ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.485
Committed: Wed Dec 17 04:57:07 2008 UTC (15 years, 5 months ago) by root
Branch: MAIN
Changes since 1.484: +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
8 TODO: suspend/resume, unready_all etc.
9 - implement and document Coro->suspend, Coro->resume.
10
11 5.13 Mon Dec 15 21:51:42 CET 2008
12 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
13 - don't set diehook to C<undef> but instead to NULL, to avoid
14 spurious warnings.
15 - fix a lot of bugs in Coro::SemaphoreSet.
16 - Coro::SemaphoreSet will less often create a semaphore needlessly.
17 - add Coro::SemaphoreSet::count and wait methods.
18 - take advantage of the new Guard module.
19 - deprecate Coro::guard.
20 - try to fix the dreaded 01_unblock tests once more. I hate it when
21 testsuites need more fixing than the code they are supposed to test.
22 - croak in more cases when a required callback isn't resolvable.
23 - fix some minor issues in Coro::State->call/eval.
24 - use current coroutine context instead of a temporary one
25 when temporarily switching to another coroutine.
26 - do not call C-level on_destroy handlers during global destruction,
27 to avoid needless segfaults.
28
29 5.12 Sun Dec 7 13:30:38 CET 2008
30 - add default config for MirOS, which seems to be a bug-to-bug
31 compatible fork of openbsd ("world domination by releasing
32 openbsd cvs before the openbsd folks do it" or so :).
33 - free_padlist did destroy the names pad, not good, but didn't
34 seem to bother perl - this could fix issues such as eval ""
35 inside a function called from multiple coroutines.
36 - use a different method to detect destruction time.
37 - be more careful when freeing padlists just before global
38 destruction.
39 - fixed and expanded the call/cc example.
40 - renamed _terminate to _coro_run.
41 - new method Coro::Channel->shutdown.
42 - try pthreads on openbsd <4.4 (broken sigaltstack, will
43 pthreads fare better?).
44 - be less picky about destroying "a" running coroutine.
45
46 5.11 Tue Nov 25 21:49:05 CET 2008
47 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
48 working code with failed assertions, introducing more bugs than
49 it fixes, requiring elaborate workarounds :(
50
51 5.1 Mon Nov 24 08:54:59 CET 2008
52 - wrote a small introductory tutorial - Coro::Intro.
53 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
54 - Coro::Select did errornously dup the file descriptors
55 and didn't work with all AnyEvent backends.
56 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
57 LWP data transfers.
58 - disassociate c contexts from coro objects - this is agruably more
59 correct, but mostly allows sharing of cctxs between coro and state
60 objects, for added memory savings and speed increases.
61 - bumped $Coro::POOL_RSS up to 32kb by default.
62 - no longer set the optype to OP_CUSTOM, as B::* understandably
63 doesn't like this very much (and we *are* a type of entersub).
64 - implement state cloning, just to prove that call/cc can be done.
65 - automatically load Coro::AnyEvent in Coro::Handle.
66 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
67 do not have this method (and don't need it either).
68 - speed up generic anyevent methods in Coro::Handle by using rouse
69 callbacks.
70 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
71 others. It also makes the debugger happier, as you can trace
72 through the idle threads now.
73 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
74 - eg/event was pretty broken.
75 - better 5.8.6 compatibility.
76
77 5.0 Thu Nov 20 10:35:05 CET 2008
78 - NEW ARCHITECTURE: use the latest 4.x version if you experience
79 stability issues.
80 - bump API version to 7 - all dependents must be recompiled.
81 - removed timed_* functions - they were not being used anyways
82 and should be replaced by a more generic mechanism -
83 and were annoying to support anyways :)
84 - removed SemaphoreSet's waiter method - use sem method instead.
85 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
86 - async_pool did free a scalar value twice
87 ("Attempt to unreference...").
88 - fix a longstanding bug where calling terminate on a coro that
89 was waiting for a semaphore that was just becoming available
90 would cause a deadlock (semaphore would get into a state where
91 it was available but waiters were still blocked).
92 - calling throw on a coroutine that is waiting for a semaphore will
93 no longer make it acquire the semaphore (and thus leak a count).
94 - perl's process emulation is now not even theoretically supported
95 anymore.
96 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
97 conversion of callback-style to blocking-style.
98 - new methods $coro->schedule_to and ->cede_to, to specifically
99 schedule or cede to a specific coroutine.
100 - new function Coro::Semaphore::wait.
101 - use named constants in Coro::Channel (Richard Hundt).
102 - directly patch the entersub opcode calling SLF functions (cede,
103 transfer and so on). this does speed up context switching, but
104 more importanly, it frees us from the hardcoded behaviour of
105 entersub, so we might actually be able to return something from
106 those functions and atcually create new ones.
107 - take advantage of __builtin_frame_address on gcc.
108 - expose THX in coroapi (not sure whether this was a wise decision,
109 as "threaded" perls are running at half speed anyways).
110 - implement execute_slf (schedule-like-function) interface that makes
111 it possible to implement schedule-like-functions in XS.
112 - use new SLF interface to massively speed up Coro::EV by roughly a
113 factor of two.
114 - used new SLF interface to massively speed up Coro::Semaphore by a
115 factor of three.
116 - used new SLF interface to speed up Coro::AIO by roughly a factor of
117 four and reduce its memory usage considerably.
118 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
119 for a nice speedup and vastly more correct behaviour. Also implement
120 a new method "sem" to get at the underlying semaphore object.
121 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
122 (in comparison) 20-40% speedup.
123 - used new SLF interface to reimplement Coro::Signal gaining
124 some unknown (because I was too lazy), but certain, speedup, and also
125 making signals reliable for the first time.
126 - used new SLF interface and other optimisations to speed up async_pool
127 by a factor of two. It also doesn't rely on perl's exception mechanism
128 to exit anymore. The overhead for terminating an async_pool, coro over
129 a normal async is now very small.
130 - sped up coroutine creation/destruction by 40%.
131 - forgot to include Coro/libcoro/README in the dist for all these years.
132 - work around a freebsd pthreads bug (manual testcancel is required as
133 pthread_cond_wait isn't a cancellation point on freebsd).
134 - use new rouse functions to speed up and simplify Coro::BDB.
135 - make "prefer perl native functions" work with threaded perls.
136 - condense Coro::Debug ps output, hint at v and w flags.
137 - (libcoro) lots of minor cleanups and portability improvements.
138
139 4.914 Wed Nov 19 12:54:18 CET 2008
140 - fix a disastrous bug in the readline optimisation
141 introduced in 4.801.
142
143 4.913 Sat Nov 15 07:58:28 CET 2008
144 - async_pool did free a scalar value twice
145 ("Attempt to unreference...").
146
147 4.912 Thu Nov 13 18:31:23 CET 2008
148 - minor cleanups.
149 - use much larger stacks on linux and perl < 5.8.8.
150 - Coro::Debug::new_unix_server did not unlink the socket
151 when destroyed.
152
153 4.911 Tue Nov 11 04:26:17 CET 2008
154 - "port" to threaded perls.
155
156 4.91 Mon Nov 10 05:36:38 CET 2008
157 - the ->throw exception object no longer leaks.
158 - creating a new cctx leaked a scopestack entry (memleak).
159 - new coroutines didn't get created with a zero flags field
160 (unknown impact).
161 - calling ->throw on a not-yet-started coroutine should now work
162 instead of being ignored.
163 - ->throw is now supported on Coro::State objects.
164 - clean up cctx creation code a bit.
165 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
166
167 4.9 Sat Nov 8 17:45:27 CET 2008
168 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
169 - (libcoro) no longer leak threads in the experimental pthread backend,
170 also speed it up considerably.
171 - (libcoro) do not rely on makecontext passing void *'s unscathed.
172 - fix compiletime dependencies on libcoro in the Makefile.
173 - cctx_count wasn't always updated properly.
174 - Coro::State::cctx_stacksize wasn't applied correctly.
175 - new function Coro::State::cctx_max_idle.
176 - the default max number of idle C contexts is now 4.
177 - (libcoro) try harder to get _setjmp/_longjmp.
178 - (libcoro) cleanup and extend the libcoro API to officially
179 allow the creation of empty source contexts.
180 - very experimental workaround for the totally broken netbsd platform.
181 - tried to hack around openbsd bugs.
182
183 4.804 Wed Nov 5 16:36:00 CET 2008
184 - Coro::Debug::new_unix_server would not create a non-blocking listening
185 socket, sometimes causing freezes.
186 - (libcoro) fix misaligned stack points for setjmp and assembly
187 methods, which can cause crashes on x86/x86_64 with a sufficiently
188 aggressive compiler.
189 - new function: Coro::Debug::new_tcp_server.
190 - move ->throw into the Coro class because it only works on coro objects.
191
192 4.803 Mon Nov 3 17:16:12 CET 2008
193 - (libcoro) use a global asm statement to become independent of gcc
194 otpimisations for CORO_ASM (thanks to pippijn for the idea).
195 - try to workaround yet another broken bsd, this time dragonfly.
196
197 4.802 Thu Oct 30 10:56:12 CET 2008
198 - support -fno-omit-frame-pointer on x86 with the assembly method.
199 - tune 01_unblock.t tests a bit.
200
201 4.801 Wed Oct 22 18:33:37 CEST 2008
202 - improve readline speed for very long "lines".
203 - backport to 5.8.8.
204
205 4.8 Thu Oct 2 13:34:40 CEST 2008
206 - new function Coro::AIO::aio_wait.
207 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
208 - greatly speed up and reduce memory usage of Coro::AIO requests.
209 - implement some other ยต-optimisations.
210
211 4.749 Mon Sep 29 14:40:12 CEST 2008
212 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
213
214 4.748 Sat Sep 27 14:03:19 CEST 2008
215 - implement, but do not document, PerlIO::cede(granularity).
216 - Coro::Storable forgot to wrap Storable::pstore.
217 - work around the multitude of leaks and memory corruption
218 bugs in PerlIO::via by using our own C-level perliol. As a side
219 effect, Coro::Storable is now much, much, much faster.
220
221 4.747 Tue Sep 23 01:59:41 CEST 2008
222 - fix a per-cv memleak (one empty array was leaked per
223 code reference).
224 - avoid a crash in coro->call|rss when the coroutine was already
225 destroyed (most noticably when using Coro::Debug::ps :)
226 - also protect *Storable::FILE.
227 - push up default storable granularity to 20ms.
228
229 4.746 Sun Sep 21 03:22:20 CEST 2008
230 - be more insistent on locking Storable against reentrancy
231 in Coro::Storable.
232 - move swap_def?v and throw to Coro::State, as documented.
233
234 4.745 Thu Jul 24 00:14:38 CEST 2008
235 - remove debugging code related to MgPV_nolen_const, also try to
236 make it compile with perl 5.8.6 (yes, apple apparently loves
237 outdated software). Reported by John S.
238
239 4.744 Tue Jul 8 22:06:35 CEST 2008
240 - correctly provide default DIE/WARN handlers as documented.
241 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
242 implementations inside perl work fine for us.
243
244 4.743 Mon Jun 16 00:21:57 CEST 2008
245 - when using Coro::EV without running EV::loop it could
246 result in busy-waiting for events, this has been fixed.
247 - reduce codesize and improve performance by using EV_DEFAULT_UC.
248
249 4.742 Sat May 31 14:10:21 CEST 2008
250 - implement a workaround for (some) perl <5.8.8 versions.
251 - require EV 3.3+.
252
253 4.741 Fri May 30 23:33:09 CEST 2008
254 - tell netbsd how utterly broken their imitation of an OS is
255 and refuse to build by default if pthreads are in use.
256 - switch to "s" method on all bsd's by default, as their ucontext
257 stuff seems just too broken.
258 - fix a bug in Coro::Select.
259
260 4.74 Thu May 29 20:05:31 CEST 2008
261 - do not test Coro::LWP for lack of dependencies.
262 - do not test Coro::Select for lack of working perls.
263
264 4.73 Thu May 29 2008
265 - fix a bug in Coro::EV which would cause it to block despite
266 there being runnable coroutines.
267 - sprinkle "no warnings" freely over everything, also suppress
268 warnings for some other modules.
269 - fix typo in WSAEWOULDBLOCK.
270
271 4.72 Sun May 25 05:14:36 CEST 2008
272 - tweak META.yaml a bit, unfortunately, there is no documented way
273 to have optional dependencies with CPAN. doh :(
274 - avoid running some tests on windows because they would fail due to
275 perl bug (broken fork, broken pipes...).
276 - work around perl on windows bugs where perl returns undocumented
277 error codes for sysread, syswrite etc. by taking advantage
278 of AnyEvent's workaround for that problem.
279 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
280 a common perl implementation bug on windows.
281 - use unix domain sockets in testsuite to work around
282 common perl implementation bugs on widows (they are emulated by
283 tcp sockets on windows. ugh.)
284
285 4.71 Sat May 24 20:01:27 CEST 2008
286 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
287 - take advantage of async name resolution of AnyEvent::Util.
288 - work around brutal inet_aton override in Coro::LWP.
289 - take advantage of the readyhook in Coro::EV, for smoother
290 scheduling.
291
292 4.7 Sun May 11 00:32:19 CEST 2008
293 - completely reworked the Coro manpage.
294 - added Coro::AnyEvent, generic event loop integration.
295 - implement cancel, ready and kill commands in Coro::Debug.
296 - document find_coro in Coro::Debug.
297 - incompatible change: rename has_stack to has_cctx.
298 - Coro::AIO and Coro::BDB no longer force event model detection,
299 use AnyEvent::AIO and AnyEvent::BDB.
300
301 4.6 Sat Apr 26 10:05:14 CEST 2008
302 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
303 by the Coro module. It was a mistake to have it in the first place.
304 - (experimental) support for activestate perl 5.10 (method "w").
305 - (experimental) support for strawberry perl (method "a").
306 - coro_sigelem_set did not return a value although it had to,
307 actual impact unknown.
308
309 4.51 Mon Apr 14 13:28:27 CEST 2008
310 - make it compile again on 5.8.
311
312 4.50 Thu Apr 10 09:43:17 CEST 2008
313 - I did it twice! (see 4.49).
314
315 4.49 Sun Apr 6 21:23:31 CEST 2008
316 - grr, instead of compiling the recent changes on 5.10 only they
317 were compiled on 5.8 only.
318
319 4.48 Sun Apr 6 20:36:46 CEST 2008
320 - allow coroutine switches during eval's under 5.10.x, as it seems
321 the parser is a per-interpreter option now, so this is safe
322 (this might fix the odd crashing bug).
323 - drop support for 5.9.x versions: they are dead, jim.
324
325 4.47 Sun Apr 6 00:37:52 CEST 2008
326 - force cctx allocation on API calls: we know we need to force one
327 and gcc actually manages to confuse our heuristic nowadays,
328 leading to crashes and worse.
329 - document force_cctx.
330
331 4.46 Fri Apr 4 22:05:43 CEST 2008
332 - upgrade libcoro, resulting in pthread-backend (which was only created
333 to fulfill the rules of the programming languages shootout).
334
335 4.45 Thu Mar 13 11:55:36 CET 2008
336 - fix a file leaking bug in eg/mhyttpd that would allow
337 downloading of any file (reported by oesi).
338 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
339 (also add testcase).
340 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
341 apparently having loads of fun with that).
342 - implement $state->swap_defsv, ->swap_defav and document ->throw.
343
344 4.4 Wed Feb 13 16:44:29 CET 2008
345 - only assume .cfi pseudo isns exist on GNU/Linux.
346 - add get_timed method to Coro::Channels.
347 - fixed Coro::Debug manpage.
348 - perl 5.11 compatbility improvement based on patch by
349 Andreas Kรถnig.
350
351 4.37 Sun Jan 20 11:25:23 CET 2008
352 - soften the check for an active parser for perl 5.10 (avoiding
353 "Coro::State::transfer called while parsing" in many cases).
354
355 4.36 Sun Jan 13 10:53:56 CET 2008
356 - reset diehook when terminating from an async_pool as to not
357 trigger any __DIE__ handlers.
358
359 4.35 Sun Jan 13 04:14:13 CET 2008
360 - "bt" debug command now displays any exceptions
361 from longmess and also skips the innermost
362 stackframes, giving more useufl output.
363 - allow backtraces in coroutines blocked in Coro::EV,
364 at a <1% speed hit.
365 - handle localising of $SIG{__WARN__} and __DIE__
366 properly (with a proper amount of dirty hacking).
367
368 4.34 Sat Dec 22 17:49:53 CET 2007
369 - upgrade to EV version 2.0 API.
370
371 4.33 Mon Dec 17 08:36:12 CET 2007
372 - make Coro::AIO etc. loadable in the absence of EV.
373
374 4.32 Mon Dec 17 07:46:02 CET 2007
375 - majorly improved Coro::Handle's performance with Coro::EV.
376 - implemented the until now mythical Coro::BDB module.
377 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
378
379 4.31 Wed Dec 5 12:32:39 CET 2007
380 - remove warn statement form Coro::Util.
381
382 4.3 Tue Dec 4 20:33:14 CET 2007
383 - calls to the idle function could cause stack corruption
384 when the stack changed.
385 - do no longer rely on the presence of EV::DNS if EV is used
386 (because it is gone), but instead take avdantage of EV::ADNS
387 if available.
388 - add ($) prototypes to all functions in Coro::Storable.
389 - use a conventional (safer) idle callback in Coro::EV.
390 - do accept quickly in Coro::Debug to avoid endless loops.
391
392 4.22 Fri Nov 30 16:04:04 CET 2007
393 - really use optimised versions for Event and EV in Coro::Util
394 and Coro::Handle.
395
396 4.21 Sun Nov 25 10:48:59 CET 2007
397 - fix a spurious memory read.
398 - Coro::EV no longer keeps the eventloop "alive".
399
400 4.2 Fri Nov 9 20:47:05 CET 2007
401 - enable/disable tracing from a new coroutine, not a pooled one.
402 - fix a memleak in Coro::Event.
403 - removed killall call from fork_eval.
404 - made sure store_fd is already loaded so that fork_eval
405 does not have to parse autoload in each subprocess.
406 - only use assembly method if -O switch is in $Config{optimize}.
407 - add (optional) Coro::EV module, so far the best event loop module
408 directly supported by Coro.
409 - if the event model is EV, use EV::DNS to resolve
410 stuff in Coro::Util.
411 - don't get confused by multiple event notifications in Coro::Handle.
412 - initial support for EV (libevent interface).
413 - require Event and EV using configure_requires, to force their existance.
414
415 4.13 Wed Oct 24 07:26:45 CEST 2007
416 - add Coro::Storable::blocking_thaw.
417 - use a vastly more complicated technique to localise
418 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
419 - use a coroutine for the idle callback Coro::Event,
420 instead of running Event in the current coroutine context.
421 This also catches recursive invocations.
422 - actually report fork errors in gethostbyname and inet_aton.
423
424 4.11 Thu Oct 11 02:40:24 CEST 2007
425 - port to threaded perls.
426
427 4.1 Thu Oct 11 02:38:16 CEST 2007
428 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
429 be local to each coro (see Coro::State).
430 - incompatible change: for very deep reasons, cede and cede_notself
431 cannot return anything, so nothing will be returned.
432 - possibly bring back 5.10 compatibility (untested).
433 - work around stupid (and wrong) warning on 5.10 :(.
434 - overlay the saved state over the context stack. This saves
435 a few hundred bytes per coroutine on average and also
436 speeds up context switching a bit.
437 - further tune default stack sizes.
438 - (more) correctly calculate stack usage in coro_rss.
439 - Coro::Storable::blocking_* did not properly lock
440 resulting in races between coroutines.
441 - added Coro::Storable::guard.
442 - stopping to trace a coroutine could destroy the cctx of
443 an unrelated coroutine.
444 - explain the relationship between Perl and C coroutines in
445 more detail in Coro::State.
446 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
447 causing a fork per resolve. This also affected Coro::Socket.
448 - switch to a separate stack in $coro->call/eval to avoid
449 invalidating pointers.
450
451 4.03 Sat Oct 6 21:24:00 CEST 2007
452 - added Coro::throw method.
453 - minor code cleanups.
454
455 4.02 Sat Oct 6 02:36:47 CEST 2007
456 - fix a very minor per-coroutine memleak (a single codereference).
457 - fixed a bug where the currently in-use c context would be freed
458 prematurely (can happen only when programs change the stacksize
459 or use tracing).
460 - tracing can no longer keep a coro alive after it terminated.
461 - do static branch prediction in the common path for gcc. gives
462 about 2-5% speed improvement here.
463
464 4.01 Fri Oct 5 22:10:49 CEST 2007
465 - instead of recreating *a* standard output handle we simply
466 use STDOUT, which is faster and hopefully more robust.
467
468 4.0 Fri Oct 5 12:56:00 CEST 2007
469 - incompatibly changed Coro::Storable::freeze.
470 - major new feature: added Coro::Debug, for interactive coroutine
471 debugging, tracing and much more.
472 - major bug fix: unbelievable, but true: $_, $/ and many other
473 "saved" variables actually weren't being saved. This has been fixed,
474 of course, while increasing performance while losing all the save
475 flags.
476 - save flags are gone, and all the api functions dealing with them.
477 - added Coro::Semaphore::adjust.
478 - added Coro::Util::fork_eval.
479 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
480 - added Coro::killall.
481 - reduce initial stack sizes to allow for "micro-coroutines".
482 - better async_pool resource management, moved parts of async_pool
483 handling to XS (major speed improvement).
484 - actually croak before modifying important data structures.
485 - refuse to transfer while compiling.
486 - possibly support eval EXPR better now.
487 - enable assembly per default on linux+bsd x86+amd64.
488 - all internal members were renamed _something for easier subclassing.
489 - many minor tweaks.
490
491 3.63 Wed May 16 14:10:06 CEST 2007
492 - implement handcoded assembly for x86/amd64 SVR ABI.
493
494 3.62 Fri Apr 27 21:36:06 CEST 2007
495 - upgrade libcoro (which might set unwind info correctly).
496 - change default on linux to setjmp/longjmp.
497
498 3.61 Thu Apr 19 12:36:18 CEST 2007
499 - Coro::Storable caused an endless loop when thawing invalid
500 pst's sometimes.
501 - use a Semaphore in Coro::Storable, as Storable doesn't
502 seem to be reentrant (although it is documented to
503 be threadsafe...).
504 - fix Coro::Signal to bring back the original unreliable
505 but stateful semantics.
506 - fixed a lot of typos in Coro.pm (patch submitted by David
507 Steinbrunner, which applied flawlessly).
508
509 3.6 Sat Apr 14 17:13:31 CEST 2007
510 - added some bugfixes to get eg/myhttpd working again.
511 - added Coro::Storable for often-cede'ing freeze/thaw.
512 - try to do a clean exit when a coroutine calls exit
513 (EXPERIMENTAL).
514 - got rid of indirect call through _coro_init.
515 - updated the partly antique examples in eg/ to
516 work again and be a bit less magic, too.
517 - fixed Coro::Signal semantics to work as documented again.
518
519 3.55 Sun Mar 25 01:20:47 CET 2007
520 - add SAVE_DEFFH to save the default filehandle and enable
521 it by default.
522 - finally move socket-operations from Coro::Socket to Coro::Handle
523 to be able to unblock foreign sockets.
524 - add Coro::State::save_also and guarded_save.
525 - add count accessor to Coro::Semaphore.
526 - add Coro::State::cctx_stacksize.
527 - just for the fun of it, do not rely on implicit context,
528 which can dramatically improve performance, but people
529 using windows-process-emulation perls do not care much about
530 performance.
531
532 3.51 Sun Mar 4 14:18:04 CET 2007
533 - fixed a problem when you weakened references to Coro::State's
534 (patch by Yuval Kogman).
535
536 3.501 Wed Feb 28 12:44:07 CET 2007
537 - rename some global symbols as macosx from hell redefines
538 them without asking.
539
540 3.5 Tue Feb 13 20:22:53 CET 2007
541 - do AnyEvent model detection earlier, avoiding problems
542 caused by first using AnyEvent and later Coro::Event.
543 - implement and document Coro::Event event objects.
544 - fix a potential problem in Coro::Event causing crashes.
545 - initialise PL_comppad when creating a new coroutine,
546 avoids crashes on early coro destruction.
547
548 3.41 Mon Jan 22 19:19:49 CET 2007
549 - readline on Coro::Handle did not support undefined $/,
550 nor did it deliver partial lines on EOF or error.
551 - implement malloc fallback for stack allocation because
552 stupid broken idiotic OSX has a stupid broken
553 idiotic fits-the-whole-os mmap "implementation" and
554 my dick feels longer if Coro is portable even to
555 obsolete platforms.
556
557 3.4 Fri Jan 19 21:52:54 CET 2007
558 - remove t/09_timer.t, as it isn't really testing much
559 but was rather flaky in practise.
560 - async_pool coro would keep arguments and callback alive until
561 it was reused.
562 - cancellation of a coroutine could cause spurious idle calls
563 in cede_notself.
564
565 3.3 Sat Jan 6 03:45:00 CET 2007
566 - implement $coro->on_destroy.
567 - Coro::Event blocking semantics have been changed,
568 documented and - hopefully - improved.
569 - fix nice adding, not subtracting, from priority.
570 - fix ->prio and api_is_ready (patch by Mark Hinds).
571 - fixed an assert ("... == prev__cctx->idle_te")
572 that could errronously trigger.
573 - fix various large and small memleaks.
574 - use a (hopefully) more stable cancel implementation
575 that immediately frees the coroutine data.
576 - cede/cede_notself return a status now.
577 - added Coro::guard function.
578 - added a global coroutine pool for jobs (on my machine,
579 I can create and execute 48k simple coros/s with async,
580 and 128k coros with async_pool).
581 - Coro::AIO now uses the coroutine priority as io priority.
582
583 3.2 Fri Dec 22 05:07:09 CET 2006
584 - improve portability to slightly older perls.
585 - use cleaner coroutine destruction.
586 - simplify configuration for users.
587 - optionally (unrecommended) prefer perl functions over
588 their coro replacements.
589
590 3.11 Tue Dec 5 13:11:24 CET 2006
591 - fixed some bogus assert's, but as perl.h disables assert even
592 without NDEBUG (thank you very much), not too many people should
593 notice (that did include myself). Andreas Kรถnig noticed, though :)
594 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
595 it. Also noticed by Andreas Kรถnig :)
596 - save/restore tainted status.
597 - verified to pass the testsuite on my 5.9.5.
598
599 3.1 Mon Dec 4 23:03:40 CET 2006
600 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
601 - incompatible change: transfer flags are now per-state.
602 - give a better error message on deadlock.
603 - document Coro::nready.
604 - enhanced testsuite.
605
606 3.01 Sun Dec 3 23:47:42 CET 2006
607 - forgot to include Coro::Timer.
608
609 3.0 Sun Dec 3 22:57:25 CET 2006
610 - the "FINALLY COMPLETELY STABLE" release (coming soon:
611 the "FAMOUS LAST WORDS" release).
612 - implement a new stack sharing algorithm, which uses a stack
613 pool (size currently hardcoded).
614 - make stack sharing mandatory (it no longer uses a heuristic).
615 - eval/die no longer cause weird problems under heavy use.
616 - Coro::Event could cause livelocks if it was used but
617 no Coro::Event watchers were used.
618 - Coro::Event now uses asynccheck as prepare does not
619 check for changed watchers.
620 - Coro::Event allows multiple waiting coros and will wake up one
621 per event.
622 - Coro::Event should be cleaner and more efficient now.
623 - new utility function Coro::unblock_sub.
624 - document the sad fact that Event is no longer reentrant.
625 - putting a coroutine into the ready queue twice could under
626 some circumstances lead to stack corruption.
627 - minor incompatibility: subclassing Coro::State is supported
628 directly now without going through a _coro_state member.
629 - state/coro switching is much faster now.
630 - very minor optimisations and code/documentation cleanup.
631 - avoid problems due to compiler inlining.
632 - removed timers from Coro::Timer -> use AnyEvent instead.
633 - replaced Coro::idle coroutine by (cleaner) idle handler.
634 - updated to newest libcoro.
635 - implement enhanced support for valgrind.
636 - implement is_ready and return value for ready.
637 - removed Coro::Cont, it was a misnomer (it's generators),
638 and it was somewhat annoying to get it right. Will come back
639 if somebody asks for it :)
640 - many ->wait methods and Coro::Event could return
641 spuriously without the event having happened.
642
643 2.5 Tue Nov 7 12:22:33 CET 2006
644 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
645 use AnyEvent, moved them to Coro/.
646 - added Coro::LWP which contains all the uglyness required to
647 make LWP non-blocking.
648 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
649 - fixed another bug in Coro::Select when the timeout was undef.
650 - reuse PL_start_env for all coros, saves some memory per coroutine.
651 - manage PL_top_env differently, hopefully to avoid panic: top_env.
652 - timeout argument was not properly used in Coro::Socket.
653 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
654 - emulate undocumented(!) functionality of IO::Socket required
655 by LWP(!!).
656 - updated eg/lwp to work with newer lwp's.
657 - remove "FATAL: uncaught exception" prefix. Coroutines that die
658 kill the whole process, just as exceptions in the main "coroutine"
659 did already.
660
661 2.1 Wed Nov 1 23:01:13 CET 2006
662 - fix a long-standing bug in Coro::Select where select with
663 zero timeout would instead change the current default filehandle.
664 - use a simpler and hopefully more robust way to clone padlists
665 (uses less memory and a perl function instead of our own).
666 - coro can now create a stack guard area on many architectures.
667 - Coro::AIO properly reexports additional functions from IO::AIO.
668 - updated libcoro with a workaround for OS X,
669 pach and testing by Michael Schwern.
670
671 2.0 Tue Oct 24 05:47:17 CEST 2006
672 - support additional aio requests in Coro::AIO.
673
674 1.9 (never properly released due to a glitch)
675
676 1.8 Thu Feb 2 00:59:06 CET 2006
677 - applied suggested patch by SAMV to avoid problems during stupid
678 mark & sweep gc run.
679 - applied patch by Scott Walters for 5.9.3 compatibility.
680
681 1.7 Tue Dec 27 01:41:58 CET 2005
682 - added Coro::AIO, a thin wrapper around IO::AIO.
683 - improved Makefile.PL explanations.
684
685 1.6 Mon Dec 12 21:30:05 CET 2005
686 - additionally save PM_curpm might fix as-of-yet
687 unseen problems with regex matches being attributed to the
688 wrong package.
689 - add t/10_bugs.t, which currently checks against imho broken
690 perls that use 0.26MB of stack space per Perl_magic_get
691 invocation (newer linuxes) as opposed to the 0.0002MB perl
692 normally uses.
693 - make stacksize configurable for the ultimate debian experience.
694
695 1.51 Mon Dec 12 18:48:36 CET 2005
696 - remove debugging warn() accidentally left in Coro::Select.
697
698 1.5 Tue Nov 29 13:32:44 CET 2005
699 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
700 - fix Coro::Select 'select'.
701 - strict'ify some modules.
702 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
703 - use XSLoader in selected modules.
704 - remove some 5.6 compatibility cruft.
705
706 1.4 Tue Sep 6 00:11:05 CEST 2005
707 - libcoro did not take into account the trampoline on amd64,
708 when 'l'inux method was used.
709
710 1.31 Tue Aug 30 23:31:33 CEST 2005
711 - some portability fixes/workarounds.
712
713 1.3 Sat Aug 20 03:08:56 CEST 2005
714 - no code changes. module seems to work fine.
715 - improve Coro::State docs, remove reference to nonexisting
716 Coro::State::flush method.
717 - no longer autodetect windows, present it as an option instead.
718
719 1.2 Mon May 16 02:00:55 CEST 2005
720 - included libcoro.c earlier in State.xs, might improve portability.
721 - use a faster and possibly more stable Coro::Cont implementation.
722 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
723 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
724 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
725
726 1.11 Thu Mar 3 18:00:52 CET 2005
727 - change of contact address.
728
729 1.1 Tue Feb 22 20:51:16 CET 2005
730 - support [l]inux method on amd64.
731 - allow some context switches while compiling/eval'ing.
732
733 1.0 Tue Aug 10 03:47:27 CEST 2004
734 - Event 0.89 fixes the bug again.
735 - fix the remaining(??) memleak. I didn't free the cache-AV for
736 padlists, so one AV per created anon-sub leaked.
737
738 0.97 Fri May 14 15:23:32 CEST 2004
739 - remove stupid gcc extension that was an experiment only anyway.
740
741 0.96 Thu May 13 18:09:29 CEST 2004
742 - tell people if they have yet another broken Event variant.
743 - made Coro to work _better_ with threads (it seems that, as long
744 as you use Coros in one thread only everything will work fine.
745 Looking at the perl source the problem might be missing locking
746 between Coro and some perl internal routines. That's not easily
747 fixable).
748 - (hopefully) got rid of the static xnext inside transfer.
749
750 0.95 Sun Feb 15 03:20:28 CET 2004
751 - removed Coro::State::flush, implemented a new and much
752 faster caching mechanism by attaching magic to the CV.
753 - a memleak with (real) closures remains.
754 - dramatically reduced size of stacks to 65536 (linux/x86), after
755 observing that even myhttpd never had more than a kilobyte
756 of stack in use (stack is still only physically allocated in
757 pagesize increments for systems suporting mmap).
758 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
759 - fix for threaded perl.
760
761 0.9 Sun Nov 23 04:32:39 CET 2003
762 - fix a few very hard to track down but (of course) stupid
763 memory leaks.
764 - stringify version number to avoid locale problems :(.
765 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
766 hopefully ignoring it "just works".
767
768 0.8 Wed Nov 5 19:38:40 CET 2003
769 - port to perl5.005_03, but only 5.8.x+ is supported!
770 - honor LocalAddr even without LocalPort in Coro::Socket.
771
772 0.7 Tue May 27 03:12:38 CEST 2003
773 - the version jump indicates some level of testing, not gobs
774 of new features.
775 - uh... I found the prompt function in ExtUtils::MakeMaker.
776 Highly correct stuff, that is...
777 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
778 the fix is not well understood by the author, so beware :(.
779
780 0.652 Thu May 8 02:54:46 CEST 2003
781 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
782 cause version 4 doesn't have ucontext.
783 - Benjamin Reed reported that setjmp works fine on darwin,
784 so preselect it.
785
786 0.651 Sat Mar 29 15:00:23 CET 2003
787 - fix a bug in Coro::Handle where some bytes could get lost
788 on reading, reported by jason@nichego.net.
789
790 0.65 Sun Mar 23 00:08:26 CET 2003
791 - added a README.linux-glibc.
792 - new module Coro::Select.
793 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
794 (I never claimed that I know what I am doing ;).
795 - set default to setjmp/longjmp on non-x86-linux.
796
797 0.6 Thu Nov 21 11:09:06 CET 2002
798 - made lazy context switching the default.
799 - fixed the bug where SemaphoreSet::guard
800 was the same as timed_guard and timed_guard was missing.
801 - fix a memleak in Coro::Timer.
802 - reclassified context sharing as not experimental.
803
804 0.534 Sun Apr 14 03:05:12 CEST 2002
805 - fixed backspace => tab in header parsing (myhttpd).
806 - added eg/readline, for elmex.
807 - Coro::Event now calls ->start on first call instead of
808 ->again, which created wrong timeouts for the first ->next.
809 - fixed a bug where Coro::Socket returned a closed socket instead
810 of nothing on ECONNREFUSED and similar error conditions.
811
812 0.533 Mon Feb 18 18:49:40 CET 2002
813 - adapt to finally working Event-0.85.
814 - fixed buggy event initialization (trapped under windows, but how
815 could it possibly have worked anywhere?)
816
817 0.532 Wed Jan 16 02:45:32 CET 2002
818 - removed the reference to pp_entersub, might work on
819 (some) win32 perls now (testsuite works under cygwin,
820 without Event, which freezes).
821
822 0.531 Mon Dec 10 22:18:44 CET 2001
823 - Fixed a bug in SemaphoreSet::timed_down.
824
825 0.53 Tue Nov 27 21:11:13 CET 2001
826 - some tests for Event.
827 - slightly different internal architecture to get refcnt into the object
828 and not the reference. might fix a few bugs, certainly did introduce
829 new ones.
830 - Coro::Timer (independent of Event).
831 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
832
833 0.52 Tue Nov 6 21:36:18 CET 2001
834 - ported to cygwin (trivial).
835 patch by Gerrit P. Haase <gp@familiehaase.de>.
836 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
837
838 0.51 Thu Nov 1 20:39:01 CET 2001
839 - terminate/cancel now work properly (otherwise termination
840 could cause a "next coroutine is not and contains not..."
841 error.
842 - added Coro::Socket::shutdown.
843 - Coro::Event::loop no is the same as Event::loop.
844 - implemented terminate with args + join.
845
846 0.5 Fri Sep 28 16:15:35 CEST 2001
847 - fixed "print" on a Coro::Handle. The print method worked.
848 - small tweaks (seem to reduce memory consumption a lot)
849 in various modules.
850 - splendid use of "no warnings" scattered throughout.
851 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
852
853 0.49 Sun Sep 16 02:42:45 CEST 2001
854 - changed some method calls to function calls for speed
855 inside Coro::Handle.
856 - make Coro::Handle use an array instead of a hash for
857 speed reasons.
858 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
859 sjlj and ucontext should now work.
860 - IRIX-specific port for libcoro.
861 - swapped order of accept results to match IO::Socket.
862 - changed getsock/peername to sock/peername to match IO::Socket.
863 - fixed a bug that caused segfault when returning to main task
864 under some circumstances.
865 - other bugfixes.
866
867 0.45 Sun Sep 2 02:54:01 CEST 2001
868 - new method Coro::Handle::timeout.
869 - corrected speling of set...name to get...name in Coro::Socket.
870 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
871 - Coro::Handle now supports fileno and FILENO.
872 - added eg/myhttpd, a web server.
873 - Coro::Socket now accepts numeric ports as well as "port(number)"
874 syntax.
875 - moved some scheduling primitives into xs code. more to come.
876 - new simple priority system.
877 - implemented Coro::Event using XS, almost four times faster!
878 - small memory corruption problem fixed (boy that was difficult).
879
880 0.13 Wed Aug 8 16:53:07 CEST 2001
881 - new method Coro::Handle::readline.
882 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
883 (see eg/lwp on how to make LWP non-blocking).
884 - renamed Coro::Socket::new_inet to new.
885 - added Coro::Util, some utility functions.
886 - die/eval now works better (still get annoying (but true ;)
887 "Callback called exit" messages).
888 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
889 - much better Coro::Socket support.
890
891 0.12 Fri Jul 27 04:19:01 CEST 2001
892 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
893 - very experimental stack sharing algorithm. not 100% safe but
894 should work well in practise ;)
895 - added Coro::RWLock.
896 - Coro::Specific now works.
897
898 0.11 Tue Jul 24 22:49:21 CEST 2001
899 - added specialized hack for newer and older linux versions (fast).
900 - renamed Coro::Event::IO to Coro::Handle.
901 - new module Coro::Socket.
902
903 0.1 Tue Jul 24 01:47:53 CEST 2001
904 - release candidate 3. A rather new internal structure :(
905 - the great renaming: Damian Conway gave me a suitable replacement
906 for yield.
907 - added Coro/libcoro, a portable coroutine implementation for C
908 (not even perl-dependent!!!), which can be used by Coro::State.
909 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
910 perl-5.6.
911 - new function Coro::Event::idle.
912 - the idle process is now overriden
913 by default in Coro::Event.
914 - Coro::Channel now enforces the size.
915 - canceling events now works in all cases.
916 - Coro state now include $_ and $@.
917 - yet another bug workaround that I do not really understand :(
918 - new module Coro::Event::IO (very undocumented) to do
919 non-blocking i/o.
920 - performance tuning.
921
922 0.08 Thu Jul 19 06:13:25 CEST 2001
923 - release candidate 2 (new functionality).
924 - add Coro::State::flush function.
925 - transfer now supports a flags value (mostly for speed ATM).
926 - might compile and work in the presence of threads now.
927 - continuations have a saner syntax.
928 - no more memleaks.
929
930 0.07 Tue Jul 17 17:40:18 CEST 2001
931 - release candidate 1 ;)
932 - slightly nicer code.
933 - fixed a scheduling bug in Coro::Event.
934
935 0.06 Tue Jul 17 04:23:24 CEST 2001
936 - ok, I found the showstopper - the same sub must not be
937 re-entered in two different coroutines, otherwise => crash. I
938 see no easy solution to this problem, except by walking the call
939 chaing and saving/restoring the cv's, which is what I do now.
940 - memory leaks still latent, especially at thread termination.
941 - Coro::Event now works (haha).
942
943 0.05 Sun Jul 15 17:32:20 CEST 2001
944 - fixed a few issues in Coro::Event.
945 - I forgot to include Coro::Event in 0.04 :(:(:(
946
947 0.04 Sun Jul 15 05:24:59 CEST 2001
948 - @_ is now properly localized.
949 - Coro::State is now easier subclassable.
950 - Coro::Cont now coroutine-aware.
951 - Coro::Specific is a low-overhead module to create
952 coroutine-specific vars.
953 - Coro::Event provides a simple interface to Event.
954
955 0.03 Fri Jul 13 14:51:52 CEST 2001
956 - transfer() now implemented in XS (beware).
957 - new module Coro::Cont for really faked continuations.
958 - big internal architecture changes: Coro::State is now
959 really low-level and can thus be used to implement other
960 interesting things, While "Coro::" implements a process-like
961 model. Still crude and subject to change.
962 - $_ and $@ are no longer being localized.
963
964 0.02 Tue Jul 10 01:38:17 CEST 2001
965 - implemented "async" attribute.
966 - $_ and $@ are now localized.
967 - added Coro::Channel.
968 - more testcases, still no docs.
969
970 0.01 Tue Jul 3 02:18:41 CEST 2001
971 - original version; copied from Convert::Scalar.
972