ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.463
Committed: Mon Nov 24 07:55:28 2008 UTC (15 years, 5 months ago) by root
Branch: MAIN
CVS Tags: rel-5_1
Changes since 1.462: +4 -2 lines
Log Message:
5.1

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