ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.495
Committed: Sat Jun 20 07:16:31 2009 UTC (14 years, 11 months ago) by root
Branch: MAIN
Changes since 1.494: +3 -0 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 Revision history for Perl extension Coro.
2
3 TODO: should explore PerlIO::coroaio (perl leaks like hell).
4 TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7 TODO: unready_all
8 TODO: myhttpd header parsing
9 TODO: channel->maxsize(newsize)?
10
11 - clarify Coro::EV/Event/AnyEvent manpages: you have to actually run
12 the event loop somewhere to be able to poll for events
13 in the presence of ready threads.
14 - speed up ready queue management/context switching by using a linked
15 list instead of an array (~5-10%).
16 - implement "watch" command in Coro::Debug shells.
17 - for fun, implement time-slicing as an example in the manpage.
18
19 5.132 Fri May 29 09:00:39 CEST 2009
20 - do not keep a reference to the argument itself in
21 Coro::Semaphore::guard, as it could change later.
22 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
23 should support a prepare callback.
24
25 5.131 Mon Mar 16 23:20:37 CET 2009
26 - implement and document Coro->suspend, Coro->resume.
27 - fix Coro::Select implementation to not (often) close
28 the passed file descriptors (testcase provided by pippijn).
29
30 5.13 Mon Dec 15 21:51:42 CET 2008
31 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
32 - don't set diehook to C<undef> but instead to NULL, to avoid
33 spurious warnings.
34 - fix a lot of bugs in Coro::SemaphoreSet.
35 - Coro::SemaphoreSet will less often create a semaphore needlessly.
36 - add Coro::SemaphoreSet::count and wait methods.
37 - take advantage of the new Guard module.
38 - deprecate Coro::guard.
39 - try to fix the dreaded 01_unblock tests once more. I hate it when
40 testsuites need more fixing than the code they are supposed to test.
41 - croak in more cases when a required callback isn't resolvable.
42 - fix some minor issues in Coro::State->call/eval.
43 - use current coroutine context instead of a temporary one
44 when temporarily switching to another coroutine.
45 - do not call C-level on_destroy handlers during global destruction,
46 to avoid needless segfaults.
47
48 5.12 Sun Dec 7 13:30:38 CET 2008
49 - add default config for MirOS, which seems to be a bug-to-bug
50 compatible fork of openbsd ("world domination by releasing
51 openbsd cvs before the openbsd folks do it" or so :).
52 - free_padlist did destroy the names pad, not good, but didn't
53 seem to bother perl - this could fix issues such as eval ""
54 inside a function called from multiple coroutines.
55 - use a different method to detect destruction time.
56 - be more careful when freeing padlists just before global
57 destruction.
58 - fixed and expanded the call/cc example.
59 - renamed _terminate to _coro_run.
60 - new method Coro::Channel->shutdown.
61 - try pthreads on openbsd <4.4 (broken sigaltstack, will
62 pthreads fare better?).
63 - be less picky about destroying "a" running coroutine.
64
65 5.11 Tue Nov 25 21:49:05 CET 2008
66 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
67 working code with failed assertions, introducing more bugs than
68 it fixes, requiring elaborate workarounds :(
69
70 5.1 Mon Nov 24 08:54:59 CET 2008
71 - wrote a small introductory tutorial - Coro::Intro.
72 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
73 - Coro::Select did errornously dup the file descriptors
74 and didn't work with all AnyEvent backends.
75 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
76 LWP data transfers.
77 - disassociate c contexts from coro objects - this is agruably more
78 correct, but mostly allows sharing of cctxs between coro and state
79 objects, for added memory savings and speed increases.
80 - bumped $Coro::POOL_RSS up to 32kb by default.
81 - no longer set the optype to OP_CUSTOM, as B::* understandably
82 doesn't like this very much (and we *are* a type of entersub).
83 - implement state cloning, just to prove that call/cc can be done.
84 - automatically load Coro::AnyEvent in Coro::Handle.
85 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
86 do not have this method (and don't need it either).
87 - speed up generic anyevent methods in Coro::Handle by using rouse
88 callbacks.
89 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
90 others. It also makes the debugger happier, as you can trace
91 through the idle threads now.
92 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
93 - eg/event was pretty broken.
94 - better 5.8.6 compatibility.
95
96 5.0 Thu Nov 20 10:35:05 CET 2008
97 - NEW ARCHITECTURE: use the latest 4.x version if you experience
98 stability issues.
99 - bump API version to 7 - all dependents must be recompiled.
100 - removed timed_* functions - they were not being used anyways
101 and should be replaced by a more generic mechanism -
102 and were annoying to support anyways :)
103 - removed SemaphoreSet's waiter method - use sem method instead.
104 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
105 - async_pool did free a scalar value twice
106 ("Attempt to unreference...").
107 - fix a longstanding bug where calling terminate on a coro that
108 was waiting for a semaphore that was just becoming available
109 would cause a deadlock (semaphore would get into a state where
110 it was available but waiters were still blocked).
111 - calling throw on a coroutine that is waiting for a semaphore will
112 no longer make it acquire the semaphore (and thus leak a count).
113 - perl's process emulation is now not even theoretically supported
114 anymore.
115 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
116 conversion of callback-style to blocking-style.
117 - new methods $coro->schedule_to and ->cede_to, to specifically
118 schedule or cede to a specific coroutine.
119 - new function Coro::Semaphore::wait.
120 - use named constants in Coro::Channel (Richard Hundt).
121 - directly patch the entersub opcode calling SLF functions (cede,
122 transfer and so on). this does speed up context switching, but
123 more importanly, it frees us from the hardcoded behaviour of
124 entersub, so we might actually be able to return something from
125 those functions and atcually create new ones.
126 - take advantage of __builtin_frame_address on gcc.
127 - expose THX in coroapi (not sure whether this was a wise decision,
128 as "threaded" perls are running at half speed anyways).
129 - implement execute_slf (schedule-like-function) interface that makes
130 it possible to implement schedule-like-functions in XS.
131 - use new SLF interface to massively speed up Coro::EV by roughly a
132 factor of two.
133 - used new SLF interface to massively speed up Coro::Semaphore by a
134 factor of three.
135 - used new SLF interface to speed up Coro::AIO by roughly a factor of
136 four and reduce its memory usage considerably.
137 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
138 for a nice speedup and vastly more correct behaviour. Also implement
139 a new method "sem" to get at the underlying semaphore object.
140 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
141 (in comparison) 20-40% speedup.
142 - used new SLF interface to reimplement Coro::Signal gaining
143 some unknown (because I was too lazy), but certain, speedup, and also
144 making signals reliable for the first time.
145 - used new SLF interface and other optimisations to speed up async_pool
146 by a factor of two. It also doesn't rely on perl's exception mechanism
147 to exit anymore. The overhead for terminating an async_pool, coro over
148 a normal async is now very small.
149 - sped up coroutine creation/destruction by 40%.
150 - forgot to include Coro/libcoro/README in the dist for all these years.
151 - work around a freebsd pthreads bug (manual testcancel is required as
152 pthread_cond_wait isn't a cancellation point on freebsd).
153 - use new rouse functions to speed up and simplify Coro::BDB.
154 - make "prefer perl native functions" work with threaded perls.
155 - condense Coro::Debug ps output, hint at v and w flags.
156 - (libcoro) lots of minor cleanups and portability improvements.
157
158 4.914 Wed Nov 19 12:54:18 CET 2008
159 - fix a disastrous bug in the readline optimisation
160 introduced in 4.801.
161
162 4.913 Sat Nov 15 07:58:28 CET 2008
163 - async_pool did free a scalar value twice
164 ("Attempt to unreference...").
165
166 4.912 Thu Nov 13 18:31:23 CET 2008
167 - minor cleanups.
168 - use much larger stacks on linux and perl < 5.8.8.
169 - Coro::Debug::new_unix_server did not unlink the socket
170 when destroyed.
171
172 4.911 Tue Nov 11 04:26:17 CET 2008
173 - "port" to threaded perls.
174
175 4.91 Mon Nov 10 05:36:38 CET 2008
176 - the ->throw exception object no longer leaks.
177 - creating a new cctx leaked a scopestack entry (memleak).
178 - new coroutines didn't get created with a zero flags field
179 (unknown impact).
180 - calling ->throw on a not-yet-started coroutine should now work
181 instead of being ignored.
182 - ->throw is now supported on Coro::State objects.
183 - clean up cctx creation code a bit.
184 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
185
186 4.9 Sat Nov 8 17:45:27 CET 2008
187 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
188 - (libcoro) no longer leak threads in the experimental pthread backend,
189 also speed it up considerably.
190 - (libcoro) do not rely on makecontext passing void *'s unscathed.
191 - fix compiletime dependencies on libcoro in the Makefile.
192 - cctx_count wasn't always updated properly.
193 - Coro::State::cctx_stacksize wasn't applied correctly.
194 - new function Coro::State::cctx_max_idle.
195 - the default max number of idle C contexts is now 4.
196 - (libcoro) try harder to get _setjmp/_longjmp.
197 - (libcoro) cleanup and extend the libcoro API to officially
198 allow the creation of empty source contexts.
199 - very experimental workaround for the totally broken netbsd platform.
200 - tried to hack around openbsd bugs.
201
202 4.804 Wed Nov 5 16:36:00 CET 2008
203 - Coro::Debug::new_unix_server would not create a non-blocking listening
204 socket, sometimes causing freezes.
205 - (libcoro) fix misaligned stack points for setjmp and assembly
206 methods, which can cause crashes on x86/x86_64 with a sufficiently
207 aggressive compiler.
208 - new function: Coro::Debug::new_tcp_server.
209 - move ->throw into the Coro class because it only works on coro objects.
210
211 4.803 Mon Nov 3 17:16:12 CET 2008
212 - (libcoro) use a global asm statement to become independent of gcc
213 otpimisations for CORO_ASM (thanks to pippijn for the idea).
214 - try to workaround yet another broken bsd, this time dragonfly.
215
216 4.802 Thu Oct 30 10:56:12 CET 2008
217 - support -fno-omit-frame-pointer on x86 with the assembly method.
218 - tune 01_unblock.t tests a bit.
219
220 4.801 Wed Oct 22 18:33:37 CEST 2008
221 - improve readline speed for very long "lines".
222 - backport to 5.8.8.
223
224 4.8 Thu Oct 2 13:34:40 CEST 2008
225 - new function Coro::AIO::aio_wait.
226 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
227 - greatly speed up and reduce memory usage of Coro::AIO requests.
228 - implement some other ยต-optimisations.
229
230 4.749 Mon Sep 29 14:40:12 CEST 2008
231 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
232
233 4.748 Sat Sep 27 14:03:19 CEST 2008
234 - implement, but do not document, PerlIO::cede(granularity).
235 - Coro::Storable forgot to wrap Storable::pstore.
236 - work around the multitude of leaks and memory corruption
237 bugs in PerlIO::via by using our own C-level perliol. As a side
238 effect, Coro::Storable is now much, much, much faster.
239
240 4.747 Tue Sep 23 01:59:41 CEST 2008
241 - fix a per-cv memleak (one empty array was leaked per
242 code reference).
243 - avoid a crash in coro->call|rss when the coroutine was already
244 destroyed (most noticably when using Coro::Debug::ps :)
245 - also protect *Storable::FILE.
246 - push up default storable granularity to 20ms.
247
248 4.746 Sun Sep 21 03:22:20 CEST 2008
249 - be more insistent on locking Storable against reentrancy
250 in Coro::Storable.
251 - move swap_def?v and throw to Coro::State, as documented.
252
253 4.745 Thu Jul 24 00:14:38 CEST 2008
254 - remove debugging code related to MgPV_nolen_const, also try to
255 make it compile with perl 5.8.6 (yes, apple apparently loves
256 outdated software). Reported by John S.
257
258 4.744 Tue Jul 8 22:06:35 CEST 2008
259 - correctly provide default DIE/WARN handlers as documented.
260 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
261 implementations inside perl work fine for us.
262
263 4.743 Mon Jun 16 00:21:57 CEST 2008
264 - when using Coro::EV without running EV::loop it could
265 result in busy-waiting for events, this has been fixed.
266 - reduce codesize and improve performance by using EV_DEFAULT_UC.
267
268 4.742 Sat May 31 14:10:21 CEST 2008
269 - implement a workaround for (some) perl <5.8.8 versions.
270 - require EV 3.3+.
271
272 4.741 Fri May 30 23:33:09 CEST 2008
273 - tell netbsd how utterly broken their imitation of an OS is
274 and refuse to build by default if pthreads are in use.
275 - switch to "s" method on all bsd's by default, as their ucontext
276 stuff seems just too broken.
277 - fix a bug in Coro::Select.
278
279 4.74 Thu May 29 20:05:31 CEST 2008
280 - do not test Coro::LWP for lack of dependencies.
281 - do not test Coro::Select for lack of working perls.
282
283 4.73 Thu May 29 2008
284 - fix a bug in Coro::EV which would cause it to block despite
285 there being runnable coroutines.
286 - sprinkle "no warnings" freely over everything, also suppress
287 warnings for some other modules.
288 - fix typo in WSAEWOULDBLOCK.
289
290 4.72 Sun May 25 05:14:36 CEST 2008
291 - tweak META.yaml a bit, unfortunately, there is no documented way
292 to have optional dependencies with CPAN. doh :(
293 - avoid running some tests on windows because they would fail due to
294 perl bug (broken fork, broken pipes...).
295 - work around perl on windows bugs where perl returns undocumented
296 error codes for sysread, syswrite etc. by taking advantage
297 of AnyEvent's workaround for that problem.
298 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
299 a common perl implementation bug on windows.
300 - use unix domain sockets in testsuite to work around
301 common perl implementation bugs on widows (they are emulated by
302 tcp sockets on windows. ugh.)
303
304 4.71 Sat May 24 20:01:27 CEST 2008
305 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
306 - take advantage of async name resolution of AnyEvent::Util.
307 - work around brutal inet_aton override in Coro::LWP.
308 - take advantage of the readyhook in Coro::EV, for smoother
309 scheduling.
310
311 4.7 Sun May 11 00:32:19 CEST 2008
312 - completely reworked the Coro manpage.
313 - added Coro::AnyEvent, generic event loop integration.
314 - implement cancel, ready and kill commands in Coro::Debug.
315 - document find_coro in Coro::Debug.
316 - incompatible change: rename has_stack to has_cctx.
317 - Coro::AIO and Coro::BDB no longer force event model detection,
318 use AnyEvent::AIO and AnyEvent::BDB.
319
320 4.6 Sat Apr 26 10:05:14 CEST 2008
321 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
322 by the Coro module. It was a mistake to have it in the first place.
323 - (experimental) support for activestate perl 5.10 (method "w").
324 - (experimental) support for strawberry perl (method "a").
325 - coro_sigelem_set did not return a value although it had to,
326 actual impact unknown.
327
328 4.51 Mon Apr 14 13:28:27 CEST 2008
329 - make it compile again on 5.8.
330
331 4.50 Thu Apr 10 09:43:17 CEST 2008
332 - I did it twice! (see 4.49).
333
334 4.49 Sun Apr 6 21:23:31 CEST 2008
335 - grr, instead of compiling the recent changes on 5.10 only they
336 were compiled on 5.8 only.
337
338 4.48 Sun Apr 6 20:36:46 CEST 2008
339 - allow coroutine switches during eval's under 5.10.x, as it seems
340 the parser is a per-interpreter option now, so this is safe
341 (this might fix the odd crashing bug).
342 - drop support for 5.9.x versions: they are dead, jim.
343
344 4.47 Sun Apr 6 00:37:52 CEST 2008
345 - force cctx allocation on API calls: we know we need to force one
346 and gcc actually manages to confuse our heuristic nowadays,
347 leading to crashes and worse.
348 - document force_cctx.
349
350 4.46 Fri Apr 4 22:05:43 CEST 2008
351 - upgrade libcoro, resulting in pthread-backend (which was only created
352 to fulfill the rules of the programming languages shootout).
353
354 4.45 Thu Mar 13 11:55:36 CET 2008
355 - fix a file leaking bug in eg/mhyttpd that would allow
356 downloading of any file (reported by oesi).
357 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
358 (also add testcase).
359 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
360 apparently having loads of fun with that).
361 - implement $state->swap_defsv, ->swap_defav and document ->throw.
362
363 4.4 Wed Feb 13 16:44:29 CET 2008
364 - only assume .cfi pseudo isns exist on GNU/Linux.
365 - add get_timed method to Coro::Channels.
366 - fixed Coro::Debug manpage.
367 - perl 5.11 compatbility improvement based on patch by
368 Andreas Kรถnig.
369
370 4.37 Sun Jan 20 11:25:23 CET 2008
371 - soften the check for an active parser for perl 5.10 (avoiding
372 "Coro::State::transfer called while parsing" in many cases).
373
374 4.36 Sun Jan 13 10:53:56 CET 2008
375 - reset diehook when terminating from an async_pool as to not
376 trigger any __DIE__ handlers.
377
378 4.35 Sun Jan 13 04:14:13 CET 2008
379 - "bt" debug command now displays any exceptions
380 from longmess and also skips the innermost
381 stackframes, giving more useufl output.
382 - allow backtraces in coroutines blocked in Coro::EV,
383 at a <1% speed hit.
384 - handle localising of $SIG{__WARN__} and __DIE__
385 properly (with a proper amount of dirty hacking).
386
387 4.34 Sat Dec 22 17:49:53 CET 2007
388 - upgrade to EV version 2.0 API.
389
390 4.33 Mon Dec 17 08:36:12 CET 2007
391 - make Coro::AIO etc. loadable in the absence of EV.
392
393 4.32 Mon Dec 17 07:46:02 CET 2007
394 - majorly improved Coro::Handle's performance with Coro::EV.
395 - implemented the until now mythical Coro::BDB module.
396 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
397
398 4.31 Wed Dec 5 12:32:39 CET 2007
399 - remove warn statement form Coro::Util.
400
401 4.3 Tue Dec 4 20:33:14 CET 2007
402 - calls to the idle function could cause stack corruption
403 when the stack changed.
404 - do no longer rely on the presence of EV::DNS if EV is used
405 (because it is gone), but instead take avdantage of EV::ADNS
406 if available.
407 - add ($) prototypes to all functions in Coro::Storable.
408 - use a conventional (safer) idle callback in Coro::EV.
409 - do accept quickly in Coro::Debug to avoid endless loops.
410
411 4.22 Fri Nov 30 16:04:04 CET 2007
412 - really use optimised versions for Event and EV in Coro::Util
413 and Coro::Handle.
414
415 4.21 Sun Nov 25 10:48:59 CET 2007
416 - fix a spurious memory read.
417 - Coro::EV no longer keeps the eventloop "alive".
418
419 4.2 Fri Nov 9 20:47:05 CET 2007
420 - enable/disable tracing from a new coroutine, not a pooled one.
421 - fix a memleak in Coro::Event.
422 - removed killall call from fork_eval.
423 - made sure store_fd is already loaded so that fork_eval
424 does not have to parse autoload in each subprocess.
425 - only use assembly method if -O switch is in $Config{optimize}.
426 - add (optional) Coro::EV module, so far the best event loop module
427 directly supported by Coro.
428 - if the event model is EV, use EV::DNS to resolve
429 stuff in Coro::Util.
430 - don't get confused by multiple event notifications in Coro::Handle.
431 - initial support for EV (libevent interface).
432 - require Event and EV using configure_requires, to force their existance.
433
434 4.13 Wed Oct 24 07:26:45 CEST 2007
435 - add Coro::Storable::blocking_thaw.
436 - use a vastly more complicated technique to localise
437 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
438 - use a coroutine for the idle callback Coro::Event,
439 instead of running Event in the current coroutine context.
440 This also catches recursive invocations.
441 - actually report fork errors in gethostbyname and inet_aton.
442
443 4.11 Thu Oct 11 02:40:24 CEST 2007
444 - port to threaded perls.
445
446 4.1 Thu Oct 11 02:38:16 CEST 2007
447 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
448 be local to each coro (see Coro::State).
449 - incompatible change: for very deep reasons, cede and cede_notself
450 cannot return anything, so nothing will be returned.
451 - possibly bring back 5.10 compatibility (untested).
452 - work around stupid (and wrong) warning on 5.10 :(.
453 - overlay the saved state over the context stack. This saves
454 a few hundred bytes per coroutine on average and also
455 speeds up context switching a bit.
456 - further tune default stack sizes.
457 - (more) correctly calculate stack usage in coro_rss.
458 - Coro::Storable::blocking_* did not properly lock
459 resulting in races between coroutines.
460 - added Coro::Storable::guard.
461 - stopping to trace a coroutine could destroy the cctx of
462 an unrelated coroutine.
463 - explain the relationship between Perl and C coroutines in
464 more detail in Coro::State.
465 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
466 causing a fork per resolve. This also affected Coro::Socket.
467 - switch to a separate stack in $coro->call/eval to avoid
468 invalidating pointers.
469
470 4.03 Sat Oct 6 21:24:00 CEST 2007
471 - added Coro::throw method.
472 - minor code cleanups.
473
474 4.02 Sat Oct 6 02:36:47 CEST 2007
475 - fix a very minor per-coroutine memleak (a single codereference).
476 - fixed a bug where the currently in-use c context would be freed
477 prematurely (can happen only when programs change the stacksize
478 or use tracing).
479 - tracing can no longer keep a coro alive after it terminated.
480 - do static branch prediction in the common path for gcc. gives
481 about 2-5% speed improvement here.
482
483 4.01 Fri Oct 5 22:10:49 CEST 2007
484 - instead of recreating *a* standard output handle we simply
485 use STDOUT, which is faster and hopefully more robust.
486
487 4.0 Fri Oct 5 12:56:00 CEST 2007
488 - incompatibly changed Coro::Storable::freeze.
489 - major new feature: added Coro::Debug, for interactive coroutine
490 debugging, tracing and much more.
491 - major bug fix: unbelievable, but true: $_, $/ and many other
492 "saved" variables actually weren't being saved. This has been fixed,
493 of course, while increasing performance while losing all the save
494 flags.
495 - save flags are gone, and all the api functions dealing with them.
496 - added Coro::Semaphore::adjust.
497 - added Coro::Util::fork_eval.
498 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
499 - added Coro::killall.
500 - reduce initial stack sizes to allow for "micro-coroutines".
501 - better async_pool resource management, moved parts of async_pool
502 handling to XS (major speed improvement).
503 - actually croak before modifying important data structures.
504 - refuse to transfer while compiling.
505 - possibly support eval EXPR better now.
506 - enable assembly per default on linux+bsd x86+amd64.
507 - all internal members were renamed _something for easier subclassing.
508 - many minor tweaks.
509
510 3.63 Wed May 16 14:10:06 CEST 2007
511 - implement handcoded assembly for x86/amd64 SVR ABI.
512
513 3.62 Fri Apr 27 21:36:06 CEST 2007
514 - upgrade libcoro (which might set unwind info correctly).
515 - change default on linux to setjmp/longjmp.
516
517 3.61 Thu Apr 19 12:36:18 CEST 2007
518 - Coro::Storable caused an endless loop when thawing invalid
519 pst's sometimes.
520 - use a Semaphore in Coro::Storable, as Storable doesn't
521 seem to be reentrant (although it is documented to
522 be threadsafe...).
523 - fix Coro::Signal to bring back the original unreliable
524 but stateful semantics.
525 - fixed a lot of typos in Coro.pm (patch submitted by David
526 Steinbrunner, which applied flawlessly).
527
528 3.6 Sat Apr 14 17:13:31 CEST 2007
529 - added some bugfixes to get eg/myhttpd working again.
530 - added Coro::Storable for often-cede'ing freeze/thaw.
531 - try to do a clean exit when a coroutine calls exit
532 (EXPERIMENTAL).
533 - got rid of indirect call through _coro_init.
534 - updated the partly antique examples in eg/ to
535 work again and be a bit less magic, too.
536 - fixed Coro::Signal semantics to work as documented again.
537
538 3.55 Sun Mar 25 01:20:47 CET 2007
539 - add SAVE_DEFFH to save the default filehandle and enable
540 it by default.
541 - finally move socket-operations from Coro::Socket to Coro::Handle
542 to be able to unblock foreign sockets.
543 - add Coro::State::save_also and guarded_save.
544 - add count accessor to Coro::Semaphore.
545 - add Coro::State::cctx_stacksize.
546 - just for the fun of it, do not rely on implicit context,
547 which can dramatically improve performance, but people
548 using windows-process-emulation perls do not care much about
549 performance.
550
551 3.51 Sun Mar 4 14:18:04 CET 2007
552 - fixed a problem when you weakened references to Coro::State's
553 (patch by Yuval Kogman).
554
555 3.501 Wed Feb 28 12:44:07 CET 2007
556 - rename some global symbols as macosx from hell redefines
557 them without asking.
558
559 3.5 Tue Feb 13 20:22:53 CET 2007
560 - do AnyEvent model detection earlier, avoiding problems
561 caused by first using AnyEvent and later Coro::Event.
562 - implement and document Coro::Event event objects.
563 - fix a potential problem in Coro::Event causing crashes.
564 - initialise PL_comppad when creating a new coroutine,
565 avoids crashes on early coro destruction.
566
567 3.41 Mon Jan 22 19:19:49 CET 2007
568 - readline on Coro::Handle did not support undefined $/,
569 nor did it deliver partial lines on EOF or error.
570 - implement malloc fallback for stack allocation because
571 stupid broken idiotic OSX has a stupid broken
572 idiotic fits-the-whole-os mmap "implementation" and
573 my dick feels longer if Coro is portable even to
574 obsolete platforms.
575
576 3.4 Fri Jan 19 21:52:54 CET 2007
577 - remove t/09_timer.t, as it isn't really testing much
578 but was rather flaky in practise.
579 - async_pool coro would keep arguments and callback alive until
580 it was reused.
581 - cancellation of a coroutine could cause spurious idle calls
582 in cede_notself.
583
584 3.3 Sat Jan 6 03:45:00 CET 2007
585 - implement $coro->on_destroy.
586 - Coro::Event blocking semantics have been changed,
587 documented and - hopefully - improved.
588 - fix nice adding, not subtracting, from priority.
589 - fix ->prio and api_is_ready (patch by Mark Hinds).
590 - fixed an assert ("... == prev__cctx->idle_te")
591 that could errronously trigger.
592 - fix various large and small memleaks.
593 - use a (hopefully) more stable cancel implementation
594 that immediately frees the coroutine data.
595 - cede/cede_notself return a status now.
596 - added Coro::guard function.
597 - added a global coroutine pool for jobs (on my machine,
598 I can create and execute 48k simple coros/s with async,
599 and 128k coros with async_pool).
600 - Coro::AIO now uses the coroutine priority as io priority.
601
602 3.2 Fri Dec 22 05:07:09 CET 2006
603 - improve portability to slightly older perls.
604 - use cleaner coroutine destruction.
605 - simplify configuration for users.
606 - optionally (unrecommended) prefer perl functions over
607 their coro replacements.
608
609 3.11 Tue Dec 5 13:11:24 CET 2006
610 - fixed some bogus assert's, but as perl.h disables assert even
611 without NDEBUG (thank you very much), not too many people should
612 notice (that did include myself). Andreas Kรถnig noticed, though :)
613 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
614 it. Also noticed by Andreas Kรถnig :)
615 - save/restore tainted status.
616 - verified to pass the testsuite on my 5.9.5.
617
618 3.1 Mon Dec 4 23:03:40 CET 2006
619 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
620 - incompatible change: transfer flags are now per-state.
621 - give a better error message on deadlock.
622 - document Coro::nready.
623 - enhanced testsuite.
624
625 3.01 Sun Dec 3 23:47:42 CET 2006
626 - forgot to include Coro::Timer.
627
628 3.0 Sun Dec 3 22:57:25 CET 2006
629 - the "FINALLY COMPLETELY STABLE" release (coming soon:
630 the "FAMOUS LAST WORDS" release).
631 - implement a new stack sharing algorithm, which uses a stack
632 pool (size currently hardcoded).
633 - make stack sharing mandatory (it no longer uses a heuristic).
634 - eval/die no longer cause weird problems under heavy use.
635 - Coro::Event could cause livelocks if it was used but
636 no Coro::Event watchers were used.
637 - Coro::Event now uses asynccheck as prepare does not
638 check for changed watchers.
639 - Coro::Event allows multiple waiting coros and will wake up one
640 per event.
641 - Coro::Event should be cleaner and more efficient now.
642 - new utility function Coro::unblock_sub.
643 - document the sad fact that Event is no longer reentrant.
644 - putting a coroutine into the ready queue twice could under
645 some circumstances lead to stack corruption.
646 - minor incompatibility: subclassing Coro::State is supported
647 directly now without going through a _coro_state member.
648 - state/coro switching is much faster now.
649 - very minor optimisations and code/documentation cleanup.
650 - avoid problems due to compiler inlining.
651 - removed timers from Coro::Timer -> use AnyEvent instead.
652 - replaced Coro::idle coroutine by (cleaner) idle handler.
653 - updated to newest libcoro.
654 - implement enhanced support for valgrind.
655 - implement is_ready and return value for ready.
656 - removed Coro::Cont, it was a misnomer (it's generators),
657 and it was somewhat annoying to get it right. Will come back
658 if somebody asks for it :)
659 - many ->wait methods and Coro::Event could return
660 spuriously without the event having happened.
661
662 2.5 Tue Nov 7 12:22:33 CET 2006
663 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
664 use AnyEvent, moved them to Coro/.
665 - added Coro::LWP which contains all the uglyness required to
666 make LWP non-blocking.
667 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
668 - fixed another bug in Coro::Select when the timeout was undef.
669 - reuse PL_start_env for all coros, saves some memory per coroutine.
670 - manage PL_top_env differently, hopefully to avoid panic: top_env.
671 - timeout argument was not properly used in Coro::Socket.
672 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
673 - emulate undocumented(!) functionality of IO::Socket required
674 by LWP(!!).
675 - updated eg/lwp to work with newer lwp's.
676 - remove "FATAL: uncaught exception" prefix. Coroutines that die
677 kill the whole process, just as exceptions in the main "coroutine"
678 did already.
679
680 2.1 Wed Nov 1 23:01:13 CET 2006
681 - fix a long-standing bug in Coro::Select where select with
682 zero timeout would instead change the current default filehandle.
683 - use a simpler and hopefully more robust way to clone padlists
684 (uses less memory and a perl function instead of our own).
685 - coro can now create a stack guard area on many architectures.
686 - Coro::AIO properly reexports additional functions from IO::AIO.
687 - updated libcoro with a workaround for OS X,
688 pach and testing by Michael Schwern.
689
690 2.0 Tue Oct 24 05:47:17 CEST 2006
691 - support additional aio requests in Coro::AIO.
692
693 1.9 (never properly released due to a glitch)
694
695 1.8 Thu Feb 2 00:59:06 CET 2006
696 - applied suggested patch by SAMV to avoid problems during stupid
697 mark & sweep gc run.
698 - applied patch by Scott Walters for 5.9.3 compatibility.
699
700 1.7 Tue Dec 27 01:41:58 CET 2005
701 - added Coro::AIO, a thin wrapper around IO::AIO.
702 - improved Makefile.PL explanations.
703
704 1.6 Mon Dec 12 21:30:05 CET 2005
705 - additionally save PM_curpm might fix as-of-yet
706 unseen problems with regex matches being attributed to the
707 wrong package.
708 - add t/10_bugs.t, which currently checks against imho broken
709 perls that use 0.26MB of stack space per Perl_magic_get
710 invocation (newer linuxes) as opposed to the 0.0002MB perl
711 normally uses.
712 - make stacksize configurable for the ultimate debian experience.
713
714 1.51 Mon Dec 12 18:48:36 CET 2005
715 - remove debugging warn() accidentally left in Coro::Select.
716
717 1.5 Tue Nov 29 13:32:44 CET 2005
718 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
719 - fix Coro::Select 'select'.
720 - strict'ify some modules.
721 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
722 - use XSLoader in selected modules.
723 - remove some 5.6 compatibility cruft.
724
725 1.4 Tue Sep 6 00:11:05 CEST 2005
726 - libcoro did not take into account the trampoline on amd64,
727 when 'l'inux method was used.
728
729 1.31 Tue Aug 30 23:31:33 CEST 2005
730 - some portability fixes/workarounds.
731
732 1.3 Sat Aug 20 03:08:56 CEST 2005
733 - no code changes. module seems to work fine.
734 - improve Coro::State docs, remove reference to nonexisting
735 Coro::State::flush method.
736 - no longer autodetect windows, present it as an option instead.
737
738 1.2 Mon May 16 02:00:55 CEST 2005
739 - included libcoro.c earlier in State.xs, might improve portability.
740 - use a faster and possibly more stable Coro::Cont implementation.
741 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
742 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
743 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
744
745 1.11 Thu Mar 3 18:00:52 CET 2005
746 - change of contact address.
747
748 1.1 Tue Feb 22 20:51:16 CET 2005
749 - support [l]inux method on amd64.
750 - allow some context switches while compiling/eval'ing.
751
752 1.0 Tue Aug 10 03:47:27 CEST 2004
753 - Event 0.89 fixes the bug again.
754 - fix the remaining(??) memleak. I didn't free the cache-AV for
755 padlists, so one AV per created anon-sub leaked.
756
757 0.97 Fri May 14 15:23:32 CEST 2004
758 - remove stupid gcc extension that was an experiment only anyway.
759
760 0.96 Thu May 13 18:09:29 CEST 2004
761 - tell people if they have yet another broken Event variant.
762 - made Coro to work _better_ with threads (it seems that, as long
763 as you use Coros in one thread only everything will work fine.
764 Looking at the perl source the problem might be missing locking
765 between Coro and some perl internal routines. That's not easily
766 fixable).
767 - (hopefully) got rid of the static xnext inside transfer.
768
769 0.95 Sun Feb 15 03:20:28 CET 2004
770 - removed Coro::State::flush, implemented a new and much
771 faster caching mechanism by attaching magic to the CV.
772 - a memleak with (real) closures remains.
773 - dramatically reduced size of stacks to 65536 (linux/x86), after
774 observing that even myhttpd never had more than a kilobyte
775 of stack in use (stack is still only physically allocated in
776 pagesize increments for systems suporting mmap).
777 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
778 - fix for threaded perl.
779
780 0.9 Sun Nov 23 04:32:39 CET 2003
781 - fix a few very hard to track down but (of course) stupid
782 memory leaks.
783 - stringify version number to avoid locale problems :(.
784 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
785 hopefully ignoring it "just works".
786
787 0.8 Wed Nov 5 19:38:40 CET 2003
788 - port to perl5.005_03, but only 5.8.x+ is supported!
789 - honor LocalAddr even without LocalPort in Coro::Socket.
790
791 0.7 Tue May 27 03:12:38 CEST 2003
792 - the version jump indicates some level of testing, not gobs
793 of new features.
794 - uh... I found the prompt function in ExtUtils::MakeMaker.
795 Highly correct stuff, that is...
796 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
797 the fix is not well understood by the author, so beware :(.
798
799 0.652 Thu May 8 02:54:46 CEST 2003
800 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
801 cause version 4 doesn't have ucontext.
802 - Benjamin Reed reported that setjmp works fine on darwin,
803 so preselect it.
804
805 0.651 Sat Mar 29 15:00:23 CET 2003
806 - fix a bug in Coro::Handle where some bytes could get lost
807 on reading, reported by jason@nichego.net.
808
809 0.65 Sun Mar 23 00:08:26 CET 2003
810 - added a README.linux-glibc.
811 - new module Coro::Select.
812 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
813 (I never claimed that I know what I am doing ;).
814 - set default to setjmp/longjmp on non-x86-linux.
815
816 0.6 Thu Nov 21 11:09:06 CET 2002
817 - made lazy context switching the default.
818 - fixed the bug where SemaphoreSet::guard
819 was the same as timed_guard and timed_guard was missing.
820 - fix a memleak in Coro::Timer.
821 - reclassified context sharing as not experimental.
822
823 0.534 Sun Apr 14 03:05:12 CEST 2002
824 - fixed backspace => tab in header parsing (myhttpd).
825 - added eg/readline, for elmex.
826 - Coro::Event now calls ->start on first call instead of
827 ->again, which created wrong timeouts for the first ->next.
828 - fixed a bug where Coro::Socket returned a closed socket instead
829 of nothing on ECONNREFUSED and similar error conditions.
830
831 0.533 Mon Feb 18 18:49:40 CET 2002
832 - adapt to finally working Event-0.85.
833 - fixed buggy event initialization (trapped under windows, but how
834 could it possibly have worked anywhere?)
835
836 0.532 Wed Jan 16 02:45:32 CET 2002
837 - removed the reference to pp_entersub, might work on
838 (some) win32 perls now (testsuite works under cygwin,
839 without Event, which freezes).
840
841 0.531 Mon Dec 10 22:18:44 CET 2001
842 - Fixed a bug in SemaphoreSet::timed_down.
843
844 0.53 Tue Nov 27 21:11:13 CET 2001
845 - some tests for Event.
846 - slightly different internal architecture to get refcnt into the object
847 and not the reference. might fix a few bugs, certainly did introduce
848 new ones.
849 - Coro::Timer (independent of Event).
850 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
851
852 0.52 Tue Nov 6 21:36:18 CET 2001
853 - ported to cygwin (trivial).
854 patch by Gerrit P. Haase <gp@familiehaase.de>.
855 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
856
857 0.51 Thu Nov 1 20:39:01 CET 2001
858 - terminate/cancel now work properly (otherwise termination
859 could cause a "next coroutine is not and contains not..."
860 error.
861 - added Coro::Socket::shutdown.
862 - Coro::Event::loop no is the same as Event::loop.
863 - implemented terminate with args + join.
864
865 0.5 Fri Sep 28 16:15:35 CEST 2001
866 - fixed "print" on a Coro::Handle. The print method worked.
867 - small tweaks (seem to reduce memory consumption a lot)
868 in various modules.
869 - splendid use of "no warnings" scattered throughout.
870 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
871
872 0.49 Sun Sep 16 02:42:45 CEST 2001
873 - changed some method calls to function calls for speed
874 inside Coro::Handle.
875 - make Coro::Handle use an array instead of a hash for
876 speed reasons.
877 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
878 sjlj and ucontext should now work.
879 - IRIX-specific port for libcoro.
880 - swapped order of accept results to match IO::Socket.
881 - changed getsock/peername to sock/peername to match IO::Socket.
882 - fixed a bug that caused segfault when returning to main task
883 under some circumstances.
884 - other bugfixes.
885
886 0.45 Sun Sep 2 02:54:01 CEST 2001
887 - new method Coro::Handle::timeout.
888 - corrected speling of set...name to get...name in Coro::Socket.
889 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
890 - Coro::Handle now supports fileno and FILENO.
891 - added eg/myhttpd, a web server.
892 - Coro::Socket now accepts numeric ports as well as "port(number)"
893 syntax.
894 - moved some scheduling primitives into xs code. more to come.
895 - new simple priority system.
896 - implemented Coro::Event using XS, almost four times faster!
897 - small memory corruption problem fixed (boy that was difficult).
898
899 0.13 Wed Aug 8 16:53:07 CEST 2001
900 - new method Coro::Handle::readline.
901 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
902 (see eg/lwp on how to make LWP non-blocking).
903 - renamed Coro::Socket::new_inet to new.
904 - added Coro::Util, some utility functions.
905 - die/eval now works better (still get annoying (but true ;)
906 "Callback called exit" messages).
907 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
908 - much better Coro::Socket support.
909
910 0.12 Fri Jul 27 04:19:01 CEST 2001
911 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
912 - very experimental stack sharing algorithm. not 100% safe but
913 should work well in practise ;)
914 - added Coro::RWLock.
915 - Coro::Specific now works.
916
917 0.11 Tue Jul 24 22:49:21 CEST 2001
918 - added specialized hack for newer and older linux versions (fast).
919 - renamed Coro::Event::IO to Coro::Handle.
920 - new module Coro::Socket.
921
922 0.1 Tue Jul 24 01:47:53 CEST 2001
923 - release candidate 3. A rather new internal structure :(
924 - the great renaming: Damian Conway gave me a suitable replacement
925 for yield.
926 - added Coro/libcoro, a portable coroutine implementation for C
927 (not even perl-dependent!!!), which can be used by Coro::State.
928 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
929 perl-5.6.
930 - new function Coro::Event::idle.
931 - the idle process is now overriden
932 by default in Coro::Event.
933 - Coro::Channel now enforces the size.
934 - canceling events now works in all cases.
935 - Coro state now include $_ and $@.
936 - yet another bug workaround that I do not really understand :(
937 - new module Coro::Event::IO (very undocumented) to do
938 non-blocking i/o.
939 - performance tuning.
940
941 0.08 Thu Jul 19 06:13:25 CEST 2001
942 - release candidate 2 (new functionality).
943 - add Coro::State::flush function.
944 - transfer now supports a flags value (mostly for speed ATM).
945 - might compile and work in the presence of threads now.
946 - continuations have a saner syntax.
947 - no more memleaks.
948
949 0.07 Tue Jul 17 17:40:18 CEST 2001
950 - release candidate 1 ;)
951 - slightly nicer code.
952 - fixed a scheduling bug in Coro::Event.
953
954 0.06 Tue Jul 17 04:23:24 CEST 2001
955 - ok, I found the showstopper - the same sub must not be
956 re-entered in two different coroutines, otherwise => crash. I
957 see no easy solution to this problem, except by walking the call
958 chaing and saving/restoring the cv's, which is what I do now.
959 - memory leaks still latent, especially at thread termination.
960 - Coro::Event now works (haha).
961
962 0.05 Sun Jul 15 17:32:20 CEST 2001
963 - fixed a few issues in Coro::Event.
964 - I forgot to include Coro::Event in 0.04 :(:(:(
965
966 0.04 Sun Jul 15 05:24:59 CEST 2001
967 - @_ is now properly localized.
968 - Coro::State is now easier subclassable.
969 - Coro::Cont now coroutine-aware.
970 - Coro::Specific is a low-overhead module to create
971 coroutine-specific vars.
972 - Coro::Event provides a simple interface to Event.
973
974 0.03 Fri Jul 13 14:51:52 CEST 2001
975 - transfer() now implemented in XS (beware).
976 - new module Coro::Cont for really faked continuations.
977 - big internal architecture changes: Coro::State is now
978 really low-level and can thus be used to implement other
979 interesting things, While "Coro::" implements a process-like
980 model. Still crude and subject to change.
981 - $_ and $@ are no longer being localized.
982
983 0.02 Tue Jul 10 01:38:17 CEST 2001
984 - implemented "async" attribute.
985 - $_ and $@ are now localized.
986 - added Coro::Channel.
987 - more testcases, still no docs.
988
989 0.01 Tue Jul 3 02:18:41 CEST 2001
990 - original version; copied from Convert::Scalar.
991