ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.486
Committed: Sat Feb 7 22:24:51 2009 UTC (15 years, 3 months ago) by root
Branch: MAIN
Changes since 1.485: +2 -0 lines
Log Message:
*** empty log message ***

File Contents

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