ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.421
Committed: Sun Nov 16 10:12:38 2008 UTC (15 years, 6 months ago) by root
Branch: MAIN
Changes since 1.420: +1 -0 lines
Log Message:
*** empty log message ***

File Contents

# User Rev Content
1 root 1.1 Revision history for Perl extension Coro.
2 root 1.113
3 root 1.382 TODO: should explore PerlIO::coroaio (perl leaks like hell).
4 root 1.328 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 root 1.247 but very lacking in others (higher level ipc)).
7 root 1.173
8 root 1.419 TODO: testcancel
9     TODO: aio
10     TODO: Coro::Semaphore
11 root 1.421 TODO: trow disturbs cleanup path inside pp_slf?
12 root 1.408 5.0
13 root 1.409 - NEW ARCHITECTURE: use 4.912 for the stable version.
14 root 1.411 - bump API version to 7 - all dependents must be recompiled.
15 root 1.415 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
16 root 1.416 - async_pool did free a scalar value twice
17     ("Attempt to unreference...").
18 root 1.420 - fix a longstanding bug where calling terminate on a coro that
19     was waiting for a semaphore that was just becoming available
20     would cause a deadlock (semaphore would get into a state where
21     it was available but waiters were waiting(.
22 root 1.412 - perl's process emulation is now not even theoretically supported
23     anymore.
24     - directly patch the entersub opcode calling SLF functions (cede,
25     transfer and so on). this does speed up context switching, but
26     more importanly, it frees us from the hardcoded behaviour of
27     entersub, so we might actually be able to return something from
28     those functions and atcually create new ones.
29     - take advantage of __builtin_frame_address on gcc.
30     - expose THX in coroapi (not sure whether this was a wise decision,
31     as "threaded" perls are running at half speed anyways).
32     - implement execute_slf (schedule-like-function) interface that makes
33     it possible to implement schedule-like-functions in XS.
34 root 1.416 - use new SLF interface to massively speed up Coro::EV by roughly a factor
35     of two.
36 root 1.415 - removed timed_down and timed_guard functions - they were not being used
37 root 1.416 anyways and should be replaced by a more generic mechanism - and were
38     annoying to support anyways :
39     - used new SLF interface to massively speed up Coro::Semaphore by a factor
40     of three.
41 root 1.417 - forgot to include Coro/libcoro/README in the dist for all these years.
42 root 1.418 - work around a freebsd pthreads bug (manual testcancel is required as
43     pthread_cond_wait isn't a cancellation point on freebsd).
44 root 1.416
45     4.913 Sat Nov 15 07:58:28 CET 2008
46     - async_pool did free a scalar value twice
47     ("Attempt to unreference...").
48 root 1.408
49 root 1.407 4.912 Thu Nov 13 18:31:23 CET 2008
50 root 1.404 - minor cleanups.
51 root 1.405 - use much larger stacks on linux and perl < 5.8.8.
52 root 1.406 - Coro::Debug::new_unix_server did not unlink the socket
53     when destroyed.
54 root 1.404
55 root 1.403 4.911 Tue Nov 11 04:26:17 CET 2008
56 root 1.401 - "port" to threaded perls.
57    
58 root 1.400 4.91 Mon Nov 10 05:36:38 CET 2008
59     - the ->throw exception object no longer leaks.
60     - creating a new cctx leaked a scopestack entry (memleak).
61     - new coroutines didn't get created with a zero flags field
62     (unknown impact).
63 root 1.399 - calling ->throw on a not-yet-started coroutine should now work
64     instead of being ignored.
65 root 1.400 - ->throw is now supported on Coro::State objects.
66 root 1.398 - clean up cctx creation code a bit.
67 root 1.400 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
68 root 1.398
69 root 1.397 4.9 Sat Nov 8 17:45:27 CET 2008
70 root 1.392 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
71     - (libcoro) no longer leak threads in the experimental pthread backend,
72     also speed it up considerably.
73 root 1.394 - (libcoro) do not rely on makecontext passing void *'s unscathed.
74 root 1.392 - fix compiletime dependencies on libcoro in the Makefile.
75     - cctx_count wasn't always updated properly.
76 root 1.393 - Coro::State::cctx_stacksize wasn't applied correctly.
77     - new function Coro::State::cctx_max_idle.
78     - the default max number of idle C contexts is now 4.
79 root 1.394 - (libcoro) try harder to get _setjmp/_longjmp.
80     - (libcoro) cleanup and extend the libcoro API to officially
81     allow the creation of empty source contexts.
82 root 1.395 - very experimental workaround for the totally broken netbsd platform.
83 root 1.396 - tried to hack around openbsd bugs.
84 root 1.391
85 root 1.390 4.804 Wed Nov 5 16:36:00 CET 2008
86 root 1.389 - Coro::Debug::new_unix_server would not create a non-blocking listening
87 root 1.390 socket, sometimes causing freezes.
88 root 1.392 - (libcoro) fix misaligned stack points for setjmp and assembly
89     methods, which can cause crashes on x86/x86_64 with a sufficiently
90     aggressive compiler.
91 root 1.389 - new function: Coro::Debug::new_tcp_server.
92 root 1.388 - move ->throw into the Coro class because it only works on coro objects.
93    
94 root 1.387 4.803 Mon Nov 3 17:16:12 CET 2008
95 root 1.392 - (libcoro) use a global asm statement to become independent of gcc
96     otpimisations for CORO_ASM (thanks to pippijn for the idea).
97 root 1.387 - try to workaround yet another broken bsd, this time dragonfly.
98 root 1.386
99 root 1.385 4.802 Thu Oct 30 10:56:12 CET 2008
100     - support -fno-omit-frame-pointer on x86 with the assembly method.
101 root 1.383 - tune 01_unblock.t tests a bit.
102    
103 root 1.382 4.801 Wed Oct 22 18:33:37 CEST 2008
104 root 1.381 - improve readline speed for very long "lines".
105 root 1.382 - backport to 5.8.8.
106 root 1.381
107 root 1.380 4.8 Thu Oct 2 13:34:40 CEST 2008
108 root 1.378 - new function Coro::AIO::aio_wait.
109     - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
110     - greatly speed up and reduce memory usage of Coro::AIO requests.
111 root 1.379 - implement some other ยต-optimisations.
112 root 1.378
113 root 1.377 4.749 Mon Sep 29 14:40:12 CEST 2008
114     - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
115    
116 root 1.376 4.748 Sat Sep 27 14:03:19 CEST 2008
117 root 1.374 - implement, but do not document, PerlIO::cede(granularity).
118 root 1.372 - Coro::Storable forgot to wrap Storable::pstore.
119 root 1.374 - work around the multitude of leaks and memory corruption
120 root 1.375 bugs in PerlIO::via by using our own C-level perliol. As a side
121     effect, Coro::Storable is now much, much, much faster.
122 root 1.372
123 root 1.371 4.747 Tue Sep 23 01:59:41 CEST 2008
124 root 1.369 - fix a per-cv memleak (one empty array was leaked per
125     code reference).
126 root 1.370 - avoid a crash in coro->call|rss when the coroutine was already
127     destroyed (most noticably when using Coro::Debug::ps :)
128     - also protect *Storable::FILE.
129 root 1.371 - push up default storable granularity to 20ms.
130 root 1.369
131 root 1.368 4.746 Sun Sep 21 03:22:20 CEST 2008
132     - be more insistent on locking Storable against reentrancy
133     in Coro::Storable.
134 root 1.366 - move swap_def?v and throw to Coro::State, as documented.
135    
136 root 1.365 4.745 Thu Jul 24 00:14:38 CEST 2008
137 root 1.364 - remove debugging code related to MgPV_nolen_const, also try to
138     make it compile with perl 5.8.6 (yes, apple apparently loves
139     outdated software). Reported by John S.
140 root 1.363
141 root 1.362 4.744 Tue Jul 8 22:06:35 CEST 2008
142 root 1.361 - correctly provide default DIE/WARN handlers as documented.
143     - also overwrite PL_vtbl_sigelem.svt_clear, even though current
144     implementations inside perl work fine for us.
145    
146 root 1.360 4.743 Mon Jun 16 00:21:57 CEST 2008
147     - when using Coro::EV without running EV::loop it could
148     result in busy-waiting for events, this has been fixed.
149     - reduce codesize and improve performance by using EV_DEFAULT_UC.
150    
151 root 1.359 4.742 Sat May 31 14:10:21 CEST 2008
152     - implement a workaround for (some) perl <5.8.8 versions.
153     - require EV 3.3+.
154    
155 root 1.358 4.741 Fri May 30 23:33:09 CEST 2008
156     - tell netbsd how utterly broken their imitation of an OS is
157     and refuse to build by default if pthreads are in use.
158     - switch to "s" method on all bsd's by default, as their ucontext
159     stuff seems just too broken.
160     - fix a bug in Coro::Select.
161    
162 root 1.357 4.74 Thu May 29 20:05:31 CEST 2008
163     - do not test Coro::LWP for lack of dependencies.
164     - do not test Coro::Select for lack of working perls.
165    
166     4.73 Thu May 29 2008
167 root 1.356 - fix a bug in Coro::EV which would cause it to block despite
168     there being runnable coroutines.
169 root 1.354 - sprinkle "no warnings" freely over everything, also suppress
170     warnings for some other modules.
171 root 1.355 - fix typo in WSAEWOULDBLOCK.
172 root 1.354
173 root 1.353 4.72 Sun May 25 05:14:36 CEST 2008
174 root 1.351 - tweak META.yaml a bit, unfortunately, there is no documented way
175     to have optional dependencies with CPAN. doh :(
176 root 1.350 - avoid running some tests on windows because they would fail due to
177 root 1.351 perl bug (broken fork, broken pipes...).
178     - work around perl on windows bugs where perl returns undocumented
179     error codes for sysread, syswrite etc. by taking advantage
180     of AnyEvent's workaround for that problem.
181 root 1.350 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
182     a common perl implementation bug on windows.
183     - use unix domain sockets in testsuite to work around
184 root 1.351 common perl implementation bugs on widows (they are emulated by
185     tcp sockets on windows. ugh.)
186 root 1.350
187 root 1.349 4.71 Sat May 24 20:01:27 CEST 2008
188 root 1.348 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
189 root 1.347 - take advantage of async name resolution of AnyEvent::Util.
190     - work around brutal inet_aton override in Coro::LWP.
191 root 1.348 - take advantage of the readyhook in Coro::EV, for smoother
192     scheduling.
193 root 1.347
194 root 1.346 4.7 Sun May 11 00:32:19 CEST 2008
195 root 1.342 - completely reworked the Coro manpage.
196     - added Coro::AnyEvent, generic event loop integration.
197 root 1.341 - implement cancel, ready and kill commands in Coro::Debug.
198     - document find_coro in Coro::Debug.
199 root 1.343 - incompatible change: rename has_stack to has_cctx.
200 root 1.346 - Coro::AIO and Coro::BDB no longer force event model detection,
201     use AnyEvent::AIO and AnyEvent::BDB.
202 root 1.341
203 root 1.340 4.6 Sat Apr 26 10:05:14 CEST 2008
204 root 1.336 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
205 root 1.339 by the Coro module. It was a mistake to have it in the first place.
206     - (experimental) support for activestate perl 5.10 (method "w").
207     - (experimental) support for strawberry perl (method "a").
208 root 1.337 - coro_sigelem_set did not return a value although it had to,
209     actual impact unknown.
210 root 1.336
211 root 1.335 4.51 Mon Apr 14 13:28:27 CEST 2008
212     - make it compile again on 5.8.
213    
214 root 1.334 4.50 Thu Apr 10 09:43:17 CEST 2008
215     - I did it twice! (see 4.49).
216    
217 root 1.333 4.49 Sun Apr 6 21:23:31 CEST 2008
218     - grr, instead of compiling the recent changes on 5.10 only they
219     were compiled on 5.8 only.
220    
221 root 1.332 4.48 Sun Apr 6 20:36:46 CEST 2008
222 root 1.331 - allow coroutine switches during eval's under 5.10.x, as it seems
223     the parser is a per-interpreter option now, so this is safe
224     (this might fix the odd crashing bug).
225 root 1.332 - drop support for 5.9.x versions: they are dead, jim.
226 root 1.331
227 root 1.330 4.47 Sun Apr 6 00:37:52 CEST 2008
228 root 1.329 - force cctx allocation on API calls: we know we need to force one
229     and gcc actually manages to confuse our heuristic nowadays,
230     leading to crashes and worse.
231 root 1.330 - document force_cctx.
232 root 1.329
233 root 1.328 4.46 Fri Apr 4 22:05:43 CEST 2008
234     - upgrade libcoro, resulting in pthread-backend (which was only created
235     to fulfill the rules of the programming languages shootout).
236    
237 root 1.327 4.45 Thu Mar 13 11:55:36 CET 2008
238 root 1.322 - fix a file leaking bug in eg/mhyttpd that would allow
239     downloading of any file (reported by oesi).
240 root 1.325 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
241     (also add testcase).
242 root 1.323 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
243     apparently having loads of fun with that).
244 root 1.324 - implement $state->swap_defsv, ->swap_defav and document ->throw.
245 root 1.322
246 root 1.321 4.4 Wed Feb 13 16:44:29 CET 2008
247 root 1.319 - only assume .cfi pseudo isns exist on GNU/Linux.
248 root 1.320 - add get_timed method to Coro::Channels.
249 root 1.321 - fixed Coro::Debug manpage.
250     - perl 5.11 compatbility improvement based on patch by
251     Andreas Kรถnig.
252 root 1.319
253 root 1.318 4.37 Sun Jan 20 11:25:23 CET 2008
254 root 1.317 - soften the check for an active parser for perl 5.10 (avoiding
255     "Coro::State::transfer called while parsing" in many cases).
256 root 1.316
257 root 1.315 4.36 Sun Jan 13 10:53:56 CET 2008
258 root 1.314 - reset diehook when terminating from an async_pool as to not
259     trigger any __DIE__ handlers.
260    
261 root 1.313 4.35 Sun Jan 13 04:14:13 CET 2008
262 root 1.308 - "bt" debug command now displays any exceptions
263 root 1.311 from longmess and also skips the innermost
264     stackframes, giving more useufl output.
265 root 1.309 - allow backtraces in coroutines blocked in Coro::EV,
266 root 1.310 at a <1% speed hit.
267 root 1.312 - handle localising of $SIG{__WARN__} and __DIE__
268 root 1.314 properly (with a proper amount of dirty hacking).
269 root 1.308
270 root 1.307 4.34 Sat Dec 22 17:49:53 CET 2007
271 root 1.306 - upgrade to EV version 2.0 API.
272    
273 root 1.305 4.33 Mon Dec 17 08:36:12 CET 2007
274     - make Coro::AIO etc. loadable in the absence of EV.
275    
276 root 1.304 4.32 Mon Dec 17 07:46:02 CET 2007
277 root 1.301 - majorly improved Coro::Handle's performance with Coro::EV.
278 root 1.302 - implemented the until now mythical Coro::BDB module.
279 root 1.304 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
280 root 1.301
281 root 1.300 4.31 Wed Dec 5 12:32:39 CET 2007
282     - remove warn statement form Coro::Util.
283    
284 root 1.299 4.3 Tue Dec 4 20:33:14 CET 2007
285     - calls to the idle function could cause stack corruption
286     when the stack changed.
287 root 1.295 - do no longer rely on the presence of EV::DNS if EV is used
288     (because it is gone), but instead take avdantage of EV::ADNS
289     if available.
290 root 1.296 - add ($) prototypes to all functions in Coro::Storable.
291 root 1.297 - use a conventional (safer) idle callback in Coro::EV.
292 root 1.298 - do accept quickly in Coro::Debug to avoid endless loops.
293 root 1.295
294 root 1.294 4.22 Fri Nov 30 16:04:04 CET 2007
295 root 1.293 - really use optimised versions for Event and EV in Coro::Util
296     and Coro::Handle.
297    
298 root 1.292 4.21 Sun Nov 25 10:48:59 CET 2007
299 root 1.290 - fix a spurious memory read.
300 root 1.291 - Coro::EV no longer keeps the eventloop "alive".
301 root 1.290
302 root 1.289 4.2 Fri Nov 9 20:47:05 CET 2007
303 root 1.282 - enable/disable tracing from a new coroutine, not a pooled one.
304 root 1.283 - fix a memleak in Coro::Event.
305 root 1.284 - removed killall call from fork_eval.
306     - made sure store_fd is already loaded so that fork_eval
307     does not have to parse autoload in each subprocess.
308 root 1.288 - only use assembly method if -O switch is in $Config{optimize}.
309 root 1.289 - add (optional) Coro::EV module, so far the best event loop module
310     directly supported by Coro.
311     - if the event model is EV, use EV::DNS to resolve
312     stuff in Coro::Util.
313 root 1.286 - don't get confused by multiple event notifications in Coro::Handle.
314 root 1.287 - initial support for EV (libevent interface).
315 root 1.289 - require Event and EV using configure_requires, to force their existance.
316 root 1.282
317 root 1.281 4.13 Wed Oct 24 07:26:45 CEST 2007
318 root 1.277 - add Coro::Storable::blocking_thaw.
319 root 1.278 - use a vastly more complicated technique to localise
320     $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
321 root 1.279 - use a coroutine for the idle callback Coro::Event,
322     instead of running Event in the current coroutine context.
323     This also catches recursive invocations.
324 root 1.280 - actually report fork errors in gethostbyname and inet_aton.
325 root 1.277
326 root 1.276 4.11 Thu Oct 11 02:40:24 CEST 2007
327     - port to threaded perls.
328    
329 root 1.275 4.1 Thu Oct 11 02:38:16 CEST 2007
330 root 1.264 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
331     be local to each coro (see Coro::State).
332 root 1.274 - incompatible change: for very deep reasons, cede and cede_notself
333     cannot return anything, so nothing will be returned.
334 root 1.264 - possibly bring back 5.10 compatibility (untested).
335     - work around stupid (and wrong) warning on 5.10 :(.
336     - overlay the saved state over the context stack. This saves
337     a few hundred bytes per coroutine on average and also
338     speeds up context switching a bit.
339 root 1.265 - further tune default stack sizes.
340 root 1.264 - (more) correctly calculate stack usage in coro_rss.
341 root 1.267 - Coro::Storable::blocking_* did not properly lock
342     resulting in races between coroutines.
343     - added Coro::Storable::guard.
344 root 1.268 - stopping to trace a coroutine could destroy the cctx of
345     an unrelated coroutine.
346 root 1.269 - explain the relationship between Perl and C coroutines in
347     more detail in Coro::State.
348 root 1.270 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
349     causing a fork per resolve. This also affected Coro::Socket.
350 root 1.272 - switch to a separate stack in $coro->call/eval to avoid
351     invalidating pointers.
352 root 1.264
353 root 1.263 4.03 Sat Oct 6 21:24:00 CEST 2007
354 root 1.262 - added Coro::throw method.
355     - minor code cleanups.
356    
357 root 1.261 4.02 Sat Oct 6 02:36:47 CEST 2007
358 root 1.258 - fix a very minor per-coroutine memleak (a single codereference).
359 root 1.259 - fixed a bug where the currently in-use c context would be freed
360     prematurely (can happen only when programs change the stacksize
361     or use tracing).
362     - tracing can no longer keep a coro alive after it terminated.
363 root 1.260 - do static branch prediction in the common path for gcc. gives
364     about 2-5% speed improvement here.
365 root 1.258
366 root 1.256 4.01 Fri Oct 5 22:10:49 CEST 2007
367 root 1.257 - instead of recreating *a* standard output handle we simply
368     use STDOUT, which is faster and hopefully more robust.
369 root 1.255
370 root 1.254 4.0 Fri Oct 5 12:56:00 CEST 2007
371 root 1.238 - incompatibly changed Coro::Storable::freeze.
372 root 1.244 - major new feature: added Coro::Debug, for interactive coroutine
373     debugging, tracing and much more.
374 root 1.252 - major bug fix: unbelievable, but true: $_, $/ and many other
375     "saved" variables actually weren't being saved. This has been fixed,
376     of course, while increasing performance while losing all the save
377     flags.
378 root 1.251 - save flags are gone, and all the api functions dealing with them.
379 root 1.232 - added Coro::Semaphore::adjust.
380 root 1.233 - added Coro::Util::fork_eval.
381     - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
382 root 1.248 - added Coro::killall.
383 root 1.236 - reduce initial stack sizes to allow for "micro-coroutines".
384 root 1.239 - better async_pool resource management, moved parts of async_pool
385 root 1.240 handling to XS (major speed improvement).
386 root 1.245 - actually croak before modifying important data structures.
387     - refuse to transfer while compiling.
388     - possibly support eval EXPR better now.
389 root 1.249 - enable assembly per default on linux+bsd x86+amd64.
390 root 1.253 - all internal members were renamed _something for easier subclassing.
391 root 1.237 - many minor tweaks.
392 root 1.232
393 root 1.231 3.63 Wed May 16 14:10:06 CEST 2007
394 root 1.230 - implement handcoded assembly for x86/amd64 SVR ABI.
395    
396 root 1.229 3.62 Fri Apr 27 21:36:06 CEST 2007
397     - upgrade libcoro (which might set unwind info correctly).
398 root 1.228 - change default on linux to setjmp/longjmp.
399 root 1.227
400 root 1.226 3.61 Thu Apr 19 12:36:18 CEST 2007
401 root 1.225 - Coro::Storable caused an endless loop when thawing invalid
402     pst's sometimes.
403 root 1.223 - use a Semaphore in Coro::Storable, as Storable doesn't
404 root 1.224 seem to be reentrant (although it is documented to
405     be threadsafe...).
406     - fix Coro::Signal to bring back the original unreliable
407     but stateful semantics.
408 root 1.226 - fixed a lot of typos in Coro.pm (patch submitted by David
409     Steinbrunner, which applied flawlessly).
410 root 1.223
411     3.6 Sat Apr 14 17:13:31 CEST 2007
412 root 1.217 - added some bugfixes to get eg/myhttpd working again.
413 root 1.218 - added Coro::Storable for often-cede'ing freeze/thaw.
414 root 1.219 - try to do a clean exit when a coroutine calls exit
415     (EXPERIMENTAL).
416 root 1.220 - got rid of indirect call through _coro_init.
417     - updated the partly antique examples in eg/ to
418     work again and be a bit less magic, too.
419     - fixed Coro::Signal semantics to work as documented again.
420 root 1.217
421 root 1.216 3.55 Sun Mar 25 01:20:47 CET 2007
422 root 1.215 - add SAVE_DEFFH to save the default filehandle and enable
423     it by default.
424     - finally move socket-operations from Coro::Socket to Coro::Handle
425     to be able to unblock foreign sockets.
426     - add Coro::State::save_also and guarded_save.
427 root 1.211 - add count accessor to Coro::Semaphore.
428 root 1.212 - add Coro::State::cctx_stacksize.
429 root 1.213 - just for the fun of it, do not rely on implicit context,
430     which can dramatically improve performance, but people
431     using windows-process-emulation perls do not care much about
432     performance.
433 root 1.211
434 root 1.210 3.51 Sun Mar 4 14:18:04 CET 2007
435 root 1.209 - fixed a problem when you weakened references to Coro::State's
436     (patch by Yuval Kogman).
437    
438 root 1.208 3.501 Wed Feb 28 12:44:07 CET 2007
439     - rename some global symbols as macosx from hell redefines
440     them without asking.
441    
442 root 1.207 3.5 Tue Feb 13 20:22:53 CET 2007
443 root 1.204 - do AnyEvent model detection earlier, avoiding problems
444     caused by first using AnyEvent and later Coro::Event.
445 root 1.205 - implement and document Coro::Event event objects.
446     - fix a potential problem in Coro::Event causing crashes.
447 root 1.206 - initialise PL_comppad when creating a new coroutine,
448     avoids crashes on early coro destruction.
449 root 1.204
450 root 1.203 3.41 Mon Jan 22 19:19:49 CET 2007
451 root 1.202 - readline on Coro::Handle did not support undefined $/,
452     nor did it deliver partial lines on EOF or error.
453 root 1.203 - implement malloc fallback for stack allocation because
454     stupid broken idiotic OSX has a stupid broken
455     idiotic fits-the-whole-os mmap "implementation" and
456     my dick feels longer if Coro is portable even to
457     obsolete platforms.
458 root 1.202
459 root 1.201 3.4 Fri Jan 19 21:52:54 CET 2007
460 root 1.197 - remove t/09_timer.t, as it isn't really testing much
461     but was rather flaky in practise.
462 root 1.198 - async_pool coro would keep arguments and callback alive until
463     it was reused.
464 root 1.199 - cancellation of a coroutine could cause spurious idle calls
465     in cede_notself.
466 root 1.197
467 root 1.196 3.3 Sat Jan 6 03:45:00 CET 2007
468 root 1.187 - implement $coro->on_destroy.
469 root 1.188 - Coro::Event blocking semantics have been changed,
470     documented and - hopefully - improved.
471 root 1.184 - fix nice adding, not subtracting, from priority.
472 root 1.185 - fix ->prio and api_is_ready (patch by Mark Hinds).
473 root 1.190 - fixed an assert ("... == prev__cctx->idle_te")
474     that could errronously trigger.
475 root 1.192 - fix various large and small memleaks.
476 root 1.191 - use a (hopefully) more stable cancel implementation
477     that immediately frees the coroutine data.
478 root 1.194 - cede/cede_notself return a status now.
479 root 1.191 - added Coro::guard function.
480 root 1.193 - added a global coroutine pool for jobs (on my machine,
481     I can create and execute 48k simple coros/s with async,
482     and 128k coros with async_pool).
483 root 1.195 - Coro::AIO now uses the coroutine priority as io priority.
484 root 1.184
485 root 1.183 3.2 Fri Dec 22 05:07:09 CET 2006
486 root 1.179 - improve portability to slightly older perls.
487 root 1.180 - use cleaner coroutine destruction.
488 root 1.182 - simplify configuration for users.
489     - optionally (unrecommended) prefer perl functions over
490     their coro replacements.
491 root 1.179
492 root 1.178 3.11 Tue Dec 5 13:11:24 CET 2006
493     - fixed some bogus assert's, but as perl.h disables assert even
494     without NDEBUG (thank you very much), not too many people should
495     notice (that did include myself). Andreas Kรถnig noticed, though :)
496     - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
497     it. Also noticed by Andreas Kรถnig :)
498     - save/restore tainted status.
499     - verified to pass the testsuite on my 5.9.5.
500    
501 root 1.177 3.1 Mon Dec 4 23:03:40 CET 2006
502     - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
503 root 1.175 - incompatible change: transfer flags are now per-state.
504 root 1.173 - give a better error message on deadlock.
505 root 1.175 - document Coro::nready.
506 root 1.174 - enhanced testsuite.
507 root 1.169
508 root 1.172 3.01 Sun Dec 3 23:47:42 CET 2006
509     - forgot to include Coro::Timer.
510    
511 root 1.171 3.0 Sun Dec 3 22:57:25 CET 2006
512 root 1.163 - the "FINALLY COMPLETELY STABLE" release (coming soon:
513     the "FAMOUS LAST WORDS" release).
514 root 1.166 - implement a new stack sharing algorithm, which uses a stack
515     pool (size currently hardcoded).
516 root 1.155 - make stack sharing mandatory (it no longer uses a heuristic).
517 root 1.157 - eval/die no longer cause weird problems under heavy use.
518 root 1.152 - Coro::Event could cause livelocks if it was used but
519 root 1.161 no Coro::Event watchers were used.
520 root 1.167 - Coro::Event now uses asynccheck as prepare does not
521     check for changed watchers.
522 root 1.170 - Coro::Event allows multiple waiting coros and will wake up one
523     per event.
524 root 1.152 - Coro::Event should be cleaner and more efficient now.
525 root 1.165 - new utility function Coro::unblock_sub.
526 root 1.164 - document the sad fact that Event is no longer reentrant.
527 root 1.163 - putting a coroutine into the ready queue twice could under
528     some circumstances lead to stack corruption.
529 root 1.161 - minor incompatibility: subclassing Coro::State is supported
530     directly now without going through a _coro_state member.
531 root 1.156 - state/coro switching is much faster now.
532 root 1.154 - very minor optimisations and code/documentation cleanup.
533 root 1.153 - avoid problems due to compiler inlining.
534 root 1.154 - removed timers from Coro::Timer -> use AnyEvent instead.
535     - replaced Coro::idle coroutine by (cleaner) idle handler.
536 root 1.156 - updated to newest libcoro.
537 root 1.159 - implement enhanced support for valgrind.
538 root 1.162 - implement is_ready and return value for ready.
539 root 1.168 - removed Coro::Cont, it was a misnomer (it's generators),
540     and it was somewhat annoying to get it right. Will come back
541     if somebody asks for it :)
542 root 1.163 - many ->wait methods and Coro::Event could return
543     spuriously without the event having happened.
544 root 1.146
545 root 1.145 2.5 Tue Nov 7 12:22:33 CET 2006
546 root 1.143 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
547     use AnyEvent, moved them to Coro/.
548 root 1.140 - added Coro::LWP which contains all the uglyness required to
549     make LWP non-blocking.
550     - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
551 root 1.141 - fixed another bug in Coro::Select when the timeout was undef.
552 root 1.136 - reuse PL_start_env for all coros, saves some memory per coroutine.
553     - manage PL_top_env differently, hopefully to avoid panic: top_env.
554 root 1.137 - timeout argument was not properly used in Coro::Socket.
555     - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
556 root 1.138 - emulate undocumented(!) functionality of IO::Socket required
557     by LWP(!!).
558     - updated eg/lwp to work with newer lwp's.
559 root 1.140 - remove "FATAL: uncaught exception" prefix. Coroutines that die
560     kill the whole process, just as exceptions in the main "coroutine"
561     did already.
562 root 1.135
563 root 1.134 2.1 Wed Nov 1 23:01:13 CET 2006
564     - fix a long-standing bug in Coro::Select where select with
565     zero timeout would instead change the current default filehandle.
566 root 1.132 - use a simpler and hopefully more robust way to clone padlists
567     (uses less memory and a perl function instead of our own).
568     - coro can now create a stack guard area on many architectures.
569 root 1.134 - Coro::AIO properly reexports additional functions from IO::AIO.
570 root 1.132 - updated libcoro with a workaround for OS X,
571 root 1.134 pach and testing by Michael Schwern.
572 root 1.130
573 root 1.129 2.0 Tue Oct 24 05:47:17 CEST 2006
574 root 1.128 - support additional aio requests in Coro::AIO.
575    
576 root 1.129 1.9 (never properly released due to a glitch)
577    
578 root 1.127 1.8 Thu Feb 2 00:59:06 CET 2006
579 root 1.125 - applied suggested patch by SAMV to avoid problems during stupid
580     mark & sweep gc run.
581 root 1.126 - applied patch by Scott Walters for 5.9.3 compatibility.
582 root 1.125
583 root 1.124 1.7 Tue Dec 27 01:41:58 CET 2005
584 root 1.123 - added Coro::AIO, a thin wrapper around IO::AIO.
585 root 1.122 - improved Makefile.PL explanations.
586    
587 root 1.119 1.6 Mon Dec 12 21:30:05 CET 2005
588 root 1.117 - additionally save PM_curpm might fix as-of-yet
589 root 1.121 unseen problems with regex matches being attributed to the
590 root 1.117 wrong package.
591 root 1.118 - add t/10_bugs.t, which currently checks against imho broken
592     perls that use 0.26MB of stack space per Perl_magic_get
593 root 1.121 invocation (newer linuxes) as opposed to the 0.0002MB perl
594     normally uses.
595 root 1.120 - make stacksize configurable for the ultimate debian experience.
596 root 1.117
597 root 1.116 1.51 Mon Dec 12 18:48:36 CET 2005
598     - remove debugging warn() accidentally left in Coro::Select.
599    
600 root 1.115 1.5 Tue Nov 29 13:32:44 CET 2005
601 root 1.114 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
602 root 1.115 - fix Coro::Select 'select'.
603     - strict'ify some modules.
604 root 1.112 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
605 root 1.115 - use XSLoader in selected modules.
606     - remove some 5.6 compatibility cruft.
607 root 1.112
608 root 1.111 1.4 Tue Sep 6 00:11:05 CEST 2005
609     - libcoro did not take into account the trampoline on amd64,
610     when 'l'inux method was used.
611    
612 root 1.110 1.31 Tue Aug 30 23:31:33 CEST 2005
613     - some portability fixes/workarounds.
614    
615 root 1.107 1.3 Sat Aug 20 03:08:56 CEST 2005
616 root 1.108 - no code changes. module seems to work fine.
617 root 1.106 - improve Coro::State docs, remove reference to nonexisting
618     Coro::State::flush method.
619 root 1.107 - no longer autodetect windows, present it as an option instead.
620 root 1.106
621 root 1.105 1.2 Mon May 16 02:00:55 CEST 2005
622 root 1.100 - included libcoro.c earlier in State.xs, might improve portability.
623 root 1.101 - use a faster and possibly more stable Coro::Cont implementation.
624 root 1.102 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
625 root 1.104 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
626 root 1.105 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
627 root 1.100
628 root 1.99 1.11 Thu Mar 3 18:00:52 CET 2005
629     - change of contact address.
630    
631 root 1.98 1.1 Tue Feb 22 20:51:16 CET 2005
632 root 1.97 - support [l]inux method on amd64.
633 root 1.109 - allow some context switches while compiling/eval'ing.
634 root 1.97
635 root 1.96 1.0 Tue Aug 10 03:47:27 CEST 2004
636 root 1.95 - Event 0.89 fixes the bug again.
637 root 1.96 - fix the remaining(??) memleak. I didn't free the cache-AV for
638     padlists, so one AV per created anon-sub leaked.
639 root 1.95
640 pcg 1.94 0.97 Fri May 14 15:23:32 CEST 2004
641     - remove stupid gcc extension that was an experiment only anyway.
642    
643 pcg 1.93 0.96 Thu May 13 18:09:29 CEST 2004
644 root 1.103 - tell people if they have yet another broken Event variant.
645 pcg 1.91 - made Coro to work _better_ with threads (it seems that, as long
646     as you use Coros in one thread only everything will work fine.
647     Looking at the perl source the problem might be missing locking
648     between Coro and some perl internal routines. That's not easily
649     fixable).
650 pcg 1.92 - (hopefully) got rid of the static xnext inside transfer.
651 pcg 1.91
652 pcg 1.90 0.95 Sun Feb 15 03:20:28 CET 2004
653 pcg 1.86 - removed Coro::State::flush, implemented a new and much
654     faster caching mechanism by attaching magic to the CV.
655 pcg 1.87 - a memleak with (real) closures remains.
656 pcg 1.88 - dramatically reduced size of stacks to 65536 (linux/x86), after
657     observing that even myhttpd never had more than a kilobyte
658     of stack in use (stack is still only physically allocated in
659     pagesize increments for systems suporting mmap).
660 pcg 1.89 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
661 pcg 1.90 - fix for threaded perl.
662 pcg 1.86
663 pcg 1.85 0.9 Sun Nov 23 04:32:39 CET 2003
664 pcg 1.84 - fix a few very hard to track down but (of course) stupid
665     memory leaks.
666     - stringify version number to avoid locale problems :(.
667 pcg 1.85 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
668     hopefully ignoring it "just works".
669 pcg 1.84
670 pcg 1.83 0.8 Wed Nov 5 19:38:40 CET 2003
671     - port to perl5.005_03, but only 5.8.x+ is supported!
672 pcg 1.82 - honor LocalAddr even without LocalPort in Coro::Socket.
673    
674 root 1.81 0.7 Tue May 27 03:12:38 CEST 2003
675     - the version jump indicates some level of testing, not gobs
676     of new features.
677 root 1.80 - uh... I found the prompt function in ExtUtils::MakeMaker.
678     Highly correct stuff, that is...
679 root 1.81 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
680     the fix is not well understood by the author, so beware :(.
681 root 1.80
682 root 1.79 0.652 Thu May 8 02:54:46 CEST 2003
683 root 1.76 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
684 root 1.78 cause version 4 doesn't have ucontext.
685 root 1.77 - Benjamin Reed reported that setjmp works fine on darwin,
686     so preselect it.
687 root 1.76
688 root 1.75 0.651 Sat Mar 29 15:00:23 CET 2003
689     - fix a bug in Coro::Handle where some bytes could get lost
690     on reading, reported by jason@nichego.net.
691 root 1.74
692 root 1.73 0.65 Sun Mar 23 00:08:26 CET 2003
693 root 1.72 - added a README.linux-glibc.
694     - new module Coro::Select.
695     - also save/restore PL_comppad, fixes segfaults with 5.8.1.
696     (I never claimed that I know what I am doing ;).
697 root 1.73 - set default to setjmp/longjmp on non-x86-linux.
698 root 1.72
699 root 1.71 0.6 Thu Nov 21 11:09:06 CET 2002
700     - made lazy context switching the default.
701     - fixed the bug where SemaphoreSet::guard
702     was the same as timed_guard and timed_guard was missing.
703     - fix a memleak in Coro::Timer.
704     - reclassified context sharing as not experimental.
705    
706 root 1.70 0.534 Sun Apr 14 03:05:12 CEST 2002
707 root 1.69 - fixed backspace => tab in header parsing (myhttpd).
708 root 1.70 - added eg/readline, for elmex.
709     - Coro::Event now calls ->start on first call instead of
710     ->again, which created wrong timeouts for the first ->next.
711     - fixed a bug where Coro::Socket returned a closed socket instead
712     of nothing on ECONNREFUSED and similar error conditions.
713 root 1.69
714 root 1.68 0.533 Mon Feb 18 18:49:40 CET 2002
715     - adapt to finally working Event-0.85.
716     - fixed buggy event initialization (trapped under windows, but how
717     could it possibly have worked anywhere?)
718    
719 root 1.67 0.532 Wed Jan 16 02:45:32 CET 2002
720 root 1.66 - removed the reference to pp_entersub, might work on
721     (some) win32 perls now (testsuite works under cygwin,
722     without Event, which freezes).
723 root 1.65
724 root 1.64 0.531 Mon Dec 10 22:18:44 CET 2001
725 root 1.63 - Fixed a bug in SemaphoreSet::timed_down.
726    
727 root 1.64 0.53 Tue Nov 27 21:11:13 CET 2001
728 root 1.60 - some tests for Event.
729 root 1.61 - slightly different internal architecture to get refcnt into the object
730     and not the reference. might fix a few bugs, certainly did introduce
731     new ones.
732     - Coro::Timer (independent of Event).
733 root 1.62 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
734 root 1.60
735 root 1.64 0.52 Tue Nov 6 21:36:18 CET 2001
736 root 1.58 - ported to cygwin (trivial).
737     patch by Gerrit P. Haase <gp@familiehaase.de>.
738     - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
739    
740 root 1.64 0.51 Thu Nov 1 20:39:01 CET 2001
741 root 1.54 - terminate/cancel now work properly (otherwise termination
742     could cause a "next coroutine is not and contains not..."
743     error.
744 root 1.55 - added Coro::Socket::shutdown.
745 root 1.56 - Coro::Event::loop no is the same as Event::loop.
746     - implemented terminate with args + join.
747 root 1.54
748 root 1.64 0.5 Fri Sep 28 16:15:35 CEST 2001
749 root 1.53 - fixed "print" on a Coro::Handle. The print method worked.
750 root 1.50 - small tweaks (seem to reduce memory consumption a lot)
751     in various modules.
752     - splendid use of "no warnings" scattered throughout.
753 root 1.52 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
754 root 1.48
755 root 1.47 0.49 Sun Sep 16 02:42:45 CEST 2001
756 root 1.43 - changed some method calls to function calls for speed
757     inside Coro::Handle.
758 root 1.44 - make Coro::Handle use an array instead of a hash for
759     speed reasons.
760     - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
761     sjlj and ucontext should now work.
762     - IRIX-specific port for libcoro.
763 root 1.45 - swapped order of accept results to match IO::Socket.
764 root 1.46 - changed getsock/peername to sock/peername to match IO::Socket.
765     - fixed a bug that caused segfault when returning to main task
766     under some circumstances.
767     - other bugfixes.
768 root 1.43
769 root 1.42 0.45 Sun Sep 2 02:54:01 CEST 2001
770 root 1.36 - new method Coro::Handle::timeout.
771     - corrected speling of set...name to get...name in Coro::Socket.
772     - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
773     - Coro::Handle now supports fileno and FILENO.
774     - added eg/myhttpd, a web server.
775 root 1.37 - Coro::Socket now accepts numeric ports as well as "port(number)"
776     syntax.
777 root 1.38 - moved some scheduling primitives into xs code. more to come.
778 root 1.39 - new simple priority system.
779 root 1.40 - implemented Coro::Event using XS, almost four times faster!
780 root 1.41 - small memory corruption problem fixed (boy that was difficult).
781 root 1.36
782 root 1.35 0.13 Wed Aug 8 16:53:07 CEST 2001
783 root 1.31 - new method Coro::Handle::readline.
784 root 1.34 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
785     (see eg/lwp on how to make LWP non-blocking).
786 root 1.31 - renamed Coro::Socket::new_inet to new.
787 root 1.32 - added Coro::Util, some utility functions.
788 root 1.33 - die/eval now works better (still get annoying (but true ;)
789     "Callback called exit" messages).
790 root 1.34 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
791     - much better Coro::Socket support.
792 root 1.31
793     0.12 Fri Jul 27 04:19:01 CEST 2001
794 root 1.29 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
795 root 1.34 - very experimental stack sharing algorithm. not 100% safe but
796 root 1.28 should work well in practise ;)
797 root 1.30 - added Coro::RWLock.
798 root 1.31 - Coro::Specific now works.
799 root 1.28
800     0.11 Tue Jul 24 22:49:21 CEST 2001
801 root 1.27 - added specialized hack for newer and older linux versions (fast).
802     - renamed Coro::Event::IO to Coro::Handle.
803     - new module Coro::Socket.
804    
805 root 1.26 0.1 Tue Jul 24 01:47:53 CEST 2001
806 root 1.23 - release candidate 3. A rather new internal structure :(
807 root 1.24 - the great renaming: Damian Conway gave me a suitable replacement
808     for yield.
809 root 1.23 - added Coro/libcoro, a portable coroutine implementation for C
810 root 1.22 (not even perl-dependent!!!), which can be used by Coro::State.
811 root 1.20 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
812     perl-5.6.
813 root 1.16 - new function Coro::Event::idle.
814     - the idle process is now overriden
815     by default in Coro::Event.
816 root 1.17 - Coro::Channel now enforces the size.
817 root 1.18 - canceling events now works in all cases.
818 root 1.19 - Coro state now include $_ and $@.
819     - yet another bug workaround that I do not really understand :(
820 root 1.21 - new module Coro::Event::IO (very undocumented) to do
821     non-blocking i/o.
822 root 1.25 - performance tuning.
823 root 1.16
824 root 1.15 0.08 Thu Jul 19 06:13:25 CEST 2001
825 root 1.14 - release candidate 2 (new functionality).
826 root 1.13 - add Coro::State::flush function.
827     - transfer now supports a flags value (mostly for speed ATM).
828     - might compile and work in the presence of threads now.
829 root 1.14 - continuations have a saner syntax.
830 root 1.19 - no more memleaks.
831 root 1.13
832 root 1.12 0.07 Tue Jul 17 17:40:18 CEST 2001
833     - release candidate 1 ;)
834     - slightly nicer code.
835     - fixed a scheduling bug in Coro::Event.
836    
837 root 1.11 0.06 Tue Jul 17 04:23:24 CEST 2001
838     - ok, I found the showstopper - the same sub must not be
839 root 1.9 re-entered in two different coroutines, otherwise => crash. I
840     see no easy solution to this problem, except by walking the call
841     chaing and saving/restoring the cv's, which is what I do now.
842 root 1.11 - memory leaks still latent, especially at thread termination.
843     - Coro::Event now works (haha).
844 root 1.9
845 root 1.8 0.05 Sun Jul 15 17:32:20 CEST 2001
846     - fixed a few issues in Coro::Event.
847     - I forgot to include Coro::Event in 0.04 :(:(:(
848    
849 root 1.7 0.04 Sun Jul 15 05:24:59 CEST 2001
850 root 1.5 - @_ is now properly localized.
851 root 1.4 - Coro::State is now easier subclassable.
852     - Coro::Cont now coroutine-aware.
853     - Coro::Specific is a low-overhead module to create
854     coroutine-specific vars.
855 root 1.6 - Coro::Event provides a simple interface to Event.
856 root 1.4
857 root 1.3 0.03 Fri Jul 13 14:51:52 CEST 2001
858 root 1.4 - transfer() now implemented in XS (beware).
859     - new module Coro::Cont for really faked continuations.
860     - big internal architecture changes: Coro::State is now
861 root 1.3 really low-level and can thus be used to implement other
862 root 1.4 interesting things, While "Coro::" implements a process-like
863     model. Still crude and subject to change.
864 root 1.3 - $_ and $@ are no longer being localized.
865    
866 root 1.2 0.02 Tue Jul 10 01:38:17 CEST 2001
867     - implemented "async" attribute.
868     - $_ and $@ are now localized.
869     - added Coro::Channel.
870     - more testcases, still no docs.
871    
872 root 1.1 0.01 Tue Jul 3 02:18:41 CEST 2001
873     - original version; copied from Convert::Scalar.
874