ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.477
Committed: Sat Dec 13 19:18:36 2008 UTC (15 years, 5 months ago) by root
Branch: MAIN
Changes since 1.476: +3 -1 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 new -> ready
9 new -> canceled
10 ready -> running
11 ready -> canceled
12 running -> ready
13 running -> blocked
14 running -> canceled
15 blocked -> ready
16 blocked -> canceled
17
18 - don't set diehook to C<undef> but instead to NULL, to avoid
19 spurious warnings.
20 - take advantage of the new Guard module.
21 - deprecate Coro::guard.
22
23 5.12 Sun Dec 7 13:30:38 CET 2008
24 - add default config for MirOS, which seems to be a bug-to-bug
25 compatible fork of openbsd ("world domination by releasing
26 openbsd cvs before the openbsd folks do it" or so :).
27 - free_padlist did destroy the names pad, not good, but didn't
28 seem to bother perl - this could fix issues such as eval ""
29 inside a function called from multiple coroutines.
30 - use a different method to detect destruction time.
31 - be more careful when freeing padlists just before global
32 destruction.
33 - fixed and expanded the call/cc example.
34 - renamed _terminate to _coro_run.
35 - new method Coro::Channel->shutdown.
36 - try pthreads on openbsd <4.4 (broken sigaltstack, will
37 pthreads fare better?).
38 - be less picky about destroying "a" running coroutine.
39
40 5.11 Tue Nov 25 21:49:05 CET 2008
41 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
42 working code with failed assertions, introducing more bugs than
43 it fixes, requiring elaborate workarounds :(
44
45 5.1 Mon Nov 24 08:54:59 CET 2008
46 - wrote a small introductory tutorial - Coro::Intro.
47 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
48 - Coro::Select did errornously dup the file descriptors
49 and didn't work with all AnyEvent backends.
50 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
51 LWP data transfers.
52 - disassociate c contexts from coro objects - this is agruably more
53 correct, but mostly allows sharing of cctxs between coro and state
54 objects, for added memory savings and speed increases.
55 - bumped $Coro::POOL_RSS up to 32kb by default.
56 - no longer set the optype to OP_CUSTOM, as B::* understandably
57 doesn't like this very much (and we *are* a type of entersub).
58 - implement state cloning, just to prove that call/cc can be done.
59 - automatically load Coro::AnyEvent in Coro::Handle.
60 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
61 do not have this method (and don't need it either).
62 - speed up generic anyevent methods in Coro::Handle by using rouse
63 callbacks.
64 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
65 others. It also makes the debugger happier, as you can trace
66 through the idle threads now.
67 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
68 - eg/event was pretty broken.
69 - better 5.8.6 compatibility.
70
71 5.0 Thu Nov 20 10:35:05 CET 2008
72 - NEW ARCHITECTURE: use the latest 4.x version if you experience
73 stability issues.
74 - bump API version to 7 - all dependents must be recompiled.
75 - removed timed_* functions - they were not being used anyways
76 and should be replaced by a more generic mechanism -
77 and were annoying to support anyways :)
78 - removed SemaphoreSet's waiter method - use sem method instead.
79 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
80 - async_pool did free a scalar value twice
81 ("Attempt to unreference...").
82 - fix a longstanding bug where calling terminate on a coro that
83 was waiting for a semaphore that was just becoming available
84 would cause a deadlock (semaphore would get into a state where
85 it was available but waiters were still blocked).
86 - calling throw on a coroutine that is waiting for a semaphore will
87 no longer make it acquire the semaphore (and thus leak a count).
88 - perl's process emulation is now not even theoretically supported
89 anymore.
90 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
91 conversion of callback-style to blocking-style.
92 - new methods $coro->schedule_to and ->cede_to, to specifically
93 schedule or cede to a specific coroutine.
94 - new function Coro::Semaphore::wait.
95 - use named constants in Coro::Channel (Richard Hundt).
96 - directly patch the entersub opcode calling SLF functions (cede,
97 transfer and so on). this does speed up context switching, but
98 more importanly, it frees us from the hardcoded behaviour of
99 entersub, so we might actually be able to return something from
100 those functions and atcually create new ones.
101 - take advantage of __builtin_frame_address on gcc.
102 - expose THX in coroapi (not sure whether this was a wise decision,
103 as "threaded" perls are running at half speed anyways).
104 - implement execute_slf (schedule-like-function) interface that makes
105 it possible to implement schedule-like-functions in XS.
106 - use new SLF interface to massively speed up Coro::EV by roughly a
107 factor of two.
108 - used new SLF interface to massively speed up Coro::Semaphore by a
109 factor of three.
110 - used new SLF interface to speed up Coro::AIO by roughly a factor of
111 four and reduce its memory usage considerably.
112 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
113 for a nice speedup and vastly more correct behaviour. Also implement
114 a new method "sem" to get at the underlying semaphore object.
115 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
116 (in comparison) 20-40% speedup.
117 - used new SLF interface to reimplement Coro::Signal gaining
118 some unknown (because I was too lazy), but certain, speedup, and also
119 making signals reliable for the first time.
120 - used new SLF interface and other optimisations to speed up async_pool
121 by a factor of two. It also doesn't rely on perl's exception mechanism
122 to exit anymore. The overhead for terminating an async_pool, coro over
123 a normal async is now very small.
124 - sped up coroutine creation/destruction by 40%.
125 - forgot to include Coro/libcoro/README in the dist for all these years.
126 - work around a freebsd pthreads bug (manual testcancel is required as
127 pthread_cond_wait isn't a cancellation point on freebsd).
128 - use new rouse functions to speed up and simplify Coro::BDB.
129 - make "prefer perl native functions" work with threaded perls.
130 - condense Coro::Debug ps output, hint at v and w flags.
131 - (libcoro) lots of minor cleanups and portability improvements.
132
133 4.914 Wed Nov 19 12:54:18 CET 2008
134 - fix a disastrous bug in the readline optimisation
135 introduced in 4.801.
136
137 4.913 Sat Nov 15 07:58:28 CET 2008
138 - async_pool did free a scalar value twice
139 ("Attempt to unreference...").
140
141 4.912 Thu Nov 13 18:31:23 CET 2008
142 - minor cleanups.
143 - use much larger stacks on linux and perl < 5.8.8.
144 - Coro::Debug::new_unix_server did not unlink the socket
145 when destroyed.
146
147 4.911 Tue Nov 11 04:26:17 CET 2008
148 - "port" to threaded perls.
149
150 4.91 Mon Nov 10 05:36:38 CET 2008
151 - the ->throw exception object no longer leaks.
152 - creating a new cctx leaked a scopestack entry (memleak).
153 - new coroutines didn't get created with a zero flags field
154 (unknown impact).
155 - calling ->throw on a not-yet-started coroutine should now work
156 instead of being ignored.
157 - ->throw is now supported on Coro::State objects.
158 - clean up cctx creation code a bit.
159 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
160
161 4.9 Sat Nov 8 17:45:27 CET 2008
162 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
163 - (libcoro) no longer leak threads in the experimental pthread backend,
164 also speed it up considerably.
165 - (libcoro) do not rely on makecontext passing void *'s unscathed.
166 - fix compiletime dependencies on libcoro in the Makefile.
167 - cctx_count wasn't always updated properly.
168 - Coro::State::cctx_stacksize wasn't applied correctly.
169 - new function Coro::State::cctx_max_idle.
170 - the default max number of idle C contexts is now 4.
171 - (libcoro) try harder to get _setjmp/_longjmp.
172 - (libcoro) cleanup and extend the libcoro API to officially
173 allow the creation of empty source contexts.
174 - very experimental workaround for the totally broken netbsd platform.
175 - tried to hack around openbsd bugs.
176
177 4.804 Wed Nov 5 16:36:00 CET 2008
178 - Coro::Debug::new_unix_server would not create a non-blocking listening
179 socket, sometimes causing freezes.
180 - (libcoro) fix misaligned stack points for setjmp and assembly
181 methods, which can cause crashes on x86/x86_64 with a sufficiently
182 aggressive compiler.
183 - new function: Coro::Debug::new_tcp_server.
184 - move ->throw into the Coro class because it only works on coro objects.
185
186 4.803 Mon Nov 3 17:16:12 CET 2008
187 - (libcoro) use a global asm statement to become independent of gcc
188 otpimisations for CORO_ASM (thanks to pippijn for the idea).
189 - try to workaround yet another broken bsd, this time dragonfly.
190
191 4.802 Thu Oct 30 10:56:12 CET 2008
192 - support -fno-omit-frame-pointer on x86 with the assembly method.
193 - tune 01_unblock.t tests a bit.
194
195 4.801 Wed Oct 22 18:33:37 CEST 2008
196 - improve readline speed for very long "lines".
197 - backport to 5.8.8.
198
199 4.8 Thu Oct 2 13:34:40 CEST 2008
200 - new function Coro::AIO::aio_wait.
201 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
202 - greatly speed up and reduce memory usage of Coro::AIO requests.
203 - implement some other ยต-optimisations.
204
205 4.749 Mon Sep 29 14:40:12 CEST 2008
206 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
207
208 4.748 Sat Sep 27 14:03:19 CEST 2008
209 - implement, but do not document, PerlIO::cede(granularity).
210 - Coro::Storable forgot to wrap Storable::pstore.
211 - work around the multitude of leaks and memory corruption
212 bugs in PerlIO::via by using our own C-level perliol. As a side
213 effect, Coro::Storable is now much, much, much faster.
214
215 4.747 Tue Sep 23 01:59:41 CEST 2008
216 - fix a per-cv memleak (one empty array was leaked per
217 code reference).
218 - avoid a crash in coro->call|rss when the coroutine was already
219 destroyed (most noticably when using Coro::Debug::ps :)
220 - also protect *Storable::FILE.
221 - push up default storable granularity to 20ms.
222
223 4.746 Sun Sep 21 03:22:20 CEST 2008
224 - be more insistent on locking Storable against reentrancy
225 in Coro::Storable.
226 - move swap_def?v and throw to Coro::State, as documented.
227
228 4.745 Thu Jul 24 00:14:38 CEST 2008
229 - remove debugging code related to MgPV_nolen_const, also try to
230 make it compile with perl 5.8.6 (yes, apple apparently loves
231 outdated software). Reported by John S.
232
233 4.744 Tue Jul 8 22:06:35 CEST 2008
234 - correctly provide default DIE/WARN handlers as documented.
235 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
236 implementations inside perl work fine for us.
237
238 4.743 Mon Jun 16 00:21:57 CEST 2008
239 - when using Coro::EV without running EV::loop it could
240 result in busy-waiting for events, this has been fixed.
241 - reduce codesize and improve performance by using EV_DEFAULT_UC.
242
243 4.742 Sat May 31 14:10:21 CEST 2008
244 - implement a workaround for (some) perl <5.8.8 versions.
245 - require EV 3.3+.
246
247 4.741 Fri May 30 23:33:09 CEST 2008
248 - tell netbsd how utterly broken their imitation of an OS is
249 and refuse to build by default if pthreads are in use.
250 - switch to "s" method on all bsd's by default, as their ucontext
251 stuff seems just too broken.
252 - fix a bug in Coro::Select.
253
254 4.74 Thu May 29 20:05:31 CEST 2008
255 - do not test Coro::LWP for lack of dependencies.
256 - do not test Coro::Select for lack of working perls.
257
258 4.73 Thu May 29 2008
259 - fix a bug in Coro::EV which would cause it to block despite
260 there being runnable coroutines.
261 - sprinkle "no warnings" freely over everything, also suppress
262 warnings for some other modules.
263 - fix typo in WSAEWOULDBLOCK.
264
265 4.72 Sun May 25 05:14:36 CEST 2008
266 - tweak META.yaml a bit, unfortunately, there is no documented way
267 to have optional dependencies with CPAN. doh :(
268 - avoid running some tests on windows because they would fail due to
269 perl bug (broken fork, broken pipes...).
270 - work around perl on windows bugs where perl returns undocumented
271 error codes for sysread, syswrite etc. by taking advantage
272 of AnyEvent's workaround for that problem.
273 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
274 a common perl implementation bug on windows.
275 - use unix domain sockets in testsuite to work around
276 common perl implementation bugs on widows (they are emulated by
277 tcp sockets on windows. ugh.)
278
279 4.71 Sat May 24 20:01:27 CEST 2008
280 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
281 - take advantage of async name resolution of AnyEvent::Util.
282 - work around brutal inet_aton override in Coro::LWP.
283 - take advantage of the readyhook in Coro::EV, for smoother
284 scheduling.
285
286 4.7 Sun May 11 00:32:19 CEST 2008
287 - completely reworked the Coro manpage.
288 - added Coro::AnyEvent, generic event loop integration.
289 - implement cancel, ready and kill commands in Coro::Debug.
290 - document find_coro in Coro::Debug.
291 - incompatible change: rename has_stack to has_cctx.
292 - Coro::AIO and Coro::BDB no longer force event model detection,
293 use AnyEvent::AIO and AnyEvent::BDB.
294
295 4.6 Sat Apr 26 10:05:14 CEST 2008
296 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
297 by the Coro module. It was a mistake to have it in the first place.
298 - (experimental) support for activestate perl 5.10 (method "w").
299 - (experimental) support for strawberry perl (method "a").
300 - coro_sigelem_set did not return a value although it had to,
301 actual impact unknown.
302
303 4.51 Mon Apr 14 13:28:27 CEST 2008
304 - make it compile again on 5.8.
305
306 4.50 Thu Apr 10 09:43:17 CEST 2008
307 - I did it twice! (see 4.49).
308
309 4.49 Sun Apr 6 21:23:31 CEST 2008
310 - grr, instead of compiling the recent changes on 5.10 only they
311 were compiled on 5.8 only.
312
313 4.48 Sun Apr 6 20:36:46 CEST 2008
314 - allow coroutine switches during eval's under 5.10.x, as it seems
315 the parser is a per-interpreter option now, so this is safe
316 (this might fix the odd crashing bug).
317 - drop support for 5.9.x versions: they are dead, jim.
318
319 4.47 Sun Apr 6 00:37:52 CEST 2008
320 - force cctx allocation on API calls: we know we need to force one
321 and gcc actually manages to confuse our heuristic nowadays,
322 leading to crashes and worse.
323 - document force_cctx.
324
325 4.46 Fri Apr 4 22:05:43 CEST 2008
326 - upgrade libcoro, resulting in pthread-backend (which was only created
327 to fulfill the rules of the programming languages shootout).
328
329 4.45 Thu Mar 13 11:55:36 CET 2008
330 - fix a file leaking bug in eg/mhyttpd that would allow
331 downloading of any file (reported by oesi).
332 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
333 (also add testcase).
334 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
335 apparently having loads of fun with that).
336 - implement $state->swap_defsv, ->swap_defav and document ->throw.
337
338 4.4 Wed Feb 13 16:44:29 CET 2008
339 - only assume .cfi pseudo isns exist on GNU/Linux.
340 - add get_timed method to Coro::Channels.
341 - fixed Coro::Debug manpage.
342 - perl 5.11 compatbility improvement based on patch by
343 Andreas Kรถnig.
344
345 4.37 Sun Jan 20 11:25:23 CET 2008
346 - soften the check for an active parser for perl 5.10 (avoiding
347 "Coro::State::transfer called while parsing" in many cases).
348
349 4.36 Sun Jan 13 10:53:56 CET 2008
350 - reset diehook when terminating from an async_pool as to not
351 trigger any __DIE__ handlers.
352
353 4.35 Sun Jan 13 04:14:13 CET 2008
354 - "bt" debug command now displays any exceptions
355 from longmess and also skips the innermost
356 stackframes, giving more useufl output.
357 - allow backtraces in coroutines blocked in Coro::EV,
358 at a <1% speed hit.
359 - handle localising of $SIG{__WARN__} and __DIE__
360 properly (with a proper amount of dirty hacking).
361
362 4.34 Sat Dec 22 17:49:53 CET 2007
363 - upgrade to EV version 2.0 API.
364
365 4.33 Mon Dec 17 08:36:12 CET 2007
366 - make Coro::AIO etc. loadable in the absence of EV.
367
368 4.32 Mon Dec 17 07:46:02 CET 2007
369 - majorly improved Coro::Handle's performance with Coro::EV.
370 - implemented the until now mythical Coro::BDB module.
371 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
372
373 4.31 Wed Dec 5 12:32:39 CET 2007
374 - remove warn statement form Coro::Util.
375
376 4.3 Tue Dec 4 20:33:14 CET 2007
377 - calls to the idle function could cause stack corruption
378 when the stack changed.
379 - do no longer rely on the presence of EV::DNS if EV is used
380 (because it is gone), but instead take avdantage of EV::ADNS
381 if available.
382 - add ($) prototypes to all functions in Coro::Storable.
383 - use a conventional (safer) idle callback in Coro::EV.
384 - do accept quickly in Coro::Debug to avoid endless loops.
385
386 4.22 Fri Nov 30 16:04:04 CET 2007
387 - really use optimised versions for Event and EV in Coro::Util
388 and Coro::Handle.
389
390 4.21 Sun Nov 25 10:48:59 CET 2007
391 - fix a spurious memory read.
392 - Coro::EV no longer keeps the eventloop "alive".
393
394 4.2 Fri Nov 9 20:47:05 CET 2007
395 - enable/disable tracing from a new coroutine, not a pooled one.
396 - fix a memleak in Coro::Event.
397 - removed killall call from fork_eval.
398 - made sure store_fd is already loaded so that fork_eval
399 does not have to parse autoload in each subprocess.
400 - only use assembly method if -O switch is in $Config{optimize}.
401 - add (optional) Coro::EV module, so far the best event loop module
402 directly supported by Coro.
403 - if the event model is EV, use EV::DNS to resolve
404 stuff in Coro::Util.
405 - don't get confused by multiple event notifications in Coro::Handle.
406 - initial support for EV (libevent interface).
407 - require Event and EV using configure_requires, to force their existance.
408
409 4.13 Wed Oct 24 07:26:45 CEST 2007
410 - add Coro::Storable::blocking_thaw.
411 - use a vastly more complicated technique to localise
412 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
413 - use a coroutine for the idle callback Coro::Event,
414 instead of running Event in the current coroutine context.
415 This also catches recursive invocations.
416 - actually report fork errors in gethostbyname and inet_aton.
417
418 4.11 Thu Oct 11 02:40:24 CEST 2007
419 - port to threaded perls.
420
421 4.1 Thu Oct 11 02:38:16 CEST 2007
422 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
423 be local to each coro (see Coro::State).
424 - incompatible change: for very deep reasons, cede and cede_notself
425 cannot return anything, so nothing will be returned.
426 - possibly bring back 5.10 compatibility (untested).
427 - work around stupid (and wrong) warning on 5.10 :(.
428 - overlay the saved state over the context stack. This saves
429 a few hundred bytes per coroutine on average and also
430 speeds up context switching a bit.
431 - further tune default stack sizes.
432 - (more) correctly calculate stack usage in coro_rss.
433 - Coro::Storable::blocking_* did not properly lock
434 resulting in races between coroutines.
435 - added Coro::Storable::guard.
436 - stopping to trace a coroutine could destroy the cctx of
437 an unrelated coroutine.
438 - explain the relationship between Perl and C coroutines in
439 more detail in Coro::State.
440 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
441 causing a fork per resolve. This also affected Coro::Socket.
442 - switch to a separate stack in $coro->call/eval to avoid
443 invalidating pointers.
444
445 4.03 Sat Oct 6 21:24:00 CEST 2007
446 - added Coro::throw method.
447 - minor code cleanups.
448
449 4.02 Sat Oct 6 02:36:47 CEST 2007
450 - fix a very minor per-coroutine memleak (a single codereference).
451 - fixed a bug where the currently in-use c context would be freed
452 prematurely (can happen only when programs change the stacksize
453 or use tracing).
454 - tracing can no longer keep a coro alive after it terminated.
455 - do static branch prediction in the common path for gcc. gives
456 about 2-5% speed improvement here.
457
458 4.01 Fri Oct 5 22:10:49 CEST 2007
459 - instead of recreating *a* standard output handle we simply
460 use STDOUT, which is faster and hopefully more robust.
461
462 4.0 Fri Oct 5 12:56:00 CEST 2007
463 - incompatibly changed Coro::Storable::freeze.
464 - major new feature: added Coro::Debug, for interactive coroutine
465 debugging, tracing and much more.
466 - major bug fix: unbelievable, but true: $_, $/ and many other
467 "saved" variables actually weren't being saved. This has been fixed,
468 of course, while increasing performance while losing all the save
469 flags.
470 - save flags are gone, and all the api functions dealing with them.
471 - added Coro::Semaphore::adjust.
472 - added Coro::Util::fork_eval.
473 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
474 - added Coro::killall.
475 - reduce initial stack sizes to allow for "micro-coroutines".
476 - better async_pool resource management, moved parts of async_pool
477 handling to XS (major speed improvement).
478 - actually croak before modifying important data structures.
479 - refuse to transfer while compiling.
480 - possibly support eval EXPR better now.
481 - enable assembly per default on linux+bsd x86+amd64.
482 - all internal members were renamed _something for easier subclassing.
483 - many minor tweaks.
484
485 3.63 Wed May 16 14:10:06 CEST 2007
486 - implement handcoded assembly for x86/amd64 SVR ABI.
487
488 3.62 Fri Apr 27 21:36:06 CEST 2007
489 - upgrade libcoro (which might set unwind info correctly).
490 - change default on linux to setjmp/longjmp.
491
492 3.61 Thu Apr 19 12:36:18 CEST 2007
493 - Coro::Storable caused an endless loop when thawing invalid
494 pst's sometimes.
495 - use a Semaphore in Coro::Storable, as Storable doesn't
496 seem to be reentrant (although it is documented to
497 be threadsafe...).
498 - fix Coro::Signal to bring back the original unreliable
499 but stateful semantics.
500 - fixed a lot of typos in Coro.pm (patch submitted by David
501 Steinbrunner, which applied flawlessly).
502
503 3.6 Sat Apr 14 17:13:31 CEST 2007
504 - added some bugfixes to get eg/myhttpd working again.
505 - added Coro::Storable for often-cede'ing freeze/thaw.
506 - try to do a clean exit when a coroutine calls exit
507 (EXPERIMENTAL).
508 - got rid of indirect call through _coro_init.
509 - updated the partly antique examples in eg/ to
510 work again and be a bit less magic, too.
511 - fixed Coro::Signal semantics to work as documented again.
512
513 3.55 Sun Mar 25 01:20:47 CET 2007
514 - add SAVE_DEFFH to save the default filehandle and enable
515 it by default.
516 - finally move socket-operations from Coro::Socket to Coro::Handle
517 to be able to unblock foreign sockets.
518 - add Coro::State::save_also and guarded_save.
519 - add count accessor to Coro::Semaphore.
520 - add Coro::State::cctx_stacksize.
521 - just for the fun of it, do not rely on implicit context,
522 which can dramatically improve performance, but people
523 using windows-process-emulation perls do not care much about
524 performance.
525
526 3.51 Sun Mar 4 14:18:04 CET 2007
527 - fixed a problem when you weakened references to Coro::State's
528 (patch by Yuval Kogman).
529
530 3.501 Wed Feb 28 12:44:07 CET 2007
531 - rename some global symbols as macosx from hell redefines
532 them without asking.
533
534 3.5 Tue Feb 13 20:22:53 CET 2007
535 - do AnyEvent model detection earlier, avoiding problems
536 caused by first using AnyEvent and later Coro::Event.
537 - implement and document Coro::Event event objects.
538 - fix a potential problem in Coro::Event causing crashes.
539 - initialise PL_comppad when creating a new coroutine,
540 avoids crashes on early coro destruction.
541
542 3.41 Mon Jan 22 19:19:49 CET 2007
543 - readline on Coro::Handle did not support undefined $/,
544 nor did it deliver partial lines on EOF or error.
545 - implement malloc fallback for stack allocation because
546 stupid broken idiotic OSX has a stupid broken
547 idiotic fits-the-whole-os mmap "implementation" and
548 my dick feels longer if Coro is portable even to
549 obsolete platforms.
550
551 3.4 Fri Jan 19 21:52:54 CET 2007
552 - remove t/09_timer.t, as it isn't really testing much
553 but was rather flaky in practise.
554 - async_pool coro would keep arguments and callback alive until
555 it was reused.
556 - cancellation of a coroutine could cause spurious idle calls
557 in cede_notself.
558
559 3.3 Sat Jan 6 03:45:00 CET 2007
560 - implement $coro->on_destroy.
561 - Coro::Event blocking semantics have been changed,
562 documented and - hopefully - improved.
563 - fix nice adding, not subtracting, from priority.
564 - fix ->prio and api_is_ready (patch by Mark Hinds).
565 - fixed an assert ("... == prev__cctx->idle_te")
566 that could errronously trigger.
567 - fix various large and small memleaks.
568 - use a (hopefully) more stable cancel implementation
569 that immediately frees the coroutine data.
570 - cede/cede_notself return a status now.
571 - added Coro::guard function.
572 - added a global coroutine pool for jobs (on my machine,
573 I can create and execute 48k simple coros/s with async,
574 and 128k coros with async_pool).
575 - Coro::AIO now uses the coroutine priority as io priority.
576
577 3.2 Fri Dec 22 05:07:09 CET 2006
578 - improve portability to slightly older perls.
579 - use cleaner coroutine destruction.
580 - simplify configuration for users.
581 - optionally (unrecommended) prefer perl functions over
582 their coro replacements.
583
584 3.11 Tue Dec 5 13:11:24 CET 2006
585 - fixed some bogus assert's, but as perl.h disables assert even
586 without NDEBUG (thank you very much), not too many people should
587 notice (that did include myself). Andreas Kรถnig noticed, though :)
588 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
589 it. Also noticed by Andreas Kรถnig :)
590 - save/restore tainted status.
591 - verified to pass the testsuite on my 5.9.5.
592
593 3.1 Mon Dec 4 23:03:40 CET 2006
594 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
595 - incompatible change: transfer flags are now per-state.
596 - give a better error message on deadlock.
597 - document Coro::nready.
598 - enhanced testsuite.
599
600 3.01 Sun Dec 3 23:47:42 CET 2006
601 - forgot to include Coro::Timer.
602
603 3.0 Sun Dec 3 22:57:25 CET 2006
604 - the "FINALLY COMPLETELY STABLE" release (coming soon:
605 the "FAMOUS LAST WORDS" release).
606 - implement a new stack sharing algorithm, which uses a stack
607 pool (size currently hardcoded).
608 - make stack sharing mandatory (it no longer uses a heuristic).
609 - eval/die no longer cause weird problems under heavy use.
610 - Coro::Event could cause livelocks if it was used but
611 no Coro::Event watchers were used.
612 - Coro::Event now uses asynccheck as prepare does not
613 check for changed watchers.
614 - Coro::Event allows multiple waiting coros and will wake up one
615 per event.
616 - Coro::Event should be cleaner and more efficient now.
617 - new utility function Coro::unblock_sub.
618 - document the sad fact that Event is no longer reentrant.
619 - putting a coroutine into the ready queue twice could under
620 some circumstances lead to stack corruption.
621 - minor incompatibility: subclassing Coro::State is supported
622 directly now without going through a _coro_state member.
623 - state/coro switching is much faster now.
624 - very minor optimisations and code/documentation cleanup.
625 - avoid problems due to compiler inlining.
626 - removed timers from Coro::Timer -> use AnyEvent instead.
627 - replaced Coro::idle coroutine by (cleaner) idle handler.
628 - updated to newest libcoro.
629 - implement enhanced support for valgrind.
630 - implement is_ready and return value for ready.
631 - removed Coro::Cont, it was a misnomer (it's generators),
632 and it was somewhat annoying to get it right. Will come back
633 if somebody asks for it :)
634 - many ->wait methods and Coro::Event could return
635 spuriously without the event having happened.
636
637 2.5 Tue Nov 7 12:22:33 CET 2006
638 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
639 use AnyEvent, moved them to Coro/.
640 - added Coro::LWP which contains all the uglyness required to
641 make LWP non-blocking.
642 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
643 - fixed another bug in Coro::Select when the timeout was undef.
644 - reuse PL_start_env for all coros, saves some memory per coroutine.
645 - manage PL_top_env differently, hopefully to avoid panic: top_env.
646 - timeout argument was not properly used in Coro::Socket.
647 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
648 - emulate undocumented(!) functionality of IO::Socket required
649 by LWP(!!).
650 - updated eg/lwp to work with newer lwp's.
651 - remove "FATAL: uncaught exception" prefix. Coroutines that die
652 kill the whole process, just as exceptions in the main "coroutine"
653 did already.
654
655 2.1 Wed Nov 1 23:01:13 CET 2006
656 - fix a long-standing bug in Coro::Select where select with
657 zero timeout would instead change the current default filehandle.
658 - use a simpler and hopefully more robust way to clone padlists
659 (uses less memory and a perl function instead of our own).
660 - coro can now create a stack guard area on many architectures.
661 - Coro::AIO properly reexports additional functions from IO::AIO.
662 - updated libcoro with a workaround for OS X,
663 pach and testing by Michael Schwern.
664
665 2.0 Tue Oct 24 05:47:17 CEST 2006
666 - support additional aio requests in Coro::AIO.
667
668 1.9 (never properly released due to a glitch)
669
670 1.8 Thu Feb 2 00:59:06 CET 2006
671 - applied suggested patch by SAMV to avoid problems during stupid
672 mark & sweep gc run.
673 - applied patch by Scott Walters for 5.9.3 compatibility.
674
675 1.7 Tue Dec 27 01:41:58 CET 2005
676 - added Coro::AIO, a thin wrapper around IO::AIO.
677 - improved Makefile.PL explanations.
678
679 1.6 Mon Dec 12 21:30:05 CET 2005
680 - additionally save PM_curpm might fix as-of-yet
681 unseen problems with regex matches being attributed to the
682 wrong package.
683 - add t/10_bugs.t, which currently checks against imho broken
684 perls that use 0.26MB of stack space per Perl_magic_get
685 invocation (newer linuxes) as opposed to the 0.0002MB perl
686 normally uses.
687 - make stacksize configurable for the ultimate debian experience.
688
689 1.51 Mon Dec 12 18:48:36 CET 2005
690 - remove debugging warn() accidentally left in Coro::Select.
691
692 1.5 Tue Nov 29 13:32:44 CET 2005
693 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
694 - fix Coro::Select 'select'.
695 - strict'ify some modules.
696 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
697 - use XSLoader in selected modules.
698 - remove some 5.6 compatibility cruft.
699
700 1.4 Tue Sep 6 00:11:05 CEST 2005
701 - libcoro did not take into account the trampoline on amd64,
702 when 'l'inux method was used.
703
704 1.31 Tue Aug 30 23:31:33 CEST 2005
705 - some portability fixes/workarounds.
706
707 1.3 Sat Aug 20 03:08:56 CEST 2005
708 - no code changes. module seems to work fine.
709 - improve Coro::State docs, remove reference to nonexisting
710 Coro::State::flush method.
711 - no longer autodetect windows, present it as an option instead.
712
713 1.2 Mon May 16 02:00:55 CEST 2005
714 - included libcoro.c earlier in State.xs, might improve portability.
715 - use a faster and possibly more stable Coro::Cont implementation.
716 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
717 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
718 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
719
720 1.11 Thu Mar 3 18:00:52 CET 2005
721 - change of contact address.
722
723 1.1 Tue Feb 22 20:51:16 CET 2005
724 - support [l]inux method on amd64.
725 - allow some context switches while compiling/eval'ing.
726
727 1.0 Tue Aug 10 03:47:27 CEST 2004
728 - Event 0.89 fixes the bug again.
729 - fix the remaining(??) memleak. I didn't free the cache-AV for
730 padlists, so one AV per created anon-sub leaked.
731
732 0.97 Fri May 14 15:23:32 CEST 2004
733 - remove stupid gcc extension that was an experiment only anyway.
734
735 0.96 Thu May 13 18:09:29 CEST 2004
736 - tell people if they have yet another broken Event variant.
737 - made Coro to work _better_ with threads (it seems that, as long
738 as you use Coros in one thread only everything will work fine.
739 Looking at the perl source the problem might be missing locking
740 between Coro and some perl internal routines. That's not easily
741 fixable).
742 - (hopefully) got rid of the static xnext inside transfer.
743
744 0.95 Sun Feb 15 03:20:28 CET 2004
745 - removed Coro::State::flush, implemented a new and much
746 faster caching mechanism by attaching magic to the CV.
747 - a memleak with (real) closures remains.
748 - dramatically reduced size of stacks to 65536 (linux/x86), after
749 observing that even myhttpd never had more than a kilobyte
750 of stack in use (stack is still only physically allocated in
751 pagesize increments for systems suporting mmap).
752 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
753 - fix for threaded perl.
754
755 0.9 Sun Nov 23 04:32:39 CET 2003
756 - fix a few very hard to track down but (of course) stupid
757 memory leaks.
758 - stringify version number to avoid locale problems :(.
759 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
760 hopefully ignoring it "just works".
761
762 0.8 Wed Nov 5 19:38:40 CET 2003
763 - port to perl5.005_03, but only 5.8.x+ is supported!
764 - honor LocalAddr even without LocalPort in Coro::Socket.
765
766 0.7 Tue May 27 03:12:38 CEST 2003
767 - the version jump indicates some level of testing, not gobs
768 of new features.
769 - uh... I found the prompt function in ExtUtils::MakeMaker.
770 Highly correct stuff, that is...
771 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
772 the fix is not well understood by the author, so beware :(.
773
774 0.652 Thu May 8 02:54:46 CEST 2003
775 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
776 cause version 4 doesn't have ucontext.
777 - Benjamin Reed reported that setjmp works fine on darwin,
778 so preselect it.
779
780 0.651 Sat Mar 29 15:00:23 CET 2003
781 - fix a bug in Coro::Handle where some bytes could get lost
782 on reading, reported by jason@nichego.net.
783
784 0.65 Sun Mar 23 00:08:26 CET 2003
785 - added a README.linux-glibc.
786 - new module Coro::Select.
787 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
788 (I never claimed that I know what I am doing ;).
789 - set default to setjmp/longjmp on non-x86-linux.
790
791 0.6 Thu Nov 21 11:09:06 CET 2002
792 - made lazy context switching the default.
793 - fixed the bug where SemaphoreSet::guard
794 was the same as timed_guard and timed_guard was missing.
795 - fix a memleak in Coro::Timer.
796 - reclassified context sharing as not experimental.
797
798 0.534 Sun Apr 14 03:05:12 CEST 2002
799 - fixed backspace => tab in header parsing (myhttpd).
800 - added eg/readline, for elmex.
801 - Coro::Event now calls ->start on first call instead of
802 ->again, which created wrong timeouts for the first ->next.
803 - fixed a bug where Coro::Socket returned a closed socket instead
804 of nothing on ECONNREFUSED and similar error conditions.
805
806 0.533 Mon Feb 18 18:49:40 CET 2002
807 - adapt to finally working Event-0.85.
808 - fixed buggy event initialization (trapped under windows, but how
809 could it possibly have worked anywhere?)
810
811 0.532 Wed Jan 16 02:45:32 CET 2002
812 - removed the reference to pp_entersub, might work on
813 (some) win32 perls now (testsuite works under cygwin,
814 without Event, which freezes).
815
816 0.531 Mon Dec 10 22:18:44 CET 2001
817 - Fixed a bug in SemaphoreSet::timed_down.
818
819 0.53 Tue Nov 27 21:11:13 CET 2001
820 - some tests for Event.
821 - slightly different internal architecture to get refcnt into the object
822 and not the reference. might fix a few bugs, certainly did introduce
823 new ones.
824 - Coro::Timer (independent of Event).
825 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
826
827 0.52 Tue Nov 6 21:36:18 CET 2001
828 - ported to cygwin (trivial).
829 patch by Gerrit P. Haase <gp@familiehaase.de>.
830 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
831
832 0.51 Thu Nov 1 20:39:01 CET 2001
833 - terminate/cancel now work properly (otherwise termination
834 could cause a "next coroutine is not and contains not..."
835 error.
836 - added Coro::Socket::shutdown.
837 - Coro::Event::loop no is the same as Event::loop.
838 - implemented terminate with args + join.
839
840 0.5 Fri Sep 28 16:15:35 CEST 2001
841 - fixed "print" on a Coro::Handle. The print method worked.
842 - small tweaks (seem to reduce memory consumption a lot)
843 in various modules.
844 - splendid use of "no warnings" scattered throughout.
845 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
846
847 0.49 Sun Sep 16 02:42:45 CEST 2001
848 - changed some method calls to function calls for speed
849 inside Coro::Handle.
850 - make Coro::Handle use an array instead of a hash for
851 speed reasons.
852 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
853 sjlj and ucontext should now work.
854 - IRIX-specific port for libcoro.
855 - swapped order of accept results to match IO::Socket.
856 - changed getsock/peername to sock/peername to match IO::Socket.
857 - fixed a bug that caused segfault when returning to main task
858 under some circumstances.
859 - other bugfixes.
860
861 0.45 Sun Sep 2 02:54:01 CEST 2001
862 - new method Coro::Handle::timeout.
863 - corrected speling of set...name to get...name in Coro::Socket.
864 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
865 - Coro::Handle now supports fileno and FILENO.
866 - added eg/myhttpd, a web server.
867 - Coro::Socket now accepts numeric ports as well as "port(number)"
868 syntax.
869 - moved some scheduling primitives into xs code. more to come.
870 - new simple priority system.
871 - implemented Coro::Event using XS, almost four times faster!
872 - small memory corruption problem fixed (boy that was difficult).
873
874 0.13 Wed Aug 8 16:53:07 CEST 2001
875 - new method Coro::Handle::readline.
876 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
877 (see eg/lwp on how to make LWP non-blocking).
878 - renamed Coro::Socket::new_inet to new.
879 - added Coro::Util, some utility functions.
880 - die/eval now works better (still get annoying (but true ;)
881 "Callback called exit" messages).
882 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
883 - much better Coro::Socket support.
884
885 0.12 Fri Jul 27 04:19:01 CEST 2001
886 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
887 - very experimental stack sharing algorithm. not 100% safe but
888 should work well in practise ;)
889 - added Coro::RWLock.
890 - Coro::Specific now works.
891
892 0.11 Tue Jul 24 22:49:21 CEST 2001
893 - added specialized hack for newer and older linux versions (fast).
894 - renamed Coro::Event::IO to Coro::Handle.
895 - new module Coro::Socket.
896
897 0.1 Tue Jul 24 01:47:53 CEST 2001
898 - release candidate 3. A rather new internal structure :(
899 - the great renaming: Damian Conway gave me a suitable replacement
900 for yield.
901 - added Coro/libcoro, a portable coroutine implementation for C
902 (not even perl-dependent!!!), which can be used by Coro::State.
903 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
904 perl-5.6.
905 - new function Coro::Event::idle.
906 - the idle process is now overriden
907 by default in Coro::Event.
908 - Coro::Channel now enforces the size.
909 - canceling events now works in all cases.
910 - Coro state now include $_ and $@.
911 - yet another bug workaround that I do not really understand :(
912 - new module Coro::Event::IO (very undocumented) to do
913 non-blocking i/o.
914 - performance tuning.
915
916 0.08 Thu Jul 19 06:13:25 CEST 2001
917 - release candidate 2 (new functionality).
918 - add Coro::State::flush function.
919 - transfer now supports a flags value (mostly for speed ATM).
920 - might compile and work in the presence of threads now.
921 - continuations have a saner syntax.
922 - no more memleaks.
923
924 0.07 Tue Jul 17 17:40:18 CEST 2001
925 - release candidate 1 ;)
926 - slightly nicer code.
927 - fixed a scheduling bug in Coro::Event.
928
929 0.06 Tue Jul 17 04:23:24 CEST 2001
930 - ok, I found the showstopper - the same sub must not be
931 re-entered in two different coroutines, otherwise => crash. I
932 see no easy solution to this problem, except by walking the call
933 chaing and saving/restoring the cv's, which is what I do now.
934 - memory leaks still latent, especially at thread termination.
935 - Coro::Event now works (haha).
936
937 0.05 Sun Jul 15 17:32:20 CEST 2001
938 - fixed a few issues in Coro::Event.
939 - I forgot to include Coro::Event in 0.04 :(:(:(
940
941 0.04 Sun Jul 15 05:24:59 CEST 2001
942 - @_ is now properly localized.
943 - Coro::State is now easier subclassable.
944 - Coro::Cont now coroutine-aware.
945 - Coro::Specific is a low-overhead module to create
946 coroutine-specific vars.
947 - Coro::Event provides a simple interface to Event.
948
949 0.03 Fri Jul 13 14:51:52 CEST 2001
950 - transfer() now implemented in XS (beware).
951 - new module Coro::Cont for really faked continuations.
952 - big internal architecture changes: Coro::State is now
953 really low-level and can thus be used to implement other
954 interesting things, While "Coro::" implements a process-like
955 model. Still crude and subject to change.
956 - $_ and $@ are no longer being localized.
957
958 0.02 Tue Jul 10 01:38:17 CEST 2001
959 - implemented "async" attribute.
960 - $_ and $@ are now localized.
961 - added Coro::Channel.
962 - more testcases, still no docs.
963
964 0.01 Tue Jul 3 02:18:41 CEST 2001
965 - original version; copied from Convert::Scalar.
966