ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.492
Committed: Wed Jun 10 00:00:02 2009 UTC (14 years, 11 months ago) by root
Branch: MAIN
Changes since 1.491: +3 -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
10 - speed up ready queue management/contetx switching by using a linked
11 list instead of an array (~5-10%).
12
13 5.132 Fri May 29 09:00:39 CEST 2009
14 - do not keep a reference to the argument itself in
15 Coro::Semaphore::guard, as it could change later.
16 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
17 should support a prepare callback.
18
19 5.131 Mon Mar 16 23:20:37 CET 2009
20 - implement and document Coro->suspend, Coro->resume.
21 - fix Coro::Select implementation to not (often) close
22 the passed file descriptors (testcase provided by pippijn).
23
24 5.13 Mon Dec 15 21:51:42 CET 2008
25 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
26 - don't set diehook to C<undef> but instead to NULL, to avoid
27 spurious warnings.
28 - fix a lot of bugs in Coro::SemaphoreSet.
29 - Coro::SemaphoreSet will less often create a semaphore needlessly.
30 - add Coro::SemaphoreSet::count and wait methods.
31 - take advantage of the new Guard module.
32 - deprecate Coro::guard.
33 - try to fix the dreaded 01_unblock tests once more. I hate it when
34 testsuites need more fixing than the code they are supposed to test.
35 - croak in more cases when a required callback isn't resolvable.
36 - fix some minor issues in Coro::State->call/eval.
37 - use current coroutine context instead of a temporary one
38 when temporarily switching to another coroutine.
39 - do not call C-level on_destroy handlers during global destruction,
40 to avoid needless segfaults.
41
42 5.12 Sun Dec 7 13:30:38 CET 2008
43 - add default config for MirOS, which seems to be a bug-to-bug
44 compatible fork of openbsd ("world domination by releasing
45 openbsd cvs before the openbsd folks do it" or so :).
46 - free_padlist did destroy the names pad, not good, but didn't
47 seem to bother perl - this could fix issues such as eval ""
48 inside a function called from multiple coroutines.
49 - use a different method to detect destruction time.
50 - be more careful when freeing padlists just before global
51 destruction.
52 - fixed and expanded the call/cc example.
53 - renamed _terminate to _coro_run.
54 - new method Coro::Channel->shutdown.
55 - try pthreads on openbsd <4.4 (broken sigaltstack, will
56 pthreads fare better?).
57 - be less picky about destroying "a" running coroutine.
58
59 5.11 Tue Nov 25 21:49:05 CET 2008
60 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
61 working code with failed assertions, introducing more bugs than
62 it fixes, requiring elaborate workarounds :(
63
64 5.1 Mon Nov 24 08:54:59 CET 2008
65 - wrote a small introductory tutorial - Coro::Intro.
66 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
67 - Coro::Select did errornously dup the file descriptors
68 and didn't work with all AnyEvent backends.
69 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
70 LWP data transfers.
71 - disassociate c contexts from coro objects - this is agruably more
72 correct, but mostly allows sharing of cctxs between coro and state
73 objects, for added memory savings and speed increases.
74 - bumped $Coro::POOL_RSS up to 32kb by default.
75 - no longer set the optype to OP_CUSTOM, as B::* understandably
76 doesn't like this very much (and we *are* a type of entersub).
77 - implement state cloning, just to prove that call/cc can be done.
78 - automatically load Coro::AnyEvent in Coro::Handle.
79 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
80 do not have this method (and don't need it either).
81 - speed up generic anyevent methods in Coro::Handle by using rouse
82 callbacks.
83 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
84 others. It also makes the debugger happier, as you can trace
85 through the idle threads now.
86 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
87 - eg/event was pretty broken.
88 - better 5.8.6 compatibility.
89
90 5.0 Thu Nov 20 10:35:05 CET 2008
91 - NEW ARCHITECTURE: use the latest 4.x version if you experience
92 stability issues.
93 - bump API version to 7 - all dependents must be recompiled.
94 - removed timed_* functions - they were not being used anyways
95 and should be replaced by a more generic mechanism -
96 and were annoying to support anyways :)
97 - removed SemaphoreSet's waiter method - use sem method instead.
98 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
99 - async_pool did free a scalar value twice
100 ("Attempt to unreference...").
101 - fix a longstanding bug where calling terminate on a coro that
102 was waiting for a semaphore that was just becoming available
103 would cause a deadlock (semaphore would get into a state where
104 it was available but waiters were still blocked).
105 - calling throw on a coroutine that is waiting for a semaphore will
106 no longer make it acquire the semaphore (and thus leak a count).
107 - perl's process emulation is now not even theoretically supported
108 anymore.
109 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
110 conversion of callback-style to blocking-style.
111 - new methods $coro->schedule_to and ->cede_to, to specifically
112 schedule or cede to a specific coroutine.
113 - new function Coro::Semaphore::wait.
114 - use named constants in Coro::Channel (Richard Hundt).
115 - directly patch the entersub opcode calling SLF functions (cede,
116 transfer and so on). this does speed up context switching, but
117 more importanly, it frees us from the hardcoded behaviour of
118 entersub, so we might actually be able to return something from
119 those functions and atcually create new ones.
120 - take advantage of __builtin_frame_address on gcc.
121 - expose THX in coroapi (not sure whether this was a wise decision,
122 as "threaded" perls are running at half speed anyways).
123 - implement execute_slf (schedule-like-function) interface that makes
124 it possible to implement schedule-like-functions in XS.
125 - use new SLF interface to massively speed up Coro::EV by roughly a
126 factor of two.
127 - used new SLF interface to massively speed up Coro::Semaphore by a
128 factor of three.
129 - used new SLF interface to speed up Coro::AIO by roughly a factor of
130 four and reduce its memory usage considerably.
131 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
132 for a nice speedup and vastly more correct behaviour. Also implement
133 a new method "sem" to get at the underlying semaphore object.
134 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
135 (in comparison) 20-40% speedup.
136 - used new SLF interface to reimplement Coro::Signal gaining
137 some unknown (because I was too lazy), but certain, speedup, and also
138 making signals reliable for the first time.
139 - used new SLF interface and other optimisations to speed up async_pool
140 by a factor of two. It also doesn't rely on perl's exception mechanism
141 to exit anymore. The overhead for terminating an async_pool, coro over
142 a normal async is now very small.
143 - sped up coroutine creation/destruction by 40%.
144 - forgot to include Coro/libcoro/README in the dist for all these years.
145 - work around a freebsd pthreads bug (manual testcancel is required as
146 pthread_cond_wait isn't a cancellation point on freebsd).
147 - use new rouse functions to speed up and simplify Coro::BDB.
148 - make "prefer perl native functions" work with threaded perls.
149 - condense Coro::Debug ps output, hint at v and w flags.
150 - (libcoro) lots of minor cleanups and portability improvements.
151
152 4.914 Wed Nov 19 12:54:18 CET 2008
153 - fix a disastrous bug in the readline optimisation
154 introduced in 4.801.
155
156 4.913 Sat Nov 15 07:58:28 CET 2008
157 - async_pool did free a scalar value twice
158 ("Attempt to unreference...").
159
160 4.912 Thu Nov 13 18:31:23 CET 2008
161 - minor cleanups.
162 - use much larger stacks on linux and perl < 5.8.8.
163 - Coro::Debug::new_unix_server did not unlink the socket
164 when destroyed.
165
166 4.911 Tue Nov 11 04:26:17 CET 2008
167 - "port" to threaded perls.
168
169 4.91 Mon Nov 10 05:36:38 CET 2008
170 - the ->throw exception object no longer leaks.
171 - creating a new cctx leaked a scopestack entry (memleak).
172 - new coroutines didn't get created with a zero flags field
173 (unknown impact).
174 - calling ->throw on a not-yet-started coroutine should now work
175 instead of being ignored.
176 - ->throw is now supported on Coro::State objects.
177 - clean up cctx creation code a bit.
178 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
179
180 4.9 Sat Nov 8 17:45:27 CET 2008
181 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
182 - (libcoro) no longer leak threads in the experimental pthread backend,
183 also speed it up considerably.
184 - (libcoro) do not rely on makecontext passing void *'s unscathed.
185 - fix compiletime dependencies on libcoro in the Makefile.
186 - cctx_count wasn't always updated properly.
187 - Coro::State::cctx_stacksize wasn't applied correctly.
188 - new function Coro::State::cctx_max_idle.
189 - the default max number of idle C contexts is now 4.
190 - (libcoro) try harder to get _setjmp/_longjmp.
191 - (libcoro) cleanup and extend the libcoro API to officially
192 allow the creation of empty source contexts.
193 - very experimental workaround for the totally broken netbsd platform.
194 - tried to hack around openbsd bugs.
195
196 4.804 Wed Nov 5 16:36:00 CET 2008
197 - Coro::Debug::new_unix_server would not create a non-blocking listening
198 socket, sometimes causing freezes.
199 - (libcoro) fix misaligned stack points for setjmp and assembly
200 methods, which can cause crashes on x86/x86_64 with a sufficiently
201 aggressive compiler.
202 - new function: Coro::Debug::new_tcp_server.
203 - move ->throw into the Coro class because it only works on coro objects.
204
205 4.803 Mon Nov 3 17:16:12 CET 2008
206 - (libcoro) use a global asm statement to become independent of gcc
207 otpimisations for CORO_ASM (thanks to pippijn for the idea).
208 - try to workaround yet another broken bsd, this time dragonfly.
209
210 4.802 Thu Oct 30 10:56:12 CET 2008
211 - support -fno-omit-frame-pointer on x86 with the assembly method.
212 - tune 01_unblock.t tests a bit.
213
214 4.801 Wed Oct 22 18:33:37 CEST 2008
215 - improve readline speed for very long "lines".
216 - backport to 5.8.8.
217
218 4.8 Thu Oct 2 13:34:40 CEST 2008
219 - new function Coro::AIO::aio_wait.
220 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
221 - greatly speed up and reduce memory usage of Coro::AIO requests.
222 - implement some other ยต-optimisations.
223
224 4.749 Mon Sep 29 14:40:12 CEST 2008
225 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
226
227 4.748 Sat Sep 27 14:03:19 CEST 2008
228 - implement, but do not document, PerlIO::cede(granularity).
229 - Coro::Storable forgot to wrap Storable::pstore.
230 - work around the multitude of leaks and memory corruption
231 bugs in PerlIO::via by using our own C-level perliol. As a side
232 effect, Coro::Storable is now much, much, much faster.
233
234 4.747 Tue Sep 23 01:59:41 CEST 2008
235 - fix a per-cv memleak (one empty array was leaked per
236 code reference).
237 - avoid a crash in coro->call|rss when the coroutine was already
238 destroyed (most noticably when using Coro::Debug::ps :)
239 - also protect *Storable::FILE.
240 - push up default storable granularity to 20ms.
241
242 4.746 Sun Sep 21 03:22:20 CEST 2008
243 - be more insistent on locking Storable against reentrancy
244 in Coro::Storable.
245 - move swap_def?v and throw to Coro::State, as documented.
246
247 4.745 Thu Jul 24 00:14:38 CEST 2008
248 - remove debugging code related to MgPV_nolen_const, also try to
249 make it compile with perl 5.8.6 (yes, apple apparently loves
250 outdated software). Reported by John S.
251
252 4.744 Tue Jul 8 22:06:35 CEST 2008
253 - correctly provide default DIE/WARN handlers as documented.
254 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
255 implementations inside perl work fine for us.
256
257 4.743 Mon Jun 16 00:21:57 CEST 2008
258 - when using Coro::EV without running EV::loop it could
259 result in busy-waiting for events, this has been fixed.
260 - reduce codesize and improve performance by using EV_DEFAULT_UC.
261
262 4.742 Sat May 31 14:10:21 CEST 2008
263 - implement a workaround for (some) perl <5.8.8 versions.
264 - require EV 3.3+.
265
266 4.741 Fri May 30 23:33:09 CEST 2008
267 - tell netbsd how utterly broken their imitation of an OS is
268 and refuse to build by default if pthreads are in use.
269 - switch to "s" method on all bsd's by default, as their ucontext
270 stuff seems just too broken.
271 - fix a bug in Coro::Select.
272
273 4.74 Thu May 29 20:05:31 CEST 2008
274 - do not test Coro::LWP for lack of dependencies.
275 - do not test Coro::Select for lack of working perls.
276
277 4.73 Thu May 29 2008
278 - fix a bug in Coro::EV which would cause it to block despite
279 there being runnable coroutines.
280 - sprinkle "no warnings" freely over everything, also suppress
281 warnings for some other modules.
282 - fix typo in WSAEWOULDBLOCK.
283
284 4.72 Sun May 25 05:14:36 CEST 2008
285 - tweak META.yaml a bit, unfortunately, there is no documented way
286 to have optional dependencies with CPAN. doh :(
287 - avoid running some tests on windows because they would fail due to
288 perl bug (broken fork, broken pipes...).
289 - work around perl on windows bugs where perl returns undocumented
290 error codes for sysread, syswrite etc. by taking advantage
291 of AnyEvent's workaround for that problem.
292 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
293 a common perl implementation bug on windows.
294 - use unix domain sockets in testsuite to work around
295 common perl implementation bugs on widows (they are emulated by
296 tcp sockets on windows. ugh.)
297
298 4.71 Sat May 24 20:01:27 CEST 2008
299 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
300 - take advantage of async name resolution of AnyEvent::Util.
301 - work around brutal inet_aton override in Coro::LWP.
302 - take advantage of the readyhook in Coro::EV, for smoother
303 scheduling.
304
305 4.7 Sun May 11 00:32:19 CEST 2008
306 - completely reworked the Coro manpage.
307 - added Coro::AnyEvent, generic event loop integration.
308 - implement cancel, ready and kill commands in Coro::Debug.
309 - document find_coro in Coro::Debug.
310 - incompatible change: rename has_stack to has_cctx.
311 - Coro::AIO and Coro::BDB no longer force event model detection,
312 use AnyEvent::AIO and AnyEvent::BDB.
313
314 4.6 Sat Apr 26 10:05:14 CEST 2008
315 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
316 by the Coro module. It was a mistake to have it in the first place.
317 - (experimental) support for activestate perl 5.10 (method "w").
318 - (experimental) support for strawberry perl (method "a").
319 - coro_sigelem_set did not return a value although it had to,
320 actual impact unknown.
321
322 4.51 Mon Apr 14 13:28:27 CEST 2008
323 - make it compile again on 5.8.
324
325 4.50 Thu Apr 10 09:43:17 CEST 2008
326 - I did it twice! (see 4.49).
327
328 4.49 Sun Apr 6 21:23:31 CEST 2008
329 - grr, instead of compiling the recent changes on 5.10 only they
330 were compiled on 5.8 only.
331
332 4.48 Sun Apr 6 20:36:46 CEST 2008
333 - allow coroutine switches during eval's under 5.10.x, as it seems
334 the parser is a per-interpreter option now, so this is safe
335 (this might fix the odd crashing bug).
336 - drop support for 5.9.x versions: they are dead, jim.
337
338 4.47 Sun Apr 6 00:37:52 CEST 2008
339 - force cctx allocation on API calls: we know we need to force one
340 and gcc actually manages to confuse our heuristic nowadays,
341 leading to crashes and worse.
342 - document force_cctx.
343
344 4.46 Fri Apr 4 22:05:43 CEST 2008
345 - upgrade libcoro, resulting in pthread-backend (which was only created
346 to fulfill the rules of the programming languages shootout).
347
348 4.45 Thu Mar 13 11:55:36 CET 2008
349 - fix a file leaking bug in eg/mhyttpd that would allow
350 downloading of any file (reported by oesi).
351 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
352 (also add testcase).
353 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
354 apparently having loads of fun with that).
355 - implement $state->swap_defsv, ->swap_defav and document ->throw.
356
357 4.4 Wed Feb 13 16:44:29 CET 2008
358 - only assume .cfi pseudo isns exist on GNU/Linux.
359 - add get_timed method to Coro::Channels.
360 - fixed Coro::Debug manpage.
361 - perl 5.11 compatbility improvement based on patch by
362 Andreas Kรถnig.
363
364 4.37 Sun Jan 20 11:25:23 CET 2008
365 - soften the check for an active parser for perl 5.10 (avoiding
366 "Coro::State::transfer called while parsing" in many cases).
367
368 4.36 Sun Jan 13 10:53:56 CET 2008
369 - reset diehook when terminating from an async_pool as to not
370 trigger any __DIE__ handlers.
371
372 4.35 Sun Jan 13 04:14:13 CET 2008
373 - "bt" debug command now displays any exceptions
374 from longmess and also skips the innermost
375 stackframes, giving more useufl output.
376 - allow backtraces in coroutines blocked in Coro::EV,
377 at a <1% speed hit.
378 - handle localising of $SIG{__WARN__} and __DIE__
379 properly (with a proper amount of dirty hacking).
380
381 4.34 Sat Dec 22 17:49:53 CET 2007
382 - upgrade to EV version 2.0 API.
383
384 4.33 Mon Dec 17 08:36:12 CET 2007
385 - make Coro::AIO etc. loadable in the absence of EV.
386
387 4.32 Mon Dec 17 07:46:02 CET 2007
388 - majorly improved Coro::Handle's performance with Coro::EV.
389 - implemented the until now mythical Coro::BDB module.
390 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
391
392 4.31 Wed Dec 5 12:32:39 CET 2007
393 - remove warn statement form Coro::Util.
394
395 4.3 Tue Dec 4 20:33:14 CET 2007
396 - calls to the idle function could cause stack corruption
397 when the stack changed.
398 - do no longer rely on the presence of EV::DNS if EV is used
399 (because it is gone), but instead take avdantage of EV::ADNS
400 if available.
401 - add ($) prototypes to all functions in Coro::Storable.
402 - use a conventional (safer) idle callback in Coro::EV.
403 - do accept quickly in Coro::Debug to avoid endless loops.
404
405 4.22 Fri Nov 30 16:04:04 CET 2007
406 - really use optimised versions for Event and EV in Coro::Util
407 and Coro::Handle.
408
409 4.21 Sun Nov 25 10:48:59 CET 2007
410 - fix a spurious memory read.
411 - Coro::EV no longer keeps the eventloop "alive".
412
413 4.2 Fri Nov 9 20:47:05 CET 2007
414 - enable/disable tracing from a new coroutine, not a pooled one.
415 - fix a memleak in Coro::Event.
416 - removed killall call from fork_eval.
417 - made sure store_fd is already loaded so that fork_eval
418 does not have to parse autoload in each subprocess.
419 - only use assembly method if -O switch is in $Config{optimize}.
420 - add (optional) Coro::EV module, so far the best event loop module
421 directly supported by Coro.
422 - if the event model is EV, use EV::DNS to resolve
423 stuff in Coro::Util.
424 - don't get confused by multiple event notifications in Coro::Handle.
425 - initial support for EV (libevent interface).
426 - require Event and EV using configure_requires, to force their existance.
427
428 4.13 Wed Oct 24 07:26:45 CEST 2007
429 - add Coro::Storable::blocking_thaw.
430 - use a vastly more complicated technique to localise
431 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
432 - use a coroutine for the idle callback Coro::Event,
433 instead of running Event in the current coroutine context.
434 This also catches recursive invocations.
435 - actually report fork errors in gethostbyname and inet_aton.
436
437 4.11 Thu Oct 11 02:40:24 CEST 2007
438 - port to threaded perls.
439
440 4.1 Thu Oct 11 02:38:16 CEST 2007
441 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
442 be local to each coro (see Coro::State).
443 - incompatible change: for very deep reasons, cede and cede_notself
444 cannot return anything, so nothing will be returned.
445 - possibly bring back 5.10 compatibility (untested).
446 - work around stupid (and wrong) warning on 5.10 :(.
447 - overlay the saved state over the context stack. This saves
448 a few hundred bytes per coroutine on average and also
449 speeds up context switching a bit.
450 - further tune default stack sizes.
451 - (more) correctly calculate stack usage in coro_rss.
452 - Coro::Storable::blocking_* did not properly lock
453 resulting in races between coroutines.
454 - added Coro::Storable::guard.
455 - stopping to trace a coroutine could destroy the cctx of
456 an unrelated coroutine.
457 - explain the relationship between Perl and C coroutines in
458 more detail in Coro::State.
459 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
460 causing a fork per resolve. This also affected Coro::Socket.
461 - switch to a separate stack in $coro->call/eval to avoid
462 invalidating pointers.
463
464 4.03 Sat Oct 6 21:24:00 CEST 2007
465 - added Coro::throw method.
466 - minor code cleanups.
467
468 4.02 Sat Oct 6 02:36:47 CEST 2007
469 - fix a very minor per-coroutine memleak (a single codereference).
470 - fixed a bug where the currently in-use c context would be freed
471 prematurely (can happen only when programs change the stacksize
472 or use tracing).
473 - tracing can no longer keep a coro alive after it terminated.
474 - do static branch prediction in the common path for gcc. gives
475 about 2-5% speed improvement here.
476
477 4.01 Fri Oct 5 22:10:49 CEST 2007
478 - instead of recreating *a* standard output handle we simply
479 use STDOUT, which is faster and hopefully more robust.
480
481 4.0 Fri Oct 5 12:56:00 CEST 2007
482 - incompatibly changed Coro::Storable::freeze.
483 - major new feature: added Coro::Debug, for interactive coroutine
484 debugging, tracing and much more.
485 - major bug fix: unbelievable, but true: $_, $/ and many other
486 "saved" variables actually weren't being saved. This has been fixed,
487 of course, while increasing performance while losing all the save
488 flags.
489 - save flags are gone, and all the api functions dealing with them.
490 - added Coro::Semaphore::adjust.
491 - added Coro::Util::fork_eval.
492 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
493 - added Coro::killall.
494 - reduce initial stack sizes to allow for "micro-coroutines".
495 - better async_pool resource management, moved parts of async_pool
496 handling to XS (major speed improvement).
497 - actually croak before modifying important data structures.
498 - refuse to transfer while compiling.
499 - possibly support eval EXPR better now.
500 - enable assembly per default on linux+bsd x86+amd64.
501 - all internal members were renamed _something for easier subclassing.
502 - many minor tweaks.
503
504 3.63 Wed May 16 14:10:06 CEST 2007
505 - implement handcoded assembly for x86/amd64 SVR ABI.
506
507 3.62 Fri Apr 27 21:36:06 CEST 2007
508 - upgrade libcoro (which might set unwind info correctly).
509 - change default on linux to setjmp/longjmp.
510
511 3.61 Thu Apr 19 12:36:18 CEST 2007
512 - Coro::Storable caused an endless loop when thawing invalid
513 pst's sometimes.
514 - use a Semaphore in Coro::Storable, as Storable doesn't
515 seem to be reentrant (although it is documented to
516 be threadsafe...).
517 - fix Coro::Signal to bring back the original unreliable
518 but stateful semantics.
519 - fixed a lot of typos in Coro.pm (patch submitted by David
520 Steinbrunner, which applied flawlessly).
521
522 3.6 Sat Apr 14 17:13:31 CEST 2007
523 - added some bugfixes to get eg/myhttpd working again.
524 - added Coro::Storable for often-cede'ing freeze/thaw.
525 - try to do a clean exit when a coroutine calls exit
526 (EXPERIMENTAL).
527 - got rid of indirect call through _coro_init.
528 - updated the partly antique examples in eg/ to
529 work again and be a bit less magic, too.
530 - fixed Coro::Signal semantics to work as documented again.
531
532 3.55 Sun Mar 25 01:20:47 CET 2007
533 - add SAVE_DEFFH to save the default filehandle and enable
534 it by default.
535 - finally move socket-operations from Coro::Socket to Coro::Handle
536 to be able to unblock foreign sockets.
537 - add Coro::State::save_also and guarded_save.
538 - add count accessor to Coro::Semaphore.
539 - add Coro::State::cctx_stacksize.
540 - just for the fun of it, do not rely on implicit context,
541 which can dramatically improve performance, but people
542 using windows-process-emulation perls do not care much about
543 performance.
544
545 3.51 Sun Mar 4 14:18:04 CET 2007
546 - fixed a problem when you weakened references to Coro::State's
547 (patch by Yuval Kogman).
548
549 3.501 Wed Feb 28 12:44:07 CET 2007
550 - rename some global symbols as macosx from hell redefines
551 them without asking.
552
553 3.5 Tue Feb 13 20:22:53 CET 2007
554 - do AnyEvent model detection earlier, avoiding problems
555 caused by first using AnyEvent and later Coro::Event.
556 - implement and document Coro::Event event objects.
557 - fix a potential problem in Coro::Event causing crashes.
558 - initialise PL_comppad when creating a new coroutine,
559 avoids crashes on early coro destruction.
560
561 3.41 Mon Jan 22 19:19:49 CET 2007
562 - readline on Coro::Handle did not support undefined $/,
563 nor did it deliver partial lines on EOF or error.
564 - implement malloc fallback for stack allocation because
565 stupid broken idiotic OSX has a stupid broken
566 idiotic fits-the-whole-os mmap "implementation" and
567 my dick feels longer if Coro is portable even to
568 obsolete platforms.
569
570 3.4 Fri Jan 19 21:52:54 CET 2007
571 - remove t/09_timer.t, as it isn't really testing much
572 but was rather flaky in practise.
573 - async_pool coro would keep arguments and callback alive until
574 it was reused.
575 - cancellation of a coroutine could cause spurious idle calls
576 in cede_notself.
577
578 3.3 Sat Jan 6 03:45:00 CET 2007
579 - implement $coro->on_destroy.
580 - Coro::Event blocking semantics have been changed,
581 documented and - hopefully - improved.
582 - fix nice adding, not subtracting, from priority.
583 - fix ->prio and api_is_ready (patch by Mark Hinds).
584 - fixed an assert ("... == prev__cctx->idle_te")
585 that could errronously trigger.
586 - fix various large and small memleaks.
587 - use a (hopefully) more stable cancel implementation
588 that immediately frees the coroutine data.
589 - cede/cede_notself return a status now.
590 - added Coro::guard function.
591 - added a global coroutine pool for jobs (on my machine,
592 I can create and execute 48k simple coros/s with async,
593 and 128k coros with async_pool).
594 - Coro::AIO now uses the coroutine priority as io priority.
595
596 3.2 Fri Dec 22 05:07:09 CET 2006
597 - improve portability to slightly older perls.
598 - use cleaner coroutine destruction.
599 - simplify configuration for users.
600 - optionally (unrecommended) prefer perl functions over
601 their coro replacements.
602
603 3.11 Tue Dec 5 13:11:24 CET 2006
604 - fixed some bogus assert's, but as perl.h disables assert even
605 without NDEBUG (thank you very much), not too many people should
606 notice (that did include myself). Andreas Kรถnig noticed, though :)
607 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
608 it. Also noticed by Andreas Kรถnig :)
609 - save/restore tainted status.
610 - verified to pass the testsuite on my 5.9.5.
611
612 3.1 Mon Dec 4 23:03:40 CET 2006
613 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
614 - incompatible change: transfer flags are now per-state.
615 - give a better error message on deadlock.
616 - document Coro::nready.
617 - enhanced testsuite.
618
619 3.01 Sun Dec 3 23:47:42 CET 2006
620 - forgot to include Coro::Timer.
621
622 3.0 Sun Dec 3 22:57:25 CET 2006
623 - the "FINALLY COMPLETELY STABLE" release (coming soon:
624 the "FAMOUS LAST WORDS" release).
625 - implement a new stack sharing algorithm, which uses a stack
626 pool (size currently hardcoded).
627 - make stack sharing mandatory (it no longer uses a heuristic).
628 - eval/die no longer cause weird problems under heavy use.
629 - Coro::Event could cause livelocks if it was used but
630 no Coro::Event watchers were used.
631 - Coro::Event now uses asynccheck as prepare does not
632 check for changed watchers.
633 - Coro::Event allows multiple waiting coros and will wake up one
634 per event.
635 - Coro::Event should be cleaner and more efficient now.
636 - new utility function Coro::unblock_sub.
637 - document the sad fact that Event is no longer reentrant.
638 - putting a coroutine into the ready queue twice could under
639 some circumstances lead to stack corruption.
640 - minor incompatibility: subclassing Coro::State is supported
641 directly now without going through a _coro_state member.
642 - state/coro switching is much faster now.
643 - very minor optimisations and code/documentation cleanup.
644 - avoid problems due to compiler inlining.
645 - removed timers from Coro::Timer -> use AnyEvent instead.
646 - replaced Coro::idle coroutine by (cleaner) idle handler.
647 - updated to newest libcoro.
648 - implement enhanced support for valgrind.
649 - implement is_ready and return value for ready.
650 - removed Coro::Cont, it was a misnomer (it's generators),
651 and it was somewhat annoying to get it right. Will come back
652 if somebody asks for it :)
653 - many ->wait methods and Coro::Event could return
654 spuriously without the event having happened.
655
656 2.5 Tue Nov 7 12:22:33 CET 2006
657 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
658 use AnyEvent, moved them to Coro/.
659 - added Coro::LWP which contains all the uglyness required to
660 make LWP non-blocking.
661 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
662 - fixed another bug in Coro::Select when the timeout was undef.
663 - reuse PL_start_env for all coros, saves some memory per coroutine.
664 - manage PL_top_env differently, hopefully to avoid panic: top_env.
665 - timeout argument was not properly used in Coro::Socket.
666 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
667 - emulate undocumented(!) functionality of IO::Socket required
668 by LWP(!!).
669 - updated eg/lwp to work with newer lwp's.
670 - remove "FATAL: uncaught exception" prefix. Coroutines that die
671 kill the whole process, just as exceptions in the main "coroutine"
672 did already.
673
674 2.1 Wed Nov 1 23:01:13 CET 2006
675 - fix a long-standing bug in Coro::Select where select with
676 zero timeout would instead change the current default filehandle.
677 - use a simpler and hopefully more robust way to clone padlists
678 (uses less memory and a perl function instead of our own).
679 - coro can now create a stack guard area on many architectures.
680 - Coro::AIO properly reexports additional functions from IO::AIO.
681 - updated libcoro with a workaround for OS X,
682 pach and testing by Michael Schwern.
683
684 2.0 Tue Oct 24 05:47:17 CEST 2006
685 - support additional aio requests in Coro::AIO.
686
687 1.9 (never properly released due to a glitch)
688
689 1.8 Thu Feb 2 00:59:06 CET 2006
690 - applied suggested patch by SAMV to avoid problems during stupid
691 mark & sweep gc run.
692 - applied patch by Scott Walters for 5.9.3 compatibility.
693
694 1.7 Tue Dec 27 01:41:58 CET 2005
695 - added Coro::AIO, a thin wrapper around IO::AIO.
696 - improved Makefile.PL explanations.
697
698 1.6 Mon Dec 12 21:30:05 CET 2005
699 - additionally save PM_curpm might fix as-of-yet
700 unseen problems with regex matches being attributed to the
701 wrong package.
702 - add t/10_bugs.t, which currently checks against imho broken
703 perls that use 0.26MB of stack space per Perl_magic_get
704 invocation (newer linuxes) as opposed to the 0.0002MB perl
705 normally uses.
706 - make stacksize configurable for the ultimate debian experience.
707
708 1.51 Mon Dec 12 18:48:36 CET 2005
709 - remove debugging warn() accidentally left in Coro::Select.
710
711 1.5 Tue Nov 29 13:32:44 CET 2005
712 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
713 - fix Coro::Select 'select'.
714 - strict'ify some modules.
715 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
716 - use XSLoader in selected modules.
717 - remove some 5.6 compatibility cruft.
718
719 1.4 Tue Sep 6 00:11:05 CEST 2005
720 - libcoro did not take into account the trampoline on amd64,
721 when 'l'inux method was used.
722
723 1.31 Tue Aug 30 23:31:33 CEST 2005
724 - some portability fixes/workarounds.
725
726 1.3 Sat Aug 20 03:08:56 CEST 2005
727 - no code changes. module seems to work fine.
728 - improve Coro::State docs, remove reference to nonexisting
729 Coro::State::flush method.
730 - no longer autodetect windows, present it as an option instead.
731
732 1.2 Mon May 16 02:00:55 CEST 2005
733 - included libcoro.c earlier in State.xs, might improve portability.
734 - use a faster and possibly more stable Coro::Cont implementation.
735 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
736 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
737 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
738
739 1.11 Thu Mar 3 18:00:52 CET 2005
740 - change of contact address.
741
742 1.1 Tue Feb 22 20:51:16 CET 2005
743 - support [l]inux method on amd64.
744 - allow some context switches while compiling/eval'ing.
745
746 1.0 Tue Aug 10 03:47:27 CEST 2004
747 - Event 0.89 fixes the bug again.
748 - fix the remaining(??) memleak. I didn't free the cache-AV for
749 padlists, so one AV per created anon-sub leaked.
750
751 0.97 Fri May 14 15:23:32 CEST 2004
752 - remove stupid gcc extension that was an experiment only anyway.
753
754 0.96 Thu May 13 18:09:29 CEST 2004
755 - tell people if they have yet another broken Event variant.
756 - made Coro to work _better_ with threads (it seems that, as long
757 as you use Coros in one thread only everything will work fine.
758 Looking at the perl source the problem might be missing locking
759 between Coro and some perl internal routines. That's not easily
760 fixable).
761 - (hopefully) got rid of the static xnext inside transfer.
762
763 0.95 Sun Feb 15 03:20:28 CET 2004
764 - removed Coro::State::flush, implemented a new and much
765 faster caching mechanism by attaching magic to the CV.
766 - a memleak with (real) closures remains.
767 - dramatically reduced size of stacks to 65536 (linux/x86), after
768 observing that even myhttpd never had more than a kilobyte
769 of stack in use (stack is still only physically allocated in
770 pagesize increments for systems suporting mmap).
771 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
772 - fix for threaded perl.
773
774 0.9 Sun Nov 23 04:32:39 CET 2003
775 - fix a few very hard to track down but (of course) stupid
776 memory leaks.
777 - stringify version number to avoid locale problems :(.
778 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
779 hopefully ignoring it "just works".
780
781 0.8 Wed Nov 5 19:38:40 CET 2003
782 - port to perl5.005_03, but only 5.8.x+ is supported!
783 - honor LocalAddr even without LocalPort in Coro::Socket.
784
785 0.7 Tue May 27 03:12:38 CEST 2003
786 - the version jump indicates some level of testing, not gobs
787 of new features.
788 - uh... I found the prompt function in ExtUtils::MakeMaker.
789 Highly correct stuff, that is...
790 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
791 the fix is not well understood by the author, so beware :(.
792
793 0.652 Thu May 8 02:54:46 CEST 2003
794 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
795 cause version 4 doesn't have ucontext.
796 - Benjamin Reed reported that setjmp works fine on darwin,
797 so preselect it.
798
799 0.651 Sat Mar 29 15:00:23 CET 2003
800 - fix a bug in Coro::Handle where some bytes could get lost
801 on reading, reported by jason@nichego.net.
802
803 0.65 Sun Mar 23 00:08:26 CET 2003
804 - added a README.linux-glibc.
805 - new module Coro::Select.
806 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
807 (I never claimed that I know what I am doing ;).
808 - set default to setjmp/longjmp on non-x86-linux.
809
810 0.6 Thu Nov 21 11:09:06 CET 2002
811 - made lazy context switching the default.
812 - fixed the bug where SemaphoreSet::guard
813 was the same as timed_guard and timed_guard was missing.
814 - fix a memleak in Coro::Timer.
815 - reclassified context sharing as not experimental.
816
817 0.534 Sun Apr 14 03:05:12 CEST 2002
818 - fixed backspace => tab in header parsing (myhttpd).
819 - added eg/readline, for elmex.
820 - Coro::Event now calls ->start on first call instead of
821 ->again, which created wrong timeouts for the first ->next.
822 - fixed a bug where Coro::Socket returned a closed socket instead
823 of nothing on ECONNREFUSED and similar error conditions.
824
825 0.533 Mon Feb 18 18:49:40 CET 2002
826 - adapt to finally working Event-0.85.
827 - fixed buggy event initialization (trapped under windows, but how
828 could it possibly have worked anywhere?)
829
830 0.532 Wed Jan 16 02:45:32 CET 2002
831 - removed the reference to pp_entersub, might work on
832 (some) win32 perls now (testsuite works under cygwin,
833 without Event, which freezes).
834
835 0.531 Mon Dec 10 22:18:44 CET 2001
836 - Fixed a bug in SemaphoreSet::timed_down.
837
838 0.53 Tue Nov 27 21:11:13 CET 2001
839 - some tests for Event.
840 - slightly different internal architecture to get refcnt into the object
841 and not the reference. might fix a few bugs, certainly did introduce
842 new ones.
843 - Coro::Timer (independent of Event).
844 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
845
846 0.52 Tue Nov 6 21:36:18 CET 2001
847 - ported to cygwin (trivial).
848 patch by Gerrit P. Haase <gp@familiehaase.de>.
849 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
850
851 0.51 Thu Nov 1 20:39:01 CET 2001
852 - terminate/cancel now work properly (otherwise termination
853 could cause a "next coroutine is not and contains not..."
854 error.
855 - added Coro::Socket::shutdown.
856 - Coro::Event::loop no is the same as Event::loop.
857 - implemented terminate with args + join.
858
859 0.5 Fri Sep 28 16:15:35 CEST 2001
860 - fixed "print" on a Coro::Handle. The print method worked.
861 - small tweaks (seem to reduce memory consumption a lot)
862 in various modules.
863 - splendid use of "no warnings" scattered throughout.
864 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
865
866 0.49 Sun Sep 16 02:42:45 CEST 2001
867 - changed some method calls to function calls for speed
868 inside Coro::Handle.
869 - make Coro::Handle use an array instead of a hash for
870 speed reasons.
871 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
872 sjlj and ucontext should now work.
873 - IRIX-specific port for libcoro.
874 - swapped order of accept results to match IO::Socket.
875 - changed getsock/peername to sock/peername to match IO::Socket.
876 - fixed a bug that caused segfault when returning to main task
877 under some circumstances.
878 - other bugfixes.
879
880 0.45 Sun Sep 2 02:54:01 CEST 2001
881 - new method Coro::Handle::timeout.
882 - corrected speling of set...name to get...name in Coro::Socket.
883 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
884 - Coro::Handle now supports fileno and FILENO.
885 - added eg/myhttpd, a web server.
886 - Coro::Socket now accepts numeric ports as well as "port(number)"
887 syntax.
888 - moved some scheduling primitives into xs code. more to come.
889 - new simple priority system.
890 - implemented Coro::Event using XS, almost four times faster!
891 - small memory corruption problem fixed (boy that was difficult).
892
893 0.13 Wed Aug 8 16:53:07 CEST 2001
894 - new method Coro::Handle::readline.
895 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
896 (see eg/lwp on how to make LWP non-blocking).
897 - renamed Coro::Socket::new_inet to new.
898 - added Coro::Util, some utility functions.
899 - die/eval now works better (still get annoying (but true ;)
900 "Callback called exit" messages).
901 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
902 - much better Coro::Socket support.
903
904 0.12 Fri Jul 27 04:19:01 CEST 2001
905 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
906 - very experimental stack sharing algorithm. not 100% safe but
907 should work well in practise ;)
908 - added Coro::RWLock.
909 - Coro::Specific now works.
910
911 0.11 Tue Jul 24 22:49:21 CEST 2001
912 - added specialized hack for newer and older linux versions (fast).
913 - renamed Coro::Event::IO to Coro::Handle.
914 - new module Coro::Socket.
915
916 0.1 Tue Jul 24 01:47:53 CEST 2001
917 - release candidate 3. A rather new internal structure :(
918 - the great renaming: Damian Conway gave me a suitable replacement
919 for yield.
920 - added Coro/libcoro, a portable coroutine implementation for C
921 (not even perl-dependent!!!), which can be used by Coro::State.
922 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
923 perl-5.6.
924 - new function Coro::Event::idle.
925 - the idle process is now overriden
926 by default in Coro::Event.
927 - Coro::Channel now enforces the size.
928 - canceling events now works in all cases.
929 - Coro state now include $_ and $@.
930 - yet another bug workaround that I do not really understand :(
931 - new module Coro::Event::IO (very undocumented) to do
932 non-blocking i/o.
933 - performance tuning.
934
935 0.08 Thu Jul 19 06:13:25 CEST 2001
936 - release candidate 2 (new functionality).
937 - add Coro::State::flush function.
938 - transfer now supports a flags value (mostly for speed ATM).
939 - might compile and work in the presence of threads now.
940 - continuations have a saner syntax.
941 - no more memleaks.
942
943 0.07 Tue Jul 17 17:40:18 CEST 2001
944 - release candidate 1 ;)
945 - slightly nicer code.
946 - fixed a scheduling bug in Coro::Event.
947
948 0.06 Tue Jul 17 04:23:24 CEST 2001
949 - ok, I found the showstopper - the same sub must not be
950 re-entered in two different coroutines, otherwise => crash. I
951 see no easy solution to this problem, except by walking the call
952 chaing and saving/restoring the cv's, which is what I do now.
953 - memory leaks still latent, especially at thread termination.
954 - Coro::Event now works (haha).
955
956 0.05 Sun Jul 15 17:32:20 CEST 2001
957 - fixed a few issues in Coro::Event.
958 - I forgot to include Coro::Event in 0.04 :(:(:(
959
960 0.04 Sun Jul 15 05:24:59 CEST 2001
961 - @_ is now properly localized.
962 - Coro::State is now easier subclassable.
963 - Coro::Cont now coroutine-aware.
964 - Coro::Specific is a low-overhead module to create
965 coroutine-specific vars.
966 - Coro::Event provides a simple interface to Event.
967
968 0.03 Fri Jul 13 14:51:52 CEST 2001
969 - transfer() now implemented in XS (beware).
970 - new module Coro::Cont for really faked continuations.
971 - big internal architecture changes: Coro::State is now
972 really low-level and can thus be used to implement other
973 interesting things, While "Coro::" implements a process-like
974 model. Still crude and subject to change.
975 - $_ and $@ are no longer being localized.
976
977 0.02 Tue Jul 10 01:38:17 CEST 2001
978 - implemented "async" attribute.
979 - $_ and $@ are now localized.
980 - added Coro::Channel.
981 - more testcases, still no docs.
982
983 0.01 Tue Jul 3 02:18:41 CEST 2001
984 - original version; copied from Convert::Scalar.
985