ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.494
Committed: Wed Jun 17 21:36:35 2009 UTC (14 years, 11 months ago) by root
Branch: MAIN
Changes since 1.493: +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 TODO: unready_all
8 TODO: myhttpd header parsing
9 TODO: channel->maxsize(newsize)?
10
11 - speed up ready queue management/context switching by using a linked
12 list instead of an array (~5-10%).
13 - implement "watch" command in Coro::Debug shells.
14 - for fun, implement time-slicing as an example in the manpage.
15
16 5.132 Fri May 29 09:00:39 CEST 2009
17 - do not keep a reference to the argument itself in
18 Coro::Semaphore::guard, as it could change later.
19 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
20 should support a prepare callback.
21
22 5.131 Mon Mar 16 23:20:37 CET 2009
23 - implement and document Coro->suspend, Coro->resume.
24 - fix Coro::Select implementation to not (often) close
25 the passed file descriptors (testcase provided by pippijn).
26
27 5.13 Mon Dec 15 21:51:42 CET 2008
28 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
29 - don't set diehook to C<undef> but instead to NULL, to avoid
30 spurious warnings.
31 - fix a lot of bugs in Coro::SemaphoreSet.
32 - Coro::SemaphoreSet will less often create a semaphore needlessly.
33 - add Coro::SemaphoreSet::count and wait methods.
34 - take advantage of the new Guard module.
35 - deprecate Coro::guard.
36 - try to fix the dreaded 01_unblock tests once more. I hate it when
37 testsuites need more fixing than the code they are supposed to test.
38 - croak in more cases when a required callback isn't resolvable.
39 - fix some minor issues in Coro::State->call/eval.
40 - use current coroutine context instead of a temporary one
41 when temporarily switching to another coroutine.
42 - do not call C-level on_destroy handlers during global destruction,
43 to avoid needless segfaults.
44
45 5.12 Sun Dec 7 13:30:38 CET 2008
46 - add default config for MirOS, which seems to be a bug-to-bug
47 compatible fork of openbsd ("world domination by releasing
48 openbsd cvs before the openbsd folks do it" or so :).
49 - free_padlist did destroy the names pad, not good, but didn't
50 seem to bother perl - this could fix issues such as eval ""
51 inside a function called from multiple coroutines.
52 - use a different method to detect destruction time.
53 - be more careful when freeing padlists just before global
54 destruction.
55 - fixed and expanded the call/cc example.
56 - renamed _terminate to _coro_run.
57 - new method Coro::Channel->shutdown.
58 - try pthreads on openbsd <4.4 (broken sigaltstack, will
59 pthreads fare better?).
60 - be less picky about destroying "a" running coroutine.
61
62 5.11 Tue Nov 25 21:49:05 CET 2008
63 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
64 working code with failed assertions, introducing more bugs than
65 it fixes, requiring elaborate workarounds :(
66
67 5.1 Mon Nov 24 08:54:59 CET 2008
68 - wrote a small introductory tutorial - Coro::Intro.
69 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
70 - Coro::Select did errornously dup the file descriptors
71 and didn't work with all AnyEvent backends.
72 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
73 LWP data transfers.
74 - disassociate c contexts from coro objects - this is agruably more
75 correct, but mostly allows sharing of cctxs between coro and state
76 objects, for added memory savings and speed increases.
77 - bumped $Coro::POOL_RSS up to 32kb by default.
78 - no longer set the optype to OP_CUSTOM, as B::* understandably
79 doesn't like this very much (and we *are* a type of entersub).
80 - implement state cloning, just to prove that call/cc can be done.
81 - automatically load Coro::AnyEvent in Coro::Handle.
82 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
83 do not have this method (and don't need it either).
84 - speed up generic anyevent methods in Coro::Handle by using rouse
85 callbacks.
86 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
87 others. It also makes the debugger happier, as you can trace
88 through the idle threads now.
89 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
90 - eg/event was pretty broken.
91 - better 5.8.6 compatibility.
92
93 5.0 Thu Nov 20 10:35:05 CET 2008
94 - NEW ARCHITECTURE: use the latest 4.x version if you experience
95 stability issues.
96 - bump API version to 7 - all dependents must be recompiled.
97 - removed timed_* functions - they were not being used anyways
98 and should be replaced by a more generic mechanism -
99 and were annoying to support anyways :)
100 - removed SemaphoreSet's waiter method - use sem method instead.
101 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
102 - async_pool did free a scalar value twice
103 ("Attempt to unreference...").
104 - fix a longstanding bug where calling terminate on a coro that
105 was waiting for a semaphore that was just becoming available
106 would cause a deadlock (semaphore would get into a state where
107 it was available but waiters were still blocked).
108 - calling throw on a coroutine that is waiting for a semaphore will
109 no longer make it acquire the semaphore (and thus leak a count).
110 - perl's process emulation is now not even theoretically supported
111 anymore.
112 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
113 conversion of callback-style to blocking-style.
114 - new methods $coro->schedule_to and ->cede_to, to specifically
115 schedule or cede to a specific coroutine.
116 - new function Coro::Semaphore::wait.
117 - use named constants in Coro::Channel (Richard Hundt).
118 - directly patch the entersub opcode calling SLF functions (cede,
119 transfer and so on). this does speed up context switching, but
120 more importanly, it frees us from the hardcoded behaviour of
121 entersub, so we might actually be able to return something from
122 those functions and atcually create new ones.
123 - take advantage of __builtin_frame_address on gcc.
124 - expose THX in coroapi (not sure whether this was a wise decision,
125 as "threaded" perls are running at half speed anyways).
126 - implement execute_slf (schedule-like-function) interface that makes
127 it possible to implement schedule-like-functions in XS.
128 - use new SLF interface to massively speed up Coro::EV by roughly a
129 factor of two.
130 - used new SLF interface to massively speed up Coro::Semaphore by a
131 factor of three.
132 - used new SLF interface to speed up Coro::AIO by roughly a factor of
133 four and reduce its memory usage considerably.
134 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
135 for a nice speedup and vastly more correct behaviour. Also implement
136 a new method "sem" to get at the underlying semaphore object.
137 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
138 (in comparison) 20-40% speedup.
139 - used new SLF interface to reimplement Coro::Signal gaining
140 some unknown (because I was too lazy), but certain, speedup, and also
141 making signals reliable for the first time.
142 - used new SLF interface and other optimisations to speed up async_pool
143 by a factor of two. It also doesn't rely on perl's exception mechanism
144 to exit anymore. The overhead for terminating an async_pool, coro over
145 a normal async is now very small.
146 - sped up coroutine creation/destruction by 40%.
147 - forgot to include Coro/libcoro/README in the dist for all these years.
148 - work around a freebsd pthreads bug (manual testcancel is required as
149 pthread_cond_wait isn't a cancellation point on freebsd).
150 - use new rouse functions to speed up and simplify Coro::BDB.
151 - make "prefer perl native functions" work with threaded perls.
152 - condense Coro::Debug ps output, hint at v and w flags.
153 - (libcoro) lots of minor cleanups and portability improvements.
154
155 4.914 Wed Nov 19 12:54:18 CET 2008
156 - fix a disastrous bug in the readline optimisation
157 introduced in 4.801.
158
159 4.913 Sat Nov 15 07:58:28 CET 2008
160 - async_pool did free a scalar value twice
161 ("Attempt to unreference...").
162
163 4.912 Thu Nov 13 18:31:23 CET 2008
164 - minor cleanups.
165 - use much larger stacks on linux and perl < 5.8.8.
166 - Coro::Debug::new_unix_server did not unlink the socket
167 when destroyed.
168
169 4.911 Tue Nov 11 04:26:17 CET 2008
170 - "port" to threaded perls.
171
172 4.91 Mon Nov 10 05:36:38 CET 2008
173 - the ->throw exception object no longer leaks.
174 - creating a new cctx leaked a scopestack entry (memleak).
175 - new coroutines didn't get created with a zero flags field
176 (unknown impact).
177 - calling ->throw on a not-yet-started coroutine should now work
178 instead of being ignored.
179 - ->throw is now supported on Coro::State objects.
180 - clean up cctx creation code a bit.
181 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
182
183 4.9 Sat Nov 8 17:45:27 CET 2008
184 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
185 - (libcoro) no longer leak threads in the experimental pthread backend,
186 also speed it up considerably.
187 - (libcoro) do not rely on makecontext passing void *'s unscathed.
188 - fix compiletime dependencies on libcoro in the Makefile.
189 - cctx_count wasn't always updated properly.
190 - Coro::State::cctx_stacksize wasn't applied correctly.
191 - new function Coro::State::cctx_max_idle.
192 - the default max number of idle C contexts is now 4.
193 - (libcoro) try harder to get _setjmp/_longjmp.
194 - (libcoro) cleanup and extend the libcoro API to officially
195 allow the creation of empty source contexts.
196 - very experimental workaround for the totally broken netbsd platform.
197 - tried to hack around openbsd bugs.
198
199 4.804 Wed Nov 5 16:36:00 CET 2008
200 - Coro::Debug::new_unix_server would not create a non-blocking listening
201 socket, sometimes causing freezes.
202 - (libcoro) fix misaligned stack points for setjmp and assembly
203 methods, which can cause crashes on x86/x86_64 with a sufficiently
204 aggressive compiler.
205 - new function: Coro::Debug::new_tcp_server.
206 - move ->throw into the Coro class because it only works on coro objects.
207
208 4.803 Mon Nov 3 17:16:12 CET 2008
209 - (libcoro) use a global asm statement to become independent of gcc
210 otpimisations for CORO_ASM (thanks to pippijn for the idea).
211 - try to workaround yet another broken bsd, this time dragonfly.
212
213 4.802 Thu Oct 30 10:56:12 CET 2008
214 - support -fno-omit-frame-pointer on x86 with the assembly method.
215 - tune 01_unblock.t tests a bit.
216
217 4.801 Wed Oct 22 18:33:37 CEST 2008
218 - improve readline speed for very long "lines".
219 - backport to 5.8.8.
220
221 4.8 Thu Oct 2 13:34:40 CEST 2008
222 - new function Coro::AIO::aio_wait.
223 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
224 - greatly speed up and reduce memory usage of Coro::AIO requests.
225 - implement some other ยต-optimisations.
226
227 4.749 Mon Sep 29 14:40:12 CEST 2008
228 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
229
230 4.748 Sat Sep 27 14:03:19 CEST 2008
231 - implement, but do not document, PerlIO::cede(granularity).
232 - Coro::Storable forgot to wrap Storable::pstore.
233 - work around the multitude of leaks and memory corruption
234 bugs in PerlIO::via by using our own C-level perliol. As a side
235 effect, Coro::Storable is now much, much, much faster.
236
237 4.747 Tue Sep 23 01:59:41 CEST 2008
238 - fix a per-cv memleak (one empty array was leaked per
239 code reference).
240 - avoid a crash in coro->call|rss when the coroutine was already
241 destroyed (most noticably when using Coro::Debug::ps :)
242 - also protect *Storable::FILE.
243 - push up default storable granularity to 20ms.
244
245 4.746 Sun Sep 21 03:22:20 CEST 2008
246 - be more insistent on locking Storable against reentrancy
247 in Coro::Storable.
248 - move swap_def?v and throw to Coro::State, as documented.
249
250 4.745 Thu Jul 24 00:14:38 CEST 2008
251 - remove debugging code related to MgPV_nolen_const, also try to
252 make it compile with perl 5.8.6 (yes, apple apparently loves
253 outdated software). Reported by John S.
254
255 4.744 Tue Jul 8 22:06:35 CEST 2008
256 - correctly provide default DIE/WARN handlers as documented.
257 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
258 implementations inside perl work fine for us.
259
260 4.743 Mon Jun 16 00:21:57 CEST 2008
261 - when using Coro::EV without running EV::loop it could
262 result in busy-waiting for events, this has been fixed.
263 - reduce codesize and improve performance by using EV_DEFAULT_UC.
264
265 4.742 Sat May 31 14:10:21 CEST 2008
266 - implement a workaround for (some) perl <5.8.8 versions.
267 - require EV 3.3+.
268
269 4.741 Fri May 30 23:33:09 CEST 2008
270 - tell netbsd how utterly broken their imitation of an OS is
271 and refuse to build by default if pthreads are in use.
272 - switch to "s" method on all bsd's by default, as their ucontext
273 stuff seems just too broken.
274 - fix a bug in Coro::Select.
275
276 4.74 Thu May 29 20:05:31 CEST 2008
277 - do not test Coro::LWP for lack of dependencies.
278 - do not test Coro::Select for lack of working perls.
279
280 4.73 Thu May 29 2008
281 - fix a bug in Coro::EV which would cause it to block despite
282 there being runnable coroutines.
283 - sprinkle "no warnings" freely over everything, also suppress
284 warnings for some other modules.
285 - fix typo in WSAEWOULDBLOCK.
286
287 4.72 Sun May 25 05:14:36 CEST 2008
288 - tweak META.yaml a bit, unfortunately, there is no documented way
289 to have optional dependencies with CPAN. doh :(
290 - avoid running some tests on windows because they would fail due to
291 perl bug (broken fork, broken pipes...).
292 - work around perl on windows bugs where perl returns undocumented
293 error codes for sysread, syswrite etc. by taking advantage
294 of AnyEvent's workaround for that problem.
295 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
296 a common perl implementation bug on windows.
297 - use unix domain sockets in testsuite to work around
298 common perl implementation bugs on widows (they are emulated by
299 tcp sockets on windows. ugh.)
300
301 4.71 Sat May 24 20:01:27 CEST 2008
302 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
303 - take advantage of async name resolution of AnyEvent::Util.
304 - work around brutal inet_aton override in Coro::LWP.
305 - take advantage of the readyhook in Coro::EV, for smoother
306 scheduling.
307
308 4.7 Sun May 11 00:32:19 CEST 2008
309 - completely reworked the Coro manpage.
310 - added Coro::AnyEvent, generic event loop integration.
311 - implement cancel, ready and kill commands in Coro::Debug.
312 - document find_coro in Coro::Debug.
313 - incompatible change: rename has_stack to has_cctx.
314 - Coro::AIO and Coro::BDB no longer force event model detection,
315 use AnyEvent::AIO and AnyEvent::BDB.
316
317 4.6 Sat Apr 26 10:05:14 CEST 2008
318 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
319 by the Coro module. It was a mistake to have it in the first place.
320 - (experimental) support for activestate perl 5.10 (method "w").
321 - (experimental) support for strawberry perl (method "a").
322 - coro_sigelem_set did not return a value although it had to,
323 actual impact unknown.
324
325 4.51 Mon Apr 14 13:28:27 CEST 2008
326 - make it compile again on 5.8.
327
328 4.50 Thu Apr 10 09:43:17 CEST 2008
329 - I did it twice! (see 4.49).
330
331 4.49 Sun Apr 6 21:23:31 CEST 2008
332 - grr, instead of compiling the recent changes on 5.10 only they
333 were compiled on 5.8 only.
334
335 4.48 Sun Apr 6 20:36:46 CEST 2008
336 - allow coroutine switches during eval's under 5.10.x, as it seems
337 the parser is a per-interpreter option now, so this is safe
338 (this might fix the odd crashing bug).
339 - drop support for 5.9.x versions: they are dead, jim.
340
341 4.47 Sun Apr 6 00:37:52 CEST 2008
342 - force cctx allocation on API calls: we know we need to force one
343 and gcc actually manages to confuse our heuristic nowadays,
344 leading to crashes and worse.
345 - document force_cctx.
346
347 4.46 Fri Apr 4 22:05:43 CEST 2008
348 - upgrade libcoro, resulting in pthread-backend (which was only created
349 to fulfill the rules of the programming languages shootout).
350
351 4.45 Thu Mar 13 11:55:36 CET 2008
352 - fix a file leaking bug in eg/mhyttpd that would allow
353 downloading of any file (reported by oesi).
354 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
355 (also add testcase).
356 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
357 apparently having loads of fun with that).
358 - implement $state->swap_defsv, ->swap_defav and document ->throw.
359
360 4.4 Wed Feb 13 16:44:29 CET 2008
361 - only assume .cfi pseudo isns exist on GNU/Linux.
362 - add get_timed method to Coro::Channels.
363 - fixed Coro::Debug manpage.
364 - perl 5.11 compatbility improvement based on patch by
365 Andreas Kรถnig.
366
367 4.37 Sun Jan 20 11:25:23 CET 2008
368 - soften the check for an active parser for perl 5.10 (avoiding
369 "Coro::State::transfer called while parsing" in many cases).
370
371 4.36 Sun Jan 13 10:53:56 CET 2008
372 - reset diehook when terminating from an async_pool as to not
373 trigger any __DIE__ handlers.
374
375 4.35 Sun Jan 13 04:14:13 CET 2008
376 - "bt" debug command now displays any exceptions
377 from longmess and also skips the innermost
378 stackframes, giving more useufl output.
379 - allow backtraces in coroutines blocked in Coro::EV,
380 at a <1% speed hit.
381 - handle localising of $SIG{__WARN__} and __DIE__
382 properly (with a proper amount of dirty hacking).
383
384 4.34 Sat Dec 22 17:49:53 CET 2007
385 - upgrade to EV version 2.0 API.
386
387 4.33 Mon Dec 17 08:36:12 CET 2007
388 - make Coro::AIO etc. loadable in the absence of EV.
389
390 4.32 Mon Dec 17 07:46:02 CET 2007
391 - majorly improved Coro::Handle's performance with Coro::EV.
392 - implemented the until now mythical Coro::BDB module.
393 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
394
395 4.31 Wed Dec 5 12:32:39 CET 2007
396 - remove warn statement form Coro::Util.
397
398 4.3 Tue Dec 4 20:33:14 CET 2007
399 - calls to the idle function could cause stack corruption
400 when the stack changed.
401 - do no longer rely on the presence of EV::DNS if EV is used
402 (because it is gone), but instead take avdantage of EV::ADNS
403 if available.
404 - add ($) prototypes to all functions in Coro::Storable.
405 - use a conventional (safer) idle callback in Coro::EV.
406 - do accept quickly in Coro::Debug to avoid endless loops.
407
408 4.22 Fri Nov 30 16:04:04 CET 2007
409 - really use optimised versions for Event and EV in Coro::Util
410 and Coro::Handle.
411
412 4.21 Sun Nov 25 10:48:59 CET 2007
413 - fix a spurious memory read.
414 - Coro::EV no longer keeps the eventloop "alive".
415
416 4.2 Fri Nov 9 20:47:05 CET 2007
417 - enable/disable tracing from a new coroutine, not a pooled one.
418 - fix a memleak in Coro::Event.
419 - removed killall call from fork_eval.
420 - made sure store_fd is already loaded so that fork_eval
421 does not have to parse autoload in each subprocess.
422 - only use assembly method if -O switch is in $Config{optimize}.
423 - add (optional) Coro::EV module, so far the best event loop module
424 directly supported by Coro.
425 - if the event model is EV, use EV::DNS to resolve
426 stuff in Coro::Util.
427 - don't get confused by multiple event notifications in Coro::Handle.
428 - initial support for EV (libevent interface).
429 - require Event and EV using configure_requires, to force their existance.
430
431 4.13 Wed Oct 24 07:26:45 CEST 2007
432 - add Coro::Storable::blocking_thaw.
433 - use a vastly more complicated technique to localise
434 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
435 - use a coroutine for the idle callback Coro::Event,
436 instead of running Event in the current coroutine context.
437 This also catches recursive invocations.
438 - actually report fork errors in gethostbyname and inet_aton.
439
440 4.11 Thu Oct 11 02:40:24 CEST 2007
441 - port to threaded perls.
442
443 4.1 Thu Oct 11 02:38:16 CEST 2007
444 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
445 be local to each coro (see Coro::State).
446 - incompatible change: for very deep reasons, cede and cede_notself
447 cannot return anything, so nothing will be returned.
448 - possibly bring back 5.10 compatibility (untested).
449 - work around stupid (and wrong) warning on 5.10 :(.
450 - overlay the saved state over the context stack. This saves
451 a few hundred bytes per coroutine on average and also
452 speeds up context switching a bit.
453 - further tune default stack sizes.
454 - (more) correctly calculate stack usage in coro_rss.
455 - Coro::Storable::blocking_* did not properly lock
456 resulting in races between coroutines.
457 - added Coro::Storable::guard.
458 - stopping to trace a coroutine could destroy the cctx of
459 an unrelated coroutine.
460 - explain the relationship between Perl and C coroutines in
461 more detail in Coro::State.
462 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
463 causing a fork per resolve. This also affected Coro::Socket.
464 - switch to a separate stack in $coro->call/eval to avoid
465 invalidating pointers.
466
467 4.03 Sat Oct 6 21:24:00 CEST 2007
468 - added Coro::throw method.
469 - minor code cleanups.
470
471 4.02 Sat Oct 6 02:36:47 CEST 2007
472 - fix a very minor per-coroutine memleak (a single codereference).
473 - fixed a bug where the currently in-use c context would be freed
474 prematurely (can happen only when programs change the stacksize
475 or use tracing).
476 - tracing can no longer keep a coro alive after it terminated.
477 - do static branch prediction in the common path for gcc. gives
478 about 2-5% speed improvement here.
479
480 4.01 Fri Oct 5 22:10:49 CEST 2007
481 - instead of recreating *a* standard output handle we simply
482 use STDOUT, which is faster and hopefully more robust.
483
484 4.0 Fri Oct 5 12:56:00 CEST 2007
485 - incompatibly changed Coro::Storable::freeze.
486 - major new feature: added Coro::Debug, for interactive coroutine
487 debugging, tracing and much more.
488 - major bug fix: unbelievable, but true: $_, $/ and many other
489 "saved" variables actually weren't being saved. This has been fixed,
490 of course, while increasing performance while losing all the save
491 flags.
492 - save flags are gone, and all the api functions dealing with them.
493 - added Coro::Semaphore::adjust.
494 - added Coro::Util::fork_eval.
495 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
496 - added Coro::killall.
497 - reduce initial stack sizes to allow for "micro-coroutines".
498 - better async_pool resource management, moved parts of async_pool
499 handling to XS (major speed improvement).
500 - actually croak before modifying important data structures.
501 - refuse to transfer while compiling.
502 - possibly support eval EXPR better now.
503 - enable assembly per default on linux+bsd x86+amd64.
504 - all internal members were renamed _something for easier subclassing.
505 - many minor tweaks.
506
507 3.63 Wed May 16 14:10:06 CEST 2007
508 - implement handcoded assembly for x86/amd64 SVR ABI.
509
510 3.62 Fri Apr 27 21:36:06 CEST 2007
511 - upgrade libcoro (which might set unwind info correctly).
512 - change default on linux to setjmp/longjmp.
513
514 3.61 Thu Apr 19 12:36:18 CEST 2007
515 - Coro::Storable caused an endless loop when thawing invalid
516 pst's sometimes.
517 - use a Semaphore in Coro::Storable, as Storable doesn't
518 seem to be reentrant (although it is documented to
519 be threadsafe...).
520 - fix Coro::Signal to bring back the original unreliable
521 but stateful semantics.
522 - fixed a lot of typos in Coro.pm (patch submitted by David
523 Steinbrunner, which applied flawlessly).
524
525 3.6 Sat Apr 14 17:13:31 CEST 2007
526 - added some bugfixes to get eg/myhttpd working again.
527 - added Coro::Storable for often-cede'ing freeze/thaw.
528 - try to do a clean exit when a coroutine calls exit
529 (EXPERIMENTAL).
530 - got rid of indirect call through _coro_init.
531 - updated the partly antique examples in eg/ to
532 work again and be a bit less magic, too.
533 - fixed Coro::Signal semantics to work as documented again.
534
535 3.55 Sun Mar 25 01:20:47 CET 2007
536 - add SAVE_DEFFH to save the default filehandle and enable
537 it by default.
538 - finally move socket-operations from Coro::Socket to Coro::Handle
539 to be able to unblock foreign sockets.
540 - add Coro::State::save_also and guarded_save.
541 - add count accessor to Coro::Semaphore.
542 - add Coro::State::cctx_stacksize.
543 - just for the fun of it, do not rely on implicit context,
544 which can dramatically improve performance, but people
545 using windows-process-emulation perls do not care much about
546 performance.
547
548 3.51 Sun Mar 4 14:18:04 CET 2007
549 - fixed a problem when you weakened references to Coro::State's
550 (patch by Yuval Kogman).
551
552 3.501 Wed Feb 28 12:44:07 CET 2007
553 - rename some global symbols as macosx from hell redefines
554 them without asking.
555
556 3.5 Tue Feb 13 20:22:53 CET 2007
557 - do AnyEvent model detection earlier, avoiding problems
558 caused by first using AnyEvent and later Coro::Event.
559 - implement and document Coro::Event event objects.
560 - fix a potential problem in Coro::Event causing crashes.
561 - initialise PL_comppad when creating a new coroutine,
562 avoids crashes on early coro destruction.
563
564 3.41 Mon Jan 22 19:19:49 CET 2007
565 - readline on Coro::Handle did not support undefined $/,
566 nor did it deliver partial lines on EOF or error.
567 - implement malloc fallback for stack allocation because
568 stupid broken idiotic OSX has a stupid broken
569 idiotic fits-the-whole-os mmap "implementation" and
570 my dick feels longer if Coro is portable even to
571 obsolete platforms.
572
573 3.4 Fri Jan 19 21:52:54 CET 2007
574 - remove t/09_timer.t, as it isn't really testing much
575 but was rather flaky in practise.
576 - async_pool coro would keep arguments and callback alive until
577 it was reused.
578 - cancellation of a coroutine could cause spurious idle calls
579 in cede_notself.
580
581 3.3 Sat Jan 6 03:45:00 CET 2007
582 - implement $coro->on_destroy.
583 - Coro::Event blocking semantics have been changed,
584 documented and - hopefully - improved.
585 - fix nice adding, not subtracting, from priority.
586 - fix ->prio and api_is_ready (patch by Mark Hinds).
587 - fixed an assert ("... == prev__cctx->idle_te")
588 that could errronously trigger.
589 - fix various large and small memleaks.
590 - use a (hopefully) more stable cancel implementation
591 that immediately frees the coroutine data.
592 - cede/cede_notself return a status now.
593 - added Coro::guard function.
594 - added a global coroutine pool for jobs (on my machine,
595 I can create and execute 48k simple coros/s with async,
596 and 128k coros with async_pool).
597 - Coro::AIO now uses the coroutine priority as io priority.
598
599 3.2 Fri Dec 22 05:07:09 CET 2006
600 - improve portability to slightly older perls.
601 - use cleaner coroutine destruction.
602 - simplify configuration for users.
603 - optionally (unrecommended) prefer perl functions over
604 their coro replacements.
605
606 3.11 Tue Dec 5 13:11:24 CET 2006
607 - fixed some bogus assert's, but as perl.h disables assert even
608 without NDEBUG (thank you very much), not too many people should
609 notice (that did include myself). Andreas Kรถnig noticed, though :)
610 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
611 it. Also noticed by Andreas Kรถnig :)
612 - save/restore tainted status.
613 - verified to pass the testsuite on my 5.9.5.
614
615 3.1 Mon Dec 4 23:03:40 CET 2006
616 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
617 - incompatible change: transfer flags are now per-state.
618 - give a better error message on deadlock.
619 - document Coro::nready.
620 - enhanced testsuite.
621
622 3.01 Sun Dec 3 23:47:42 CET 2006
623 - forgot to include Coro::Timer.
624
625 3.0 Sun Dec 3 22:57:25 CET 2006
626 - the "FINALLY COMPLETELY STABLE" release (coming soon:
627 the "FAMOUS LAST WORDS" release).
628 - implement a new stack sharing algorithm, which uses a stack
629 pool (size currently hardcoded).
630 - make stack sharing mandatory (it no longer uses a heuristic).
631 - eval/die no longer cause weird problems under heavy use.
632 - Coro::Event could cause livelocks if it was used but
633 no Coro::Event watchers were used.
634 - Coro::Event now uses asynccheck as prepare does not
635 check for changed watchers.
636 - Coro::Event allows multiple waiting coros and will wake up one
637 per event.
638 - Coro::Event should be cleaner and more efficient now.
639 - new utility function Coro::unblock_sub.
640 - document the sad fact that Event is no longer reentrant.
641 - putting a coroutine into the ready queue twice could under
642 some circumstances lead to stack corruption.
643 - minor incompatibility: subclassing Coro::State is supported
644 directly now without going through a _coro_state member.
645 - state/coro switching is much faster now.
646 - very minor optimisations and code/documentation cleanup.
647 - avoid problems due to compiler inlining.
648 - removed timers from Coro::Timer -> use AnyEvent instead.
649 - replaced Coro::idle coroutine by (cleaner) idle handler.
650 - updated to newest libcoro.
651 - implement enhanced support for valgrind.
652 - implement is_ready and return value for ready.
653 - removed Coro::Cont, it was a misnomer (it's generators),
654 and it was somewhat annoying to get it right. Will come back
655 if somebody asks for it :)
656 - many ->wait methods and Coro::Event could return
657 spuriously without the event having happened.
658
659 2.5 Tue Nov 7 12:22:33 CET 2006
660 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
661 use AnyEvent, moved them to Coro/.
662 - added Coro::LWP which contains all the uglyness required to
663 make LWP non-blocking.
664 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
665 - fixed another bug in Coro::Select when the timeout was undef.
666 - reuse PL_start_env for all coros, saves some memory per coroutine.
667 - manage PL_top_env differently, hopefully to avoid panic: top_env.
668 - timeout argument was not properly used in Coro::Socket.
669 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
670 - emulate undocumented(!) functionality of IO::Socket required
671 by LWP(!!).
672 - updated eg/lwp to work with newer lwp's.
673 - remove "FATAL: uncaught exception" prefix. Coroutines that die
674 kill the whole process, just as exceptions in the main "coroutine"
675 did already.
676
677 2.1 Wed Nov 1 23:01:13 CET 2006
678 - fix a long-standing bug in Coro::Select where select with
679 zero timeout would instead change the current default filehandle.
680 - use a simpler and hopefully more robust way to clone padlists
681 (uses less memory and a perl function instead of our own).
682 - coro can now create a stack guard area on many architectures.
683 - Coro::AIO properly reexports additional functions from IO::AIO.
684 - updated libcoro with a workaround for OS X,
685 pach and testing by Michael Schwern.
686
687 2.0 Tue Oct 24 05:47:17 CEST 2006
688 - support additional aio requests in Coro::AIO.
689
690 1.9 (never properly released due to a glitch)
691
692 1.8 Thu Feb 2 00:59:06 CET 2006
693 - applied suggested patch by SAMV to avoid problems during stupid
694 mark & sweep gc run.
695 - applied patch by Scott Walters for 5.9.3 compatibility.
696
697 1.7 Tue Dec 27 01:41:58 CET 2005
698 - added Coro::AIO, a thin wrapper around IO::AIO.
699 - improved Makefile.PL explanations.
700
701 1.6 Mon Dec 12 21:30:05 CET 2005
702 - additionally save PM_curpm might fix as-of-yet
703 unseen problems with regex matches being attributed to the
704 wrong package.
705 - add t/10_bugs.t, which currently checks against imho broken
706 perls that use 0.26MB of stack space per Perl_magic_get
707 invocation (newer linuxes) as opposed to the 0.0002MB perl
708 normally uses.
709 - make stacksize configurable for the ultimate debian experience.
710
711 1.51 Mon Dec 12 18:48:36 CET 2005
712 - remove debugging warn() accidentally left in Coro::Select.
713
714 1.5 Tue Nov 29 13:32:44 CET 2005
715 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
716 - fix Coro::Select 'select'.
717 - strict'ify some modules.
718 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
719 - use XSLoader in selected modules.
720 - remove some 5.6 compatibility cruft.
721
722 1.4 Tue Sep 6 00:11:05 CEST 2005
723 - libcoro did not take into account the trampoline on amd64,
724 when 'l'inux method was used.
725
726 1.31 Tue Aug 30 23:31:33 CEST 2005
727 - some portability fixes/workarounds.
728
729 1.3 Sat Aug 20 03:08:56 CEST 2005
730 - no code changes. module seems to work fine.
731 - improve Coro::State docs, remove reference to nonexisting
732 Coro::State::flush method.
733 - no longer autodetect windows, present it as an option instead.
734
735 1.2 Mon May 16 02:00:55 CEST 2005
736 - included libcoro.c earlier in State.xs, might improve portability.
737 - use a faster and possibly more stable Coro::Cont implementation.
738 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
739 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
740 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
741
742 1.11 Thu Mar 3 18:00:52 CET 2005
743 - change of contact address.
744
745 1.1 Tue Feb 22 20:51:16 CET 2005
746 - support [l]inux method on amd64.
747 - allow some context switches while compiling/eval'ing.
748
749 1.0 Tue Aug 10 03:47:27 CEST 2004
750 - Event 0.89 fixes the bug again.
751 - fix the remaining(??) memleak. I didn't free the cache-AV for
752 padlists, so one AV per created anon-sub leaked.
753
754 0.97 Fri May 14 15:23:32 CEST 2004
755 - remove stupid gcc extension that was an experiment only anyway.
756
757 0.96 Thu May 13 18:09:29 CEST 2004
758 - tell people if they have yet another broken Event variant.
759 - made Coro to work _better_ with threads (it seems that, as long
760 as you use Coros in one thread only everything will work fine.
761 Looking at the perl source the problem might be missing locking
762 between Coro and some perl internal routines. That's not easily
763 fixable).
764 - (hopefully) got rid of the static xnext inside transfer.
765
766 0.95 Sun Feb 15 03:20:28 CET 2004
767 - removed Coro::State::flush, implemented a new and much
768 faster caching mechanism by attaching magic to the CV.
769 - a memleak with (real) closures remains.
770 - dramatically reduced size of stacks to 65536 (linux/x86), after
771 observing that even myhttpd never had more than a kilobyte
772 of stack in use (stack is still only physically allocated in
773 pagesize increments for systems suporting mmap).
774 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
775 - fix for threaded perl.
776
777 0.9 Sun Nov 23 04:32:39 CET 2003
778 - fix a few very hard to track down but (of course) stupid
779 memory leaks.
780 - stringify version number to avoid locale problems :(.
781 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
782 hopefully ignoring it "just works".
783
784 0.8 Wed Nov 5 19:38:40 CET 2003
785 - port to perl5.005_03, but only 5.8.x+ is supported!
786 - honor LocalAddr even without LocalPort in Coro::Socket.
787
788 0.7 Tue May 27 03:12:38 CEST 2003
789 - the version jump indicates some level of testing, not gobs
790 of new features.
791 - uh... I found the prompt function in ExtUtils::MakeMaker.
792 Highly correct stuff, that is...
793 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
794 the fix is not well understood by the author, so beware :(.
795
796 0.652 Thu May 8 02:54:46 CEST 2003
797 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
798 cause version 4 doesn't have ucontext.
799 - Benjamin Reed reported that setjmp works fine on darwin,
800 so preselect it.
801
802 0.651 Sat Mar 29 15:00:23 CET 2003
803 - fix a bug in Coro::Handle where some bytes could get lost
804 on reading, reported by jason@nichego.net.
805
806 0.65 Sun Mar 23 00:08:26 CET 2003
807 - added a README.linux-glibc.
808 - new module Coro::Select.
809 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
810 (I never claimed that I know what I am doing ;).
811 - set default to setjmp/longjmp on non-x86-linux.
812
813 0.6 Thu Nov 21 11:09:06 CET 2002
814 - made lazy context switching the default.
815 - fixed the bug where SemaphoreSet::guard
816 was the same as timed_guard and timed_guard was missing.
817 - fix a memleak in Coro::Timer.
818 - reclassified context sharing as not experimental.
819
820 0.534 Sun Apr 14 03:05:12 CEST 2002
821 - fixed backspace => tab in header parsing (myhttpd).
822 - added eg/readline, for elmex.
823 - Coro::Event now calls ->start on first call instead of
824 ->again, which created wrong timeouts for the first ->next.
825 - fixed a bug where Coro::Socket returned a closed socket instead
826 of nothing on ECONNREFUSED and similar error conditions.
827
828 0.533 Mon Feb 18 18:49:40 CET 2002
829 - adapt to finally working Event-0.85.
830 - fixed buggy event initialization (trapped under windows, but how
831 could it possibly have worked anywhere?)
832
833 0.532 Wed Jan 16 02:45:32 CET 2002
834 - removed the reference to pp_entersub, might work on
835 (some) win32 perls now (testsuite works under cygwin,
836 without Event, which freezes).
837
838 0.531 Mon Dec 10 22:18:44 CET 2001
839 - Fixed a bug in SemaphoreSet::timed_down.
840
841 0.53 Tue Nov 27 21:11:13 CET 2001
842 - some tests for Event.
843 - slightly different internal architecture to get refcnt into the object
844 and not the reference. might fix a few bugs, certainly did introduce
845 new ones.
846 - Coro::Timer (independent of Event).
847 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
848
849 0.52 Tue Nov 6 21:36:18 CET 2001
850 - ported to cygwin (trivial).
851 patch by Gerrit P. Haase <gp@familiehaase.de>.
852 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
853
854 0.51 Thu Nov 1 20:39:01 CET 2001
855 - terminate/cancel now work properly (otherwise termination
856 could cause a "next coroutine is not and contains not..."
857 error.
858 - added Coro::Socket::shutdown.
859 - Coro::Event::loop no is the same as Event::loop.
860 - implemented terminate with args + join.
861
862 0.5 Fri Sep 28 16:15:35 CEST 2001
863 - fixed "print" on a Coro::Handle. The print method worked.
864 - small tweaks (seem to reduce memory consumption a lot)
865 in various modules.
866 - splendid use of "no warnings" scattered throughout.
867 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
868
869 0.49 Sun Sep 16 02:42:45 CEST 2001
870 - changed some method calls to function calls for speed
871 inside Coro::Handle.
872 - make Coro::Handle use an array instead of a hash for
873 speed reasons.
874 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
875 sjlj and ucontext should now work.
876 - IRIX-specific port for libcoro.
877 - swapped order of accept results to match IO::Socket.
878 - changed getsock/peername to sock/peername to match IO::Socket.
879 - fixed a bug that caused segfault when returning to main task
880 under some circumstances.
881 - other bugfixes.
882
883 0.45 Sun Sep 2 02:54:01 CEST 2001
884 - new method Coro::Handle::timeout.
885 - corrected speling of set...name to get...name in Coro::Socket.
886 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
887 - Coro::Handle now supports fileno and FILENO.
888 - added eg/myhttpd, a web server.
889 - Coro::Socket now accepts numeric ports as well as "port(number)"
890 syntax.
891 - moved some scheduling primitives into xs code. more to come.
892 - new simple priority system.
893 - implemented Coro::Event using XS, almost four times faster!
894 - small memory corruption problem fixed (boy that was difficult).
895
896 0.13 Wed Aug 8 16:53:07 CEST 2001
897 - new method Coro::Handle::readline.
898 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
899 (see eg/lwp on how to make LWP non-blocking).
900 - renamed Coro::Socket::new_inet to new.
901 - added Coro::Util, some utility functions.
902 - die/eval now works better (still get annoying (but true ;)
903 "Callback called exit" messages).
904 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
905 - much better Coro::Socket support.
906
907 0.12 Fri Jul 27 04:19:01 CEST 2001
908 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
909 - very experimental stack sharing algorithm. not 100% safe but
910 should work well in practise ;)
911 - added Coro::RWLock.
912 - Coro::Specific now works.
913
914 0.11 Tue Jul 24 22:49:21 CEST 2001
915 - added specialized hack for newer and older linux versions (fast).
916 - renamed Coro::Event::IO to Coro::Handle.
917 - new module Coro::Socket.
918
919 0.1 Tue Jul 24 01:47:53 CEST 2001
920 - release candidate 3. A rather new internal structure :(
921 - the great renaming: Damian Conway gave me a suitable replacement
922 for yield.
923 - added Coro/libcoro, a portable coroutine implementation for C
924 (not even perl-dependent!!!), which can be used by Coro::State.
925 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
926 perl-5.6.
927 - new function Coro::Event::idle.
928 - the idle process is now overriden
929 by default in Coro::Event.
930 - Coro::Channel now enforces the size.
931 - canceling events now works in all cases.
932 - Coro state now include $_ and $@.
933 - yet another bug workaround that I do not really understand :(
934 - new module Coro::Event::IO (very undocumented) to do
935 non-blocking i/o.
936 - performance tuning.
937
938 0.08 Thu Jul 19 06:13:25 CEST 2001
939 - release candidate 2 (new functionality).
940 - add Coro::State::flush function.
941 - transfer now supports a flags value (mostly for speed ATM).
942 - might compile and work in the presence of threads now.
943 - continuations have a saner syntax.
944 - no more memleaks.
945
946 0.07 Tue Jul 17 17:40:18 CEST 2001
947 - release candidate 1 ;)
948 - slightly nicer code.
949 - fixed a scheduling bug in Coro::Event.
950
951 0.06 Tue Jul 17 04:23:24 CEST 2001
952 - ok, I found the showstopper - the same sub must not be
953 re-entered in two different coroutines, otherwise => crash. I
954 see no easy solution to this problem, except by walking the call
955 chaing and saving/restoring the cv's, which is what I do now.
956 - memory leaks still latent, especially at thread termination.
957 - Coro::Event now works (haha).
958
959 0.05 Sun Jul 15 17:32:20 CEST 2001
960 - fixed a few issues in Coro::Event.
961 - I forgot to include Coro::Event in 0.04 :(:(:(
962
963 0.04 Sun Jul 15 05:24:59 CEST 2001
964 - @_ is now properly localized.
965 - Coro::State is now easier subclassable.
966 - Coro::Cont now coroutine-aware.
967 - Coro::Specific is a low-overhead module to create
968 coroutine-specific vars.
969 - Coro::Event provides a simple interface to Event.
970
971 0.03 Fri Jul 13 14:51:52 CEST 2001
972 - transfer() now implemented in XS (beware).
973 - new module Coro::Cont for really faked continuations.
974 - big internal architecture changes: Coro::State is now
975 really low-level and can thus be used to implement other
976 interesting things, While "Coro::" implements a process-like
977 model. Still crude and subject to change.
978 - $_ and $@ are no longer being localized.
979
980 0.02 Tue Jul 10 01:38:17 CEST 2001
981 - implemented "async" attribute.
982 - $_ and $@ are now localized.
983 - added Coro::Channel.
984 - more testcases, still no docs.
985
986 0.01 Tue Jul 3 02:18:41 CEST 2001
987 - original version; copied from Convert::Scalar.
988