ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.460
Committed: Sat Nov 22 16:37:11 2008 UTC (15 years, 5 months ago) by root
Branch: MAIN
Changes since 1.459: +2 -0 lines
Log Message:
*** empty log message ***

File Contents

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