module to create
885 coroutine-specific vars.
886 - Coro::Event provides a simple interface to Event.
887
888 0.03 Fri Jul 13 14:51:52 CEST 2001
889 - transfer() now implemented in XS (beware).
890 - new module Coro::Cont for really faked continuations.
891 - big internal architecture changes: Coro::State is now
892 really low-level and can thus be used to implement other
893 interesting things, While "Coro::" implements a process-like
894 model. Still crude and subject to change.
895 - $_ and $@ are no longer being localized.
896
897 0.02 Tue Jul 10 01:38:17 CEST 2001
898 - implemented "async" attribute.
899 - $_ and $@ are now localized.
900 - added Coro::Channel.
901 - more testcases, still no docs.
902
903 0.01 Tue Jul 3 02:18:41 CEST 2001
904 - original version; copied from Convert::Scalar.
905