ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.471
Committed: Thu Dec 4 04:31:41 2008 UTC (15 years, 5 months ago) by root
Branch: MAIN
Changes since 1.470: +1 -0 lines
Log Message:
*** empty log message ***

File Contents

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