ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
(Generate patch)

Comparing Coro/Changes (file contents):
Revision 1.211 by root, Wed Mar 7 13:11:10 2007 UTC vs.
Revision 1.404 by root, Wed Nov 12 04:49:06 2008 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: better (earlier) error messages on broken transfer's 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: destroy gets only called on cancel: fix(?) 4TODO: 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)).
5 7
63.52 8 - minor cleanups.
9
104.911 Tue Nov 11 04:26:17 CET 2008
11 - "port" to threaded perls.
12
134.91 Mon Nov 10 05:36:38 CET 2008
14 - the ->throw exception object no longer leaks.
15 - creating a new cctx leaked a scopestack entry (memleak).
16 - new coroutines didn't get created with a zero flags field
17 (unknown impact).
18 - calling ->throw on a not-yet-started coroutine should now work
19 instead of being ignored.
20 - ->throw is now supported on Coro::State objects.
21 - clean up cctx creation code a bit.
22 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
23
244.9 Sat Nov 8 17:45:27 CET 2008
25 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
26 - (libcoro) no longer leak threads in the experimental pthread backend,
27 also speed it up considerably.
28 - (libcoro) do not rely on makecontext passing void *'s unscathed.
29 - fix compiletime dependencies on libcoro in the Makefile.
30 - cctx_count wasn't always updated properly.
31 - Coro::State::cctx_stacksize wasn't applied correctly.
32 - new function Coro::State::cctx_max_idle.
33 - the default max number of idle C contexts is now 4.
34 - (libcoro) try harder to get _setjmp/_longjmp.
35 - (libcoro) cleanup and extend the libcoro API to officially
36 allow the creation of empty source contexts.
37 - very experimental workaround for the totally broken netbsd platform.
38 - tried to hack around openbsd bugs.
39
404.804 Wed Nov 5 16:36:00 CET 2008
41 - Coro::Debug::new_unix_server would not create a non-blocking listening
42 socket, sometimes causing freezes.
43 - (libcoro) fix misaligned stack points for setjmp and assembly
44 methods, which can cause crashes on x86/x86_64 with a sufficiently
45 aggressive compiler.
46 - new function: Coro::Debug::new_tcp_server.
47 - move ->throw into the Coro class because it only works on coro objects.
48
494.803 Mon Nov 3 17:16:12 CET 2008
50 - (libcoro) use a global asm statement to become independent of gcc
51 otpimisations for CORO_ASM (thanks to pippijn for the idea).
52 - try to workaround yet another broken bsd, this time dragonfly.
53
544.802 Thu Oct 30 10:56:12 CET 2008
55 - support -fno-omit-frame-pointer on x86 with the assembly method.
56 - tune 01_unblock.t tests a bit.
57
584.801 Wed Oct 22 18:33:37 CEST 2008
59 - improve readline speed for very long "lines".
60 - backport to 5.8.8.
61
624.8 Thu Oct 2 13:34:40 CEST 2008
63 - new function Coro::AIO::aio_wait.
64 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
65 - greatly speed up and reduce memory usage of Coro::AIO requests.
66 - implement some other µ-optimisations.
67
684.749 Mon Sep 29 14:40:12 CEST 2008
69 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
70
714.748 Sat Sep 27 14:03:19 CEST 2008
72 - implement, but do not document, PerlIO::cede(granularity).
73 - Coro::Storable forgot to wrap Storable::pstore.
74 - work around the multitude of leaks and memory corruption
75 bugs in PerlIO::via by using our own C-level perliol. As a side
76 effect, Coro::Storable is now much, much, much faster.
77
784.747 Tue Sep 23 01:59:41 CEST 2008
79 - fix a per-cv memleak (one empty array was leaked per
80 code reference).
81 - avoid a crash in coro->call|rss when the coroutine was already
82 destroyed (most noticably when using Coro::Debug::ps :)
83 - also protect *Storable::FILE.
84 - push up default storable granularity to 20ms.
85
864.746 Sun Sep 21 03:22:20 CEST 2008
87 - be more insistent on locking Storable against reentrancy
88 in Coro::Storable.
89 - move swap_def?v and throw to Coro::State, as documented.
90
914.745 Thu Jul 24 00:14:38 CEST 2008
92 - remove debugging code related to MgPV_nolen_const, also try to
93 make it compile with perl 5.8.6 (yes, apple apparently loves
94 outdated software). Reported by John S.
95
964.744 Tue Jul 8 22:06:35 CEST 2008
97 - correctly provide default DIE/WARN handlers as documented.
98 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
99 implementations inside perl work fine for us.
100
1014.743 Mon Jun 16 00:21:57 CEST 2008
102 - when using Coro::EV without running EV::loop it could
103 result in busy-waiting for events, this has been fixed.
104 - reduce codesize and improve performance by using EV_DEFAULT_UC.
105
1064.742 Sat May 31 14:10:21 CEST 2008
107 - implement a workaround for (some) perl <5.8.8 versions.
108 - require EV 3.3+.
109
1104.741 Fri May 30 23:33:09 CEST 2008
111 - tell netbsd how utterly broken their imitation of an OS is
112 and refuse to build by default if pthreads are in use.
113 - switch to "s" method on all bsd's by default, as their ucontext
114 stuff seems just too broken.
115 - fix a bug in Coro::Select.
116
1174.74 Thu May 29 20:05:31 CEST 2008
118 - do not test Coro::LWP for lack of dependencies.
119 - do not test Coro::Select for lack of working perls.
120
1214.73 Thu May 29 2008
122 - fix a bug in Coro::EV which would cause it to block despite
123 there being runnable coroutines.
124 - sprinkle "no warnings" freely over everything, also suppress
125 warnings for some other modules.
126 - fix typo in WSAEWOULDBLOCK.
127
1284.72 Sun May 25 05:14:36 CEST 2008
129 - tweak META.yaml a bit, unfortunately, there is no documented way
130 to have optional dependencies with CPAN. doh :(
131 - avoid running some tests on windows because they would fail due to
132 perl bug (broken fork, broken pipes...).
133 - work around perl on windows bugs where perl returns undocumented
134 error codes for sysread, syswrite etc. by taking advantage
135 of AnyEvent's workaround for that problem.
136 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
137 a common perl implementation bug on windows.
138 - use unix domain sockets in testsuite to work around
139 common perl implementation bugs on widows (they are emulated by
140 tcp sockets on windows. ugh.)
141
1424.71 Sat May 24 20:01:27 CEST 2008
143 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
144 - take advantage of async name resolution of AnyEvent::Util.
145 - work around brutal inet_aton override in Coro::LWP.
146 - take advantage of the readyhook in Coro::EV, for smoother
147 scheduling.
148
1494.7 Sun May 11 00:32:19 CEST 2008
150 - completely reworked the Coro manpage.
151 - added Coro::AnyEvent, generic event loop integration.
152 - implement cancel, ready and kill commands in Coro::Debug.
153 - document find_coro in Coro::Debug.
154 - incompatible change: rename has_stack to has_cctx.
155 - Coro::AIO and Coro::BDB no longer force event model detection,
156 use AnyEvent::AIO and AnyEvent::BDB.
157
1584.6 Sat Apr 26 10:05:14 CEST 2008
159 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
160 by the Coro module. It was a mistake to have it in the first place.
161 - (experimental) support for activestate perl 5.10 (method "w").
162 - (experimental) support for strawberry perl (method "a").
163 - coro_sigelem_set did not return a value although it had to,
164 actual impact unknown.
165
1664.51 Mon Apr 14 13:28:27 CEST 2008
167 - make it compile again on 5.8.
168
1694.50 Thu Apr 10 09:43:17 CEST 2008
170 - I did it twice! (see 4.49).
171
1724.49 Sun Apr 6 21:23:31 CEST 2008
173 - grr, instead of compiling the recent changes on 5.10 only they
174 were compiled on 5.8 only.
175
1764.48 Sun Apr 6 20:36:46 CEST 2008
177 - allow coroutine switches during eval's under 5.10.x, as it seems
178 the parser is a per-interpreter option now, so this is safe
179 (this might fix the odd crashing bug).
180 - drop support for 5.9.x versions: they are dead, jim.
181
1824.47 Sun Apr 6 00:37:52 CEST 2008
183 - force cctx allocation on API calls: we know we need to force one
184 and gcc actually manages to confuse our heuristic nowadays,
185 leading to crashes and worse.
186 - document force_cctx.
187
1884.46 Fri Apr 4 22:05:43 CEST 2008
189 - upgrade libcoro, resulting in pthread-backend (which was only created
190 to fulfill the rules of the programming languages shootout).
191
1924.45 Thu Mar 13 11:55:36 CET 2008
193 - fix a file leaking bug in eg/mhyttpd that would allow
194 downloading of any file (reported by oesi).
195 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
196 (also add testcase).
197 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
198 apparently having loads of fun with that).
199 - implement $state->swap_defsv, ->swap_defav and document ->throw.
200
2014.4 Wed Feb 13 16:44:29 CET 2008
202 - only assume .cfi pseudo isns exist on GNU/Linux.
203 - add get_timed method to Coro::Channels.
204 - fixed Coro::Debug manpage.
205 - perl 5.11 compatbility improvement based on patch by
206 Andreas König.
207
2084.37 Sun Jan 20 11:25:23 CET 2008
209 - soften the check for an active parser for perl 5.10 (avoiding
210 "Coro::State::transfer called while parsing" in many cases).
211
2124.36 Sun Jan 13 10:53:56 CET 2008
213 - reset diehook when terminating from an async_pool as to not
214 trigger any __DIE__ handlers.
215
2164.35 Sun Jan 13 04:14:13 CET 2008
217 - "bt" debug command now displays any exceptions
218 from longmess and also skips the innermost
219 stackframes, giving more useufl output.
220 - allow backtraces in coroutines blocked in Coro::EV,
221 at a <1% speed hit.
222 - handle localising of $SIG{__WARN__} and __DIE__
223 properly (with a proper amount of dirty hacking).
224
2254.34 Sat Dec 22 17:49:53 CET 2007
226 - upgrade to EV version 2.0 API.
227
2284.33 Mon Dec 17 08:36:12 CET 2007
229 - make Coro::AIO etc. loadable in the absence of EV.
230
2314.32 Mon Dec 17 07:46:02 CET 2007
232 - majorly improved Coro::Handle's performance with Coro::EV.
233 - implemented the until now mythical Coro::BDB module.
234 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
235
2364.31 Wed Dec 5 12:32:39 CET 2007
237 - remove warn statement form Coro::Util.
238
2394.3 Tue Dec 4 20:33:14 CET 2007
240 - calls to the idle function could cause stack corruption
241 when the stack changed.
242 - do no longer rely on the presence of EV::DNS if EV is used
243 (because it is gone), but instead take avdantage of EV::ADNS
244 if available.
245 - add ($) prototypes to all functions in Coro::Storable.
246 - use a conventional (safer) idle callback in Coro::EV.
247 - do accept quickly in Coro::Debug to avoid endless loops.
248
2494.22 Fri Nov 30 16:04:04 CET 2007
250 - really use optimised versions for Event and EV in Coro::Util
251 and Coro::Handle.
252
2534.21 Sun Nov 25 10:48:59 CET 2007
254 - fix a spurious memory read.
255 - Coro::EV no longer keeps the eventloop "alive".
256
2574.2 Fri Nov 9 20:47:05 CET 2007
258 - enable/disable tracing from a new coroutine, not a pooled one.
259 - fix a memleak in Coro::Event.
260 - removed killall call from fork_eval.
261 - made sure store_fd is already loaded so that fork_eval
262 does not have to parse autoload in each subprocess.
263 - only use assembly method if -O switch is in $Config{optimize}.
264 - add (optional) Coro::EV module, so far the best event loop module
265 directly supported by Coro.
266 - if the event model is EV, use EV::DNS to resolve
267 stuff in Coro::Util.
268 - don't get confused by multiple event notifications in Coro::Handle.
269 - initial support for EV (libevent interface).
270 - require Event and EV using configure_requires, to force their existance.
271
2724.13 Wed Oct 24 07:26:45 CEST 2007
273 - add Coro::Storable::blocking_thaw.
274 - use a vastly more complicated technique to localise
275 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
276 - use a coroutine for the idle callback Coro::Event,
277 instead of running Event in the current coroutine context.
278 This also catches recursive invocations.
279 - actually report fork errors in gethostbyname and inet_aton.
280
2814.11 Thu Oct 11 02:40:24 CEST 2007
282 - port to threaded perls.
283
2844.1 Thu Oct 11 02:38:16 CEST 2007
285 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
286 be local to each coro (see Coro::State).
287 - incompatible change: for very deep reasons, cede and cede_notself
288 cannot return anything, so nothing will be returned.
289 - possibly bring back 5.10 compatibility (untested).
290 - work around stupid (and wrong) warning on 5.10 :(.
291 - overlay the saved state over the context stack. This saves
292 a few hundred bytes per coroutine on average and also
293 speeds up context switching a bit.
294 - further tune default stack sizes.
295 - (more) correctly calculate stack usage in coro_rss.
296 - Coro::Storable::blocking_* did not properly lock
297 resulting in races between coroutines.
298 - added Coro::Storable::guard.
299 - stopping to trace a coroutine could destroy the cctx of
300 an unrelated coroutine.
301 - explain the relationship between Perl and C coroutines in
302 more detail in Coro::State.
303 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
304 causing a fork per resolve. This also affected Coro::Socket.
305 - switch to a separate stack in $coro->call/eval to avoid
306 invalidating pointers.
307
3084.03 Sat Oct 6 21:24:00 CEST 2007
309 - added Coro::throw method.
310 - minor code cleanups.
311
3124.02 Sat Oct 6 02:36:47 CEST 2007
313 - fix a very minor per-coroutine memleak (a single codereference).
314 - fixed a bug where the currently in-use c context would be freed
315 prematurely (can happen only when programs change the stacksize
316 or use tracing).
317 - tracing can no longer keep a coro alive after it terminated.
318 - do static branch prediction in the common path for gcc. gives
319 about 2-5% speed improvement here.
320
3214.01 Fri Oct 5 22:10:49 CEST 2007
322 - instead of recreating *a* standard output handle we simply
323 use STDOUT, which is faster and hopefully more robust.
324
3254.0 Fri Oct 5 12:56:00 CEST 2007
326 - incompatibly changed Coro::Storable::freeze.
327 - major new feature: added Coro::Debug, for interactive coroutine
328 debugging, tracing and much more.
329 - major bug fix: unbelievable, but true: $_, $/ and many other
330 "saved" variables actually weren't being saved. This has been fixed,
331 of course, while increasing performance while losing all the save
332 flags.
333 - save flags are gone, and all the api functions dealing with them.
334 - added Coro::Semaphore::adjust.
335 - added Coro::Util::fork_eval.
336 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
337 - added Coro::killall.
338 - reduce initial stack sizes to allow for "micro-coroutines".
339 - better async_pool resource management, moved parts of async_pool
340 handling to XS (major speed improvement).
341 - actually croak before modifying important data structures.
342 - refuse to transfer while compiling.
343 - possibly support eval EXPR better now.
344 - enable assembly per default on linux+bsd x86+amd64.
345 - all internal members were renamed _something for easier subclassing.
346 - many minor tweaks.
347
3483.63 Wed May 16 14:10:06 CEST 2007
349 - implement handcoded assembly for x86/amd64 SVR ABI.
350
3513.62 Fri Apr 27 21:36:06 CEST 2007
352 - upgrade libcoro (which might set unwind info correctly).
353 - change default on linux to setjmp/longjmp.
354
3553.61 Thu Apr 19 12:36:18 CEST 2007
356 - Coro::Storable caused an endless loop when thawing invalid
357 pst's sometimes.
358 - use a Semaphore in Coro::Storable, as Storable doesn't
359 seem to be reentrant (although it is documented to
360 be threadsafe...).
361 - fix Coro::Signal to bring back the original unreliable
362 but stateful semantics.
363 - fixed a lot of typos in Coro.pm (patch submitted by David
364 Steinbrunner, which applied flawlessly).
365
3663.6 Sat Apr 14 17:13:31 CEST 2007
367 - added some bugfixes to get eg/myhttpd working again.
368 - added Coro::Storable for often-cede'ing freeze/thaw.
369 - try to do a clean exit when a coroutine calls exit
370 (EXPERIMENTAL).
371 - got rid of indirect call through _coro_init.
372 - updated the partly antique examples in eg/ to
373 work again and be a bit less magic, too.
374 - fixed Coro::Signal semantics to work as documented again.
375
3763.55 Sun Mar 25 01:20:47 CET 2007
377 - add SAVE_DEFFH to save the default filehandle and enable
378 it by default.
379 - finally move socket-operations from Coro::Socket to Coro::Handle
380 to be able to unblock foreign sockets.
381 - add Coro::State::save_also and guarded_save.
7 - add count accessor to Coro::Semaphore. 382 - add count accessor to Coro::Semaphore.
383 - add Coro::State::cctx_stacksize.
384 - just for the fun of it, do not rely on implicit context,
385 which can dramatically improve performance, but people
386 using windows-process-emulation perls do not care much about
387 performance.
8 388
93.51 Sun Mar 4 14:18:04 CET 2007 3893.51 Sun Mar 4 14:18:04 CET 2007
10 - fixed a problem when you weakened references to Coro::State's 390 - fixed a problem when you weakened references to Coro::State's
11 (patch by Yuval Kogman). 391 (patch by Yuval Kogman).
12 392

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines