ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.457
Committed: Sat Nov 22 02:09:54 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
Changes since 1.456: +1 -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 - better 5.8.6 compatibility.
15
16 5.0 Thu Nov 20 10:35:05 CET 2008
17 - NEW ARCHITECTURE: use the latest 4.x version if you experience
18 stability issues.
19 - bump API version to 7 - all dependents must be recompiled.
20 - removed timed_* functions - they were not being used anyways
21 and should be replaced by a more generic mechanism -
22 and were annoying to support anyways :)
23 - removed SemaphoreSet's waiter method - use sem method instead.
24 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
25 - async_pool did free a scalar value twice
26 ("Attempt to unreference...").
27 - fix a longstanding bug where calling terminate on a coro that
28 was waiting for a semaphore that was just becoming available
29 would cause a deadlock (semaphore would get into a state where
30 it was available but waiters were still blocked).
31 - calling throw on a coroutine that is waiting for a semaphore will
32 no longer make it acquire the semaphore (and thus leak a count).
33 - perl's process emulation is now not even theoretically supported
34 anymore.
35 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
36 conversion of callback-style to blocking-style.
37 - new methods $coro->schedule_to and ->cede_to, to specifically
38 schedule or cede to a specific coroutine.
39 - new function Coro::Semaphore::wait.
40 - use named constants in Coro::Channel (Richard Hundt).
41 - directly patch the entersub opcode calling SLF functions (cede,
42 transfer and so on). this does speed up context switching, but
43 more importanly, it frees us from the hardcoded behaviour of
44 entersub, so we might actually be able to return something from
45 those functions and atcually create new ones.
46 - take advantage of __builtin_frame_address on gcc.
47 - expose THX in coroapi (not sure whether this was a wise decision,
48 as "threaded" perls are running at half speed anyways).
49 - implement execute_slf (schedule-like-function) interface that makes
50 it possible to implement schedule-like-functions in XS.
51 - use new SLF interface to massively speed up Coro::EV by roughly a
52 factor of two.
53 - used new SLF interface to massively speed up Coro::Semaphore by a
54 factor of three.
55 - used new SLF interface to speed up Coro::AIO by roughly a factor of
56 four and reduce its memory usage considerably.
57 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
58 for a nice speedup and vastly more correct behaviour. Also implement
59 a new method "sem" to get at the underlying semaphore object.
60 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
61 (in comparison) 20-40% speedup.
62 - used new SLF interface to reimplement Coro::Signal gaining
63 some unknown (because I was too lazy), but certain, speedup, and also
64 making signals reliable for the first time.
65 - used new SLF interface and other optimisations to speed up async_pool
66 by a factor of two. It also doesn't rely on perl's exception mechanism
67 to exit anymore. The overhead for terminating an async_pool, coro over
68 a normal async is now very small.
69 - sped up coroutine creation/destruction by 40%.
70 - forgot to include Coro/libcoro/README in the dist for all these years.
71 - work around a freebsd pthreads bug (manual testcancel is required as
72 pthread_cond_wait isn't a cancellation point on freebsd).
73 - use new rouse functions to speed up and simplify Coro::BDB.
74 - make "prefer perl native functions" work with threaded perls.
75 - condense Coro::Debug ps output, hint at v and w flags.
76 - (libcoro) lots of minor cleanups and portability improvements.
77
78 4.914 Wed Nov 19 12:54:18 CET 2008
79 - fix a disastrous bug in the readline optimisation
80 introduced in 4.801.
81
82 4.913 Sat Nov 15 07:58:28 CET 2008
83 - async_pool did free a scalar value twice
84 ("Attempt to unreference...").
85
86 4.912 Thu Nov 13 18:31:23 CET 2008
87 - minor cleanups.
88 - use much larger stacks on linux and perl < 5.8.8.
89 - Coro::Debug::new_unix_server did not unlink the socket
90 when destroyed.
91
92 4.911 Tue Nov 11 04:26:17 CET 2008
93 - "port" to threaded perls.
94
95 4.91 Mon Nov 10 05:36:38 CET 2008
96 - the ->throw exception object no longer leaks.
97 - creating a new cctx leaked a scopestack entry (memleak).
98 - new coroutines didn't get created with a zero flags field
99 (unknown impact).
100 - calling ->throw on a not-yet-started coroutine should now work
101 instead of being ignored.
102 - ->throw is now supported on Coro::State objects.
103 - clean up cctx creation code a bit.
104 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
105
106 4.9 Sat Nov 8 17:45:27 CET 2008
107 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
108 - (libcoro) no longer leak threads in the experimental pthread backend,
109 also speed it up considerably.
110 - (libcoro) do not rely on makecontext passing void *'s unscathed.
111 - fix compiletime dependencies on libcoro in the Makefile.
112 - cctx_count wasn't always updated properly.
113 - Coro::State::cctx_stacksize wasn't applied correctly.
114 - new function Coro::State::cctx_max_idle.
115 - the default max number of idle C contexts is now 4.
116 - (libcoro) try harder to get _setjmp/_longjmp.
117 - (libcoro) cleanup and extend the libcoro API to officially
118 allow the creation of empty source contexts.
119 - very experimental workaround for the totally broken netbsd platform.
120 - tried to hack around openbsd bugs.
121
122 4.804 Wed Nov 5 16:36:00 CET 2008
123 - Coro::Debug::new_unix_server would not create a non-blocking listening
124 socket, sometimes causing freezes.
125 - (libcoro) fix misaligned stack points for setjmp and assembly
126 methods, which can cause crashes on x86/x86_64 with a sufficiently
127 aggressive compiler.
128 - new function: Coro::Debug::new_tcp_server.
129 - move ->throw into the Coro class because it only works on coro objects.
130
131 4.803 Mon Nov 3 17:16:12 CET 2008
132 - (libcoro) use a global asm statement to become independent of gcc
133 otpimisations for CORO_ASM (thanks to pippijn for the idea).
134 - try to workaround yet another broken bsd, this time dragonfly.
135
136 4.802 Thu Oct 30 10:56:12 CET 2008
137 - support -fno-omit-frame-pointer on x86 with the assembly method.
138 - tune 01_unblock.t tests a bit.
139
140 4.801 Wed Oct 22 18:33:37 CEST 2008
141 - improve readline speed for very long "lines".
142 - backport to 5.8.8.
143
144 4.8 Thu Oct 2 13:34:40 CEST 2008
145 - new function Coro::AIO::aio_wait.
146 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
147 - greatly speed up and reduce memory usage of Coro::AIO requests.
148 - implement some other µ-optimisations.
149
150 4.749 Mon Sep 29 14:40:12 CEST 2008
151 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
152
153 4.748 Sat Sep 27 14:03:19 CEST 2008
154 - implement, but do not document, PerlIO::cede(granularity).
155 - Coro::Storable forgot to wrap Storable::pstore.
156 - work around the multitude of leaks and memory corruption
157 bugs in PerlIO::via by using our own C-level perliol. As a side
158 effect, Coro::Storable is now much, much, much faster.
159
160 4.747 Tue Sep 23 01:59:41 CEST 2008
161 - fix a per-cv memleak (one empty array was leaked per
162 code reference).
163 - avoid a crash in coro->call|rss when the coroutine was already
164 destroyed (most noticably when using Coro::Debug::ps :)
165 - also protect *Storable::FILE.
166 - push up default storable granularity to 20ms.
167
168 4.746 Sun Sep 21 03:22:20 CEST 2008
169 - be more insistent on locking Storable against reentrancy
170 in Coro::Storable.
171 - move swap_def?v and throw to Coro::State, as documented.
172
173 4.745 Thu Jul 24 00:14:38 CEST 2008
174 - remove debugging code related to MgPV_nolen_const, also try to
175 make it compile with perl 5.8.6 (yes, apple apparently loves
176 outdated software). Reported by John S.
177
178 4.744 Tue Jul 8 22:06:35 CEST 2008
179 - correctly provide default DIE/WARN handlers as documented.
180 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
181 implementations inside perl work fine for us.
182
183 4.743 Mon Jun 16 00:21:57 CEST 2008
184 - when using Coro::EV without running EV::loop it could
185 result in busy-waiting for events, this has been fixed.
186 - reduce codesize and improve performance by using EV_DEFAULT_UC.
187
188 4.742 Sat May 31 14:10:21 CEST 2008
189 - implement a workaround for (some) perl <5.8.8 versions.
190 - require EV 3.3+.
191
192 4.741 Fri May 30 23:33:09 CEST 2008
193 - tell netbsd how utterly broken their imitation of an OS is
194 and refuse to build by default if pthreads are in use.
195 - switch to "s" method on all bsd's by default, as their ucontext
196 stuff seems just too broken.
197 - fix a bug in Coro::Select.
198
199 4.74 Thu May 29 20:05:31 CEST 2008
200 - do not test Coro::LWP for lack of dependencies.
201 - do not test Coro::Select for lack of working perls.
202
203 4.73 Thu May 29 2008
204 - fix a bug in Coro::EV which would cause it to block despite
205 there being runnable coroutines.
206 - sprinkle "no warnings" freely over everything, also suppress
207 warnings for some other modules.
208 - fix typo in WSAEWOULDBLOCK.
209
210 4.72 Sun May 25 05:14:36 CEST 2008
211 - tweak META.yaml a bit, unfortunately, there is no documented way
212 to have optional dependencies with CPAN. doh :(
213 - avoid running some tests on windows because they would fail due to
214 perl bug (broken fork, broken pipes...).
215 - work around perl on windows bugs where perl returns undocumented
216 error codes for sysread, syswrite etc. by taking advantage
217 of AnyEvent's workaround for that problem.
218 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
219 a common perl implementation bug on windows.
220 - use unix domain sockets in testsuite to work around
221 common perl implementation bugs on widows (they are emulated by
222 tcp sockets on windows. ugh.)
223
224 4.71 Sat May 24 20:01:27 CEST 2008
225 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
226 - take advantage of async name resolution of AnyEvent::Util.
227 - work around brutal inet_aton override in Coro::LWP.
228 - take advantage of the readyhook in Coro::EV, for smoother
229 scheduling.
230
231 4.7 Sun May 11 00:32:19 CEST 2008
232 - completely reworked the Coro manpage.
233 - added Coro::AnyEvent, generic event loop integration.
234 - implement cancel, ready and kill commands in Coro::Debug.
235 - document find_coro in Coro::Debug.
236 - incompatible change: rename has_stack to has_cctx.
237 - Coro::AIO and Coro::BDB no longer force event model detection,
238 use AnyEvent::AIO and AnyEvent::BDB.
239
240 4.6 Sat Apr 26 10:05:14 CEST 2008
241 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
242 by the Coro module. It was a mistake to have it in the first place.
243 - (experimental) support for activestate perl 5.10 (method "w").
244 - (experimental) support for strawberry perl (method "a").
245 - coro_sigelem_set did not return a value although it had to,
246 actual impact unknown.
247
248 4.51 Mon Apr 14 13:28:27 CEST 2008
249 - make it compile again on 5.8.
250
251 4.50 Thu Apr 10 09:43:17 CEST 2008
252 - I did it twice! (see 4.49).
253
254 4.49 Sun Apr 6 21:23:31 CEST 2008
255 - grr, instead of compiling the recent changes on 5.10 only they
256 were compiled on 5.8 only.
257
258 4.48 Sun Apr 6 20:36:46 CEST 2008
259 - allow coroutine switches during eval's under 5.10.x, as it seems
260 the parser is a per-interpreter option now, so this is safe
261 (this might fix the odd crashing bug).
262 - drop support for 5.9.x versions: they are dead, jim.
263
264 4.47 Sun Apr 6 00:37:52 CEST 2008
265 - force cctx allocation on API calls: we know we need to force one
266 and gcc actually manages to confuse our heuristic nowadays,
267 leading to crashes and worse.
268 - document force_cctx.
269
270 4.46 Fri Apr 4 22:05:43 CEST 2008
271 - upgrade libcoro, resulting in pthread-backend (which was only created
272 to fulfill the rules of the programming languages shootout).
273
274 4.45 Thu Mar 13 11:55:36 CET 2008
275 - fix a file leaking bug in eg/mhyttpd that would allow
276 downloading of any file (reported by oesi).
277 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
278 (also add testcase).
279 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
280 apparently having loads of fun with that).
281 - implement $state->swap_defsv, ->swap_defav and document ->throw.
282
283 4.4 Wed Feb 13 16:44:29 CET 2008
284 - only assume .cfi pseudo isns exist on GNU/Linux.
285 - add get_timed method to Coro::Channels.
286 - fixed Coro::Debug manpage.
287 - perl 5.11 compatbility improvement based on patch by
288 Andreas König.
289
290 4.37 Sun Jan 20 11:25:23 CET 2008
291 - soften the check for an active parser for perl 5.10 (avoiding
292 "Coro::State::transfer called while parsing" in many cases).
293
294 4.36 Sun Jan 13 10:53:56 CET 2008
295 - reset diehook when terminating from an async_pool as to not
296 trigger any __DIE__ handlers.
297
298 4.35 Sun Jan 13 04:14:13 CET 2008
299 - "bt" debug command now displays any exceptions
300 from longmess and also skips the innermost
301 stackframes, giving more useufl output.
302 - allow backtraces in coroutines blocked in Coro::EV,
303 at a <1% speed hit.
304 - handle localising of $SIG{__WARN__} and __DIE__
305 properly (with a proper amount of dirty hacking).
306
307 4.34 Sat Dec 22 17:49:53 CET 2007
308 - upgrade to EV version 2.0 API.
309
310 4.33 Mon Dec 17 08:36:12 CET 2007
311 - make Coro::AIO etc. loadable in the absence of EV.
312
313 4.32 Mon Dec 17 07:46:02 CET 2007
314 - majorly improved Coro::Handle's performance with Coro::EV.
315 - implemented the until now mythical Coro::BDB module.
316 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
317
318 4.31 Wed Dec 5 12:32:39 CET 2007
319 - remove warn statement form Coro::Util.
320
321 4.3 Tue Dec 4 20:33:14 CET 2007
322 - calls to the idle function could cause stack corruption
323 when the stack changed.
324 - do no longer rely on the presence of EV::DNS if EV is used
325 (because it is gone), but instead take avdantage of EV::ADNS
326 if available.
327 - add ($) prototypes to all functions in Coro::Storable.
328 - use a conventional (safer) idle callback in Coro::EV.
329 - do accept quickly in Coro::Debug to avoid endless loops.
330
331 4.22 Fri Nov 30 16:04:04 CET 2007
332 - really use optimised versions for Event and EV in Coro::Util
333 and Coro::Handle.
334
335 4.21 Sun Nov 25 10:48:59 CET 2007
336 - fix a spurious memory read.
337 - Coro::EV no longer keeps the eventloop "alive".
338
339 4.2 Fri Nov 9 20:47:05 CET 2007
340 - enable/disable tracing from a new coroutine, not a pooled one.
341 - fix a memleak in Coro::Event.
342 - removed killall call from fork_eval.
343 - made sure store_fd is already loaded so that fork_eval
344 does not have to parse autoload in each subprocess.
345 - only use assembly method if -O switch is in $Config{optimize}.
346 - add (optional) Coro::EV module, so far the best event loop module
347 directly supported by Coro.
348 - if the event model is EV, use EV::DNS to resolve
349 stuff in Coro::Util.
350 - don't get confused by multiple event notifications in Coro::Handle.
351 - initial support for EV (libevent interface).
352 - require Event and EV using configure_requires, to force their existance.
353
354 4.13 Wed Oct 24 07:26:45 CEST 2007
355 - add Coro::Storable::blocking_thaw.
356 - use a vastly more complicated technique to localise
357 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
358 - use a coroutine for the idle callback Coro::Event,
359 instead of running Event in the current coroutine context.
360 This also catches recursive invocations.
361 - actually report fork errors in gethostbyname and inet_aton.
362
363 4.11 Thu Oct 11 02:40:24 CEST 2007
364 - port to threaded perls.
365
366 4.1 Thu Oct 11 02:38:16 CEST 2007
367 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
368 be local to each coro (see Coro::State).
369 - incompatible change: for very deep reasons, cede and cede_notself
370 cannot return anything, so nothing will be returned.
371 - possibly bring back 5.10 compatibility (untested).
372 - work around stupid (and wrong) warning on 5.10 :(.
373 - overlay the saved state over the context stack. This saves
374 a few hundred bytes per coroutine on average and also
375 speeds up context switching a bit.
376 - further tune default stack sizes.
377 - (more) correctly calculate stack usage in coro_rss.
378 - Coro::Storable::blocking_* did not properly lock
379 resulting in races between coroutines.
380 - added Coro::Storable::guard.
381 - stopping to trace a coroutine could destroy the cctx of
382 an unrelated coroutine.
383 - explain the relationship between Perl and C coroutines in
384 more detail in Coro::State.
385 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
386 causing a fork per resolve. This also affected Coro::Socket.
387 - switch to a separate stack in $coro->call/eval to avoid
388 invalidating pointers.
389
390 4.03 Sat Oct 6 21:24:00 CEST 2007
391 - added Coro::throw method.
392 - minor code cleanups.
393
394 4.02 Sat Oct 6 02:36:47 CEST 2007
395 - fix a very minor per-coroutine memleak (a single codereference).
396 - fixed a bug where the currently in-use c context would be freed
397 prematurely (can happen only when programs change the stacksize
398 or use tracing).
399 - tracing can no longer keep a coro alive after it terminated.
400 - do static branch prediction in the common path for gcc. gives
401 about 2-5% speed improvement here.
402
403 4.01 Fri Oct 5 22:10:49 CEST 2007
404 - instead of recreating *a* standard output handle we simply
405 use STDOUT, which is faster and hopefully more robust.
406
407 4.0 Fri Oct 5 12:56:00 CEST 2007
408 - incompatibly changed Coro::Storable::freeze.
409 - major new feature: added Coro::Debug, for interactive coroutine
410 debugging, tracing and much more.
411 - major bug fix: unbelievable, but true: $_, $/ and many other
412 "saved" variables actually weren't being saved. This has been fixed,
413 of course, while increasing performance while losing all the save
414 flags.
415 - save flags are gone, and all the api functions dealing with them.
416 - added Coro::Semaphore::adjust.
417 - added Coro::Util::fork_eval.
418 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
419 - added Coro::killall.
420 - reduce initial stack sizes to allow for "micro-coroutines".
421 - better async_pool resource management, moved parts of async_pool
422 handling to XS (major speed improvement).
423 - actually croak before modifying important data structures.
424 - refuse to transfer while compiling.
425 - possibly support eval EXPR better now.
426 - enable assembly per default on linux+bsd x86+amd64.
427 - all internal members were renamed _something for easier subclassing.
428 - many minor tweaks.
429
430 3.63 Wed May 16 14:10:06 CEST 2007
431 - implement handcoded assembly for x86/amd64 SVR ABI.
432
433 3.62 Fri Apr 27 21:36:06 CEST 2007
434 - upgrade libcoro (which might set unwind info correctly).
435 - change default on linux to setjmp/longjmp.
436
437 3.61 Thu Apr 19 12:36:18 CEST 2007
438 - Coro::Storable caused an endless loop when thawing invalid
439 pst's sometimes.
440 - use a Semaphore in Coro::Storable, as Storable doesn't
441 seem to be reentrant (although it is documented to
442 be threadsafe...).
443 - fix Coro::Signal to bring back the original unreliable
444 but stateful semantics.
445 - fixed a lot of typos in Coro.pm (patch submitted by David
446 Steinbrunner, which applied flawlessly).
447
448 3.6 Sat Apr 14 17:13:31 CEST 2007
449 - added some bugfixes to get eg/myhttpd working again.
450 - added Coro::Storable for often-cede'ing freeze/thaw.
451 - try to do a clean exit when a coroutine calls exit
452 (EXPERIMENTAL).
453 - got rid of indirect call through _coro_init.
454 - updated the partly antique examples in eg/ to
455 work again and be a bit less magic, too.
456 - fixed Coro::Signal semantics to work as documented again.
457
458 3.55 Sun Mar 25 01:20:47 CET 2007
459 - add SAVE_DEFFH to save the default filehandle and enable
460 it by default.
461 - finally move socket-operations from Coro::Socket to Coro::Handle
462 to be able to unblock foreign sockets.
463 - add Coro::State::save_also and guarded_save.
464 - add count accessor to Coro::Semaphore.
465 - add Coro::State::cctx_stacksize.
466 - just for the fun of it, do not rely on implicit context,
467 which can dramatically improve performance, but people
468 using windows-process-emulation perls do not care much about
469 performance.
470
471 3.51 Sun Mar 4 14:18:04 CET 2007
472 - fixed a problem when you weakened references to Coro::State's
473 (patch by Yuval Kogman).
474
475 3.501 Wed Feb 28 12:44:07 CET 2007
476 - rename some global symbols as macosx from hell redefines
477 them without asking.
478
479 3.5 Tue Feb 13 20:22:53 CET 2007
480 - do AnyEvent model detection earlier, avoiding problems
481 caused by first using AnyEvent and later Coro::Event.
482 - implement and document Coro::Event event objects.
483 - fix a potential problem in Coro::Event causing crashes.
484 - initialise PL_comppad when creating a new coroutine,
485 avoids crashes on early coro destruction.
486
487 3.41 Mon Jan 22 19:19:49 CET 2007
488 - readline on Coro::Handle did not support undefined $/,
489 nor did it deliver partial lines on EOF or error.
490 - implement malloc fallback for stack allocation because
491 stupid broken idiotic OSX has a stupid broken
492 idiotic fits-the-whole-os mmap "implementation" and
493 my dick feels longer if Coro is portable even to
494 obsolete platforms.
495
496 3.4 Fri Jan 19 21:52:54 CET 2007
497 - remove t/09_timer.t, as it isn't really testing much
498 but was rather flaky in practise.
499 - async_pool coro would keep arguments and callback alive until
500 it was reused.
501 - cancellation of a coroutine could cause spurious idle calls
502 in cede_notself.
503
504 3.3 Sat Jan 6 03:45:00 CET 2007
505 - implement $coro->on_destroy.
506 - Coro::Event blocking semantics have been changed,
507 documented and - hopefully - improved.
508 - fix nice adding, not subtracting, from priority.
509 - fix ->prio and api_is_ready (patch by Mark Hinds).
510 - fixed an assert ("... == prev__cctx->idle_te")
511 that could errronously trigger.
512 - fix various large and small memleaks.
513 - use a (hopefully) more stable cancel implementation
514 that immediately frees the coroutine data.
515 - cede/cede_notself return a status now.
516 - added Coro::guard function.
517 - added a global coroutine pool for jobs (on my machine,
518 I can create and execute 48k simple coros/s with async,
519 and 128k coros with async_pool).
520 - Coro::AIO now uses the coroutine priority as io priority.
521
522 3.2 Fri Dec 22 05:07:09 CET 2006
523 - improve portability to slightly older perls.
524 - use cleaner coroutine destruction.
525 - simplify configuration for users.
526 - optionally (unrecommended) prefer perl functions over
527 their coro replacements.
528
529 3.11 Tue Dec 5 13:11:24 CET 2006
530 - fixed some bogus assert's, but as perl.h disables assert even
531 without NDEBUG (thank you very much), not too many people should
532 notice (that did include myself). Andreas König noticed, though :)
533 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
534 it. Also noticed by Andreas König :)
535 - save/restore tainted status.
536 - verified to pass the testsuite on my 5.9.5.
537
538 3.1 Mon Dec 4 23:03:40 CET 2006
539 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
540 - incompatible change: transfer flags are now per-state.
541 - give a better error message on deadlock.
542 - document Coro::nready.
543 - enhanced testsuite.
544
545 3.01 Sun Dec 3 23:47:42 CET 2006
546 - forgot to include Coro::Timer.
547
548 3.0 Sun Dec 3 22:57:25 CET 2006
549 - the "FINALLY COMPLETELY STABLE" release (coming soon:
550 the "FAMOUS LAST WORDS" release).
551 - implement a new stack sharing algorithm, which uses a stack
552 pool (size currently hardcoded).
553 - make stack sharing mandatory (it no longer uses a heuristic).
554 - eval/die no longer cause weird problems under heavy use.
555 - Coro::Event could cause livelocks if it was used but
556 no Coro::Event watchers were used.
557 - Coro::Event now uses asynccheck as prepare does not
558 check for changed watchers.
559 - Coro::Event allows multiple waiting coros and will wake up one
560 per event.
561 - Coro::Event should be cleaner and more efficient now.
562 - new utility function Coro::unblock_sub.
563 - document the sad fact that Event is no longer reentrant.
564 - putting a coroutine into the ready queue twice could under
565 some circumstances lead to stack corruption.
566 - minor incompatibility: subclassing Coro::State is supported
567 directly now without going through a _coro_state member.
568 - state/coro switching is much faster now.
569 - very minor optimisations and code/documentation cleanup.
570 - avoid problems due to compiler inlining.
571 - removed timers from Coro::Timer -> use AnyEvent instead.
572 - replaced Coro::idle coroutine by (cleaner) idle handler.
573 - updated to newest libcoro.
574 - implement enhanced support for valgrind.
575 - implement is_ready and return value for ready.
576 - removed Coro::Cont, it was a misnomer (it's generators),
577 and it was somewhat annoying to get it right. Will come back
578 if somebody asks for it :)
579 - many ->wait methods and Coro::Event could return
580 spuriously without the event having happened.
581
582 2.5 Tue Nov 7 12:22:33 CET 2006
583 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
584 use AnyEvent, moved them to Coro/.
585 - added Coro::LWP which contains all the uglyness required to
586 make LWP non-blocking.
587 - should work with perl 5.9.x now (Andreas König made me do it).
588 - fixed another bug in Coro::Select when the timeout was undef.
589 - reuse PL_start_env for all coros, saves some memory per coroutine.
590 - manage PL_top_env differently, hopefully to avoid panic: top_env.
591 - timeout argument was not properly used in Coro::Socket.
592 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
593 - emulate undocumented(!) functionality of IO::Socket required
594 by LWP(!!).
595 - updated eg/lwp to work with newer lwp's.
596 - remove "FATAL: uncaught exception" prefix. Coroutines that die
597 kill the whole process, just as exceptions in the main "coroutine"
598 did already.
599
600 2.1 Wed Nov 1 23:01:13 CET 2006
601 - fix a long-standing bug in Coro::Select where select with
602 zero timeout would instead change the current default filehandle.
603 - use a simpler and hopefully more robust way to clone padlists
604 (uses less memory and a perl function instead of our own).
605 - coro can now create a stack guard area on many architectures.
606 - Coro::AIO properly reexports additional functions from IO::AIO.
607 - updated libcoro with a workaround for OS X,
608 pach and testing by Michael Schwern.
609
610 2.0 Tue Oct 24 05:47:17 CEST 2006
611 - support additional aio requests in Coro::AIO.
612
613 1.9 (never properly released due to a glitch)
614
615 1.8 Thu Feb 2 00:59:06 CET 2006
616 - applied suggested patch by SAMV to avoid problems during stupid
617 mark & sweep gc run.
618 - applied patch by Scott Walters for 5.9.3 compatibility.
619
620 1.7 Tue Dec 27 01:41:58 CET 2005
621 - added Coro::AIO, a thin wrapper around IO::AIO.
622 - improved Makefile.PL explanations.
623
624 1.6 Mon Dec 12 21:30:05 CET 2005
625 - additionally save PM_curpm might fix as-of-yet
626 unseen problems with regex matches being attributed to the
627 wrong package.
628 - add t/10_bugs.t, which currently checks against imho broken
629 perls that use 0.26MB of stack space per Perl_magic_get
630 invocation (newer linuxes) as opposed to the 0.0002MB perl
631 normally uses.
632 - make stacksize configurable for the ultimate debian experience.
633
634 1.51 Mon Dec 12 18:48:36 CET 2005
635 - remove debugging warn() accidentally left in Coro::Select.
636
637 1.5 Tue Nov 29 13:32:44 CET 2005
638 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
639 - fix Coro::Select 'select'.
640 - strict'ify some modules.
641 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
642 - use XSLoader in selected modules.
643 - remove some 5.6 compatibility cruft.
644
645 1.4 Tue Sep 6 00:11:05 CEST 2005
646 - libcoro did not take into account the trampoline on amd64,
647 when 'l'inux method was used.
648
649 1.31 Tue Aug 30 23:31:33 CEST 2005
650 - some portability fixes/workarounds.
651
652 1.3 Sat Aug 20 03:08:56 CEST 2005
653 - no code changes. module seems to work fine.
654 - improve Coro::State docs, remove reference to nonexisting
655 Coro::State::flush method.
656 - no longer autodetect windows, present it as an option instead.
657
658 1.2 Mon May 16 02:00:55 CEST 2005
659 - included libcoro.c earlier in State.xs, might improve portability.
660 - use a faster and possibly more stable Coro::Cont implementation.
661 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
662 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
663 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
664
665 1.11 Thu Mar 3 18:00:52 CET 2005
666 - change of contact address.
667
668 1.1 Tue Feb 22 20:51:16 CET 2005
669 - support [l]inux method on amd64.
670 - allow some context switches while compiling/eval'ing.
671
672 1.0 Tue Aug 10 03:47:27 CEST 2004
673 - Event 0.89 fixes the bug again.
674 - fix the remaining(??) memleak. I didn't free the cache-AV for
675 padlists, so one AV per created anon-sub leaked.
676
677 0.97 Fri May 14 15:23:32 CEST 2004
678 - remove stupid gcc extension that was an experiment only anyway.
679
680 0.96 Thu May 13 18:09:29 CEST 2004
681 - tell people if they have yet another broken Event variant.
682 - made Coro to work _better_ with threads (it seems that, as long
683 as you use Coros in one thread only everything will work fine.
684 Looking at the perl source the problem might be missing locking
685 between Coro and some perl internal routines. That's not easily
686 fixable).
687 - (hopefully) got rid of the static xnext inside transfer.
688
689 0.95 Sun Feb 15 03:20:28 CET 2004
690 - removed Coro::State::flush, implemented a new and much
691 faster caching mechanism by attaching magic to the CV.
692 - a memleak with (real) closures remains.
693 - dramatically reduced size of stacks to 65536 (linux/x86), after
694 observing that even myhttpd never had more than a kilobyte
695 of stack in use (stack is still only physically allocated in
696 pagesize increments for systems suporting mmap).
697 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
698 - fix for threaded perl.
699
700 0.9 Sun Nov 23 04:32:39 CET 2003
701 - fix a few very hard to track down but (of course) stupid
702 memory leaks.
703 - stringify version number to avoid locale problems :(.
704 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
705 hopefully ignoring it "just works".
706
707 0.8 Wed Nov 5 19:38:40 CET 2003
708 - port to perl5.005_03, but only 5.8.x+ is supported!
709 - honor LocalAddr even without LocalPort in Coro::Socket.
710
711 0.7 Tue May 27 03:12:38 CEST 2003
712 - the version jump indicates some level of testing, not gobs
713 of new features.
714 - uh... I found the prompt function in ExtUtils::MakeMaker.
715 Highly correct stuff, that is...
716 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
717 the fix is not well understood by the author, so beware :(.
718
719 0.652 Thu May 8 02:54:46 CEST 2003
720 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
721 cause version 4 doesn't have ucontext.
722 - Benjamin Reed reported that setjmp works fine on darwin,
723 so preselect it.
724
725 0.651 Sat Mar 29 15:00:23 CET 2003
726 - fix a bug in Coro::Handle where some bytes could get lost
727 on reading, reported by jason@nichego.net.
728
729 0.65 Sun Mar 23 00:08:26 CET 2003
730 - added a README.linux-glibc.
731 - new module Coro::Select.
732 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
733 (I never claimed that I know what I am doing ;).
734 - set default to setjmp/longjmp on non-x86-linux.
735
736 0.6 Thu Nov 21 11:09:06 CET 2002
737 - made lazy context switching the default.
738 - fixed the bug where SemaphoreSet::guard
739 was the same as timed_guard and timed_guard was missing.
740 - fix a memleak in Coro::Timer.
741 - reclassified context sharing as not experimental.
742
743 0.534 Sun Apr 14 03:05:12 CEST 2002
744 - fixed backspace => tab in header parsing (myhttpd).
745 - added eg/readline, for elmex.
746 - Coro::Event now calls ->start on first call instead of
747 ->again, which created wrong timeouts for the first ->next.
748 - fixed a bug where Coro::Socket returned a closed socket instead
749 of nothing on ECONNREFUSED and similar error conditions.
750
751 0.533 Mon Feb 18 18:49:40 CET 2002
752 - adapt to finally working Event-0.85.
753 - fixed buggy event initialization (trapped under windows, but how
754 could it possibly have worked anywhere?)
755
756 0.532 Wed Jan 16 02:45:32 CET 2002
757 - removed the reference to pp_entersub, might work on
758 (some) win32 perls now (testsuite works under cygwin,
759 without Event, which freezes).
760
761 0.531 Mon Dec 10 22:18:44 CET 2001
762 - Fixed a bug in SemaphoreSet::timed_down.
763
764 0.53 Tue Nov 27 21:11:13 CET 2001
765 - some tests for Event.
766 - slightly different internal architecture to get refcnt into the object
767 and not the reference. might fix a few bugs, certainly did introduce
768 new ones.
769 - Coro::Timer (independent of Event).
770 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
771
772 0.52 Tue Nov 6 21:36:18 CET 2001
773 - ported to cygwin (trivial).
774 patch by Gerrit P. Haase <gp@familiehaase.de>.
775 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
776
777 0.51 Thu Nov 1 20:39:01 CET 2001
778 - terminate/cancel now work properly (otherwise termination
779 could cause a "next coroutine is not and contains not..."
780 error.
781 - added Coro::Socket::shutdown.
782 - Coro::Event::loop no is the same as Event::loop.
783 - implemented terminate with args + join.
784
785 0.5 Fri Sep 28 16:15:35 CEST 2001
786 - fixed "print" on a Coro::Handle. The print method worked.
787 - small tweaks (seem to reduce memory consumption a lot)
788 in various modules.
789 - splendid use of "no warnings" scattered throughout.
790 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
791
792 0.49 Sun Sep 16 02:42:45 CEST 2001
793 - changed some method calls to function calls for speed
794 inside Coro::Handle.
795 - make Coro::Handle use an array instead of a hash for
796 speed reasons.
797 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
798 sjlj and ucontext should now work.
799 - IRIX-specific port for libcoro.
800 - swapped order of accept results to match IO::Socket.
801 - changed getsock/peername to sock/peername to match IO::Socket.
802 - fixed a bug that caused segfault when returning to main task
803 under some circumstances.
804 - other bugfixes.
805
806 0.45 Sun Sep 2 02:54:01 CEST 2001
807 - new method Coro::Handle::timeout.
808 - corrected speling of set...name to get...name in Coro::Socket.
809 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
810 - Coro::Handle now supports fileno and FILENO.
811 - added eg/myhttpd, a web server.
812 - Coro::Socket now accepts numeric ports as well as "port(number)"
813 syntax.
814 - moved some scheduling primitives into xs code. more to come.
815 - new simple priority system.
816 - implemented Coro::Event using XS, almost four times faster!
817 - small memory corruption problem fixed (boy that was difficult).
818
819 0.13 Wed Aug 8 16:53:07 CEST 2001
820 - new method Coro::Handle::readline.
821 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
822 (see eg/lwp on how to make LWP non-blocking).
823 - renamed Coro::Socket::new_inet to new.
824 - added Coro::Util, some utility functions.
825 - die/eval now works better (still get annoying (but true ;)
826 "Callback called exit" messages).
827 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
828 - much better Coro::Socket support.
829
830 0.12 Fri Jul 27 04:19:01 CEST 2001
831 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
832 - very experimental stack sharing algorithm. not 100% safe but
833 should work well in practise ;)
834 - added Coro::RWLock.
835 - Coro::Specific now works.
836
837 0.11 Tue Jul 24 22:49:21 CEST 2001
838 - added specialized hack for newer and older linux versions (fast).
839 - renamed Coro::Event::IO to Coro::Handle.
840 - new module Coro::Socket.
841
842 0.1 Tue Jul 24 01:47:53 CEST 2001
843 - release candidate 3. A rather new internal structure :(
844 - the great renaming: Damian Conway gave me a suitable replacement
845 for yield.
846 - added Coro/libcoro, a portable coroutine implementation for C
847 (not even perl-dependent!!!), which can be used by Coro::State.
848 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
849 perl-5.6.
850 - new function Coro::Event::idle.
851 - the idle process is now overriden
852 by default in Coro::Event.
853 - Coro::Channel now enforces the size.
854 - canceling events now works in all cases.
855 - Coro state now include $_ and $@.
856 - yet another bug workaround that I do not really understand :(
857 - new module Coro::Event::IO (very undocumented) to do
858 non-blocking i/o.
859 - performance tuning.
860
861 0.08 Thu Jul 19 06:13:25 CEST 2001
862 - release candidate 2 (new functionality).
863 - add Coro::State::flush function.
864 - transfer now supports a flags value (mostly for speed ATM).
865 - might compile and work in the presence of threads now.
866 - continuations have a saner syntax.
867 - no more memleaks.
868
869 0.07 Tue Jul 17 17:40:18 CEST 2001
870 - release candidate 1 ;)
871 - slightly nicer code.
872 - fixed a scheduling bug in Coro::Event.
873
874 0.06 Tue Jul 17 04:23:24 CEST 2001
875 - ok, I found the showstopper - the same sub must not be
876 re-entered in two different coroutines, otherwise => crash. I
877 see no easy solution to this problem, except by walking the call
878 chaing and saving/restoring the cv's, which is what I do now.
879 - memory leaks still latent, especially at thread termination.
880 - Coro::Event now works (haha).
881
882 0.05 Sun Jul 15 17:32:20 CEST 2001
883 - fixed a few issues in Coro::Event.
884 - I forgot to include Coro::Event in 0.04 :(:(:(
885
886 0.04 Sun Jul 15 05:24:59 CEST 2001
887 - @_ is now properly localized.
888 - Coro::State is now easier subclassable.
889 - Coro::Cont now coroutine-aware.
890 - Coro::Specific is a low-overhead module to create
891 coroutine-specific vars.
892 - Coro::Event provides a simple interface to Event.
893
894 0.03 Fri Jul 13 14:51:52 CEST 2001
895 - transfer() now implemented in XS (beware).
896 - new module Coro::Cont for really faked continuations.
897 - big internal architecture changes: Coro::State is now
898 really low-level and can thus be used to implement other
899 interesting things, While "Coro::" implements a process-like
900 model. Still crude and subject to change.
901 - $_ and $@ are no longer being localized.
902
903 0.02 Tue Jul 10 01:38:17 CEST 2001
904 - implemented "async" attribute.
905 - $_ and $@ are now localized.
906 - added Coro::Channel.
907 - more testcases, still no docs.
908
909 0.01 Tue Jul 3 02:18:41 CEST 2001
910 - original version; copied from Convert::Scalar.
911