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

Comparing Coro/Changes (file contents):
Revision 1.225 by root, Wed Apr 18 13:46:32 2007 UTC vs.
Revision 1.419 by root, Sun Nov 16 08:59:16 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
5TODO: should explore PerlIO::via::CoroCede and PerlIO::via::CoroAIO. 5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
6 7
73.61 8TODO: testcancel
9TODO: aio
10TODO: Coro::Semaphore
115.0
12 - NEW ARCHITECTURE: use 4.912 for the stable version.
13 - bump API version to 7 - all dependents must be recompiled.
14 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
15 - async_pool did free a scalar value twice
16 ("Attempt to unreference...").
17 - perl's process emulation is now not even theoretically supported
18 anymore.
19 - directly patch the entersub opcode calling SLF functions (cede,
20 transfer and so on). this does speed up context switching, but
21 more importanly, it frees us from the hardcoded behaviour of
22 entersub, so we might actually be able to return something from
23 those functions and atcually create new ones.
24 - take advantage of __builtin_frame_address on gcc.
25 - expose THX in coroapi (not sure whether this was a wise decision,
26 as "threaded" perls are running at half speed anyways).
27 - implement execute_slf (schedule-like-function) interface that makes
28 it possible to implement schedule-like-functions in XS.
29 - use new SLF interface to massively speed up Coro::EV by roughly a factor
30 of two.
31 - removed timed_down and timed_guard functions - they were not being used
32 anyways and should be replaced by a more generic mechanism - and were
33 annoying to support anyways :
34 - used new SLF interface to massively speed up Coro::Semaphore by a factor
35 of three.
36 - forgot to include Coro/libcoro/README in the dist for all these years.
37 - work around a freebsd pthreads bug (manual testcancel is required as
38 pthread_cond_wait isn't a cancellation point on freebsd).
39
404.913 Sat Nov 15 07:58:28 CET 2008
41 - async_pool did free a scalar value twice
42 ("Attempt to unreference...").
43
444.912 Thu Nov 13 18:31:23 CET 2008
45 - minor cleanups.
46 - use much larger stacks on linux and perl < 5.8.8.
47 - Coro::Debug::new_unix_server did not unlink the socket
48 when destroyed.
49
504.911 Tue Nov 11 04:26:17 CET 2008
51 - "port" to threaded perls.
52
534.91 Mon Nov 10 05:36:38 CET 2008
54 - the ->throw exception object no longer leaks.
55 - creating a new cctx leaked a scopestack entry (memleak).
56 - new coroutines didn't get created with a zero flags field
57 (unknown impact).
58 - calling ->throw on a not-yet-started coroutine should now work
59 instead of being ignored.
60 - ->throw is now supported on Coro::State objects.
61 - clean up cctx creation code a bit.
62 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
63
644.9 Sat Nov 8 17:45:27 CET 2008
65 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
66 - (libcoro) no longer leak threads in the experimental pthread backend,
67 also speed it up considerably.
68 - (libcoro) do not rely on makecontext passing void *'s unscathed.
69 - fix compiletime dependencies on libcoro in the Makefile.
70 - cctx_count wasn't always updated properly.
71 - Coro::State::cctx_stacksize wasn't applied correctly.
72 - new function Coro::State::cctx_max_idle.
73 - the default max number of idle C contexts is now 4.
74 - (libcoro) try harder to get _setjmp/_longjmp.
75 - (libcoro) cleanup and extend the libcoro API to officially
76 allow the creation of empty source contexts.
77 - very experimental workaround for the totally broken netbsd platform.
78 - tried to hack around openbsd bugs.
79
804.804 Wed Nov 5 16:36:00 CET 2008
81 - Coro::Debug::new_unix_server would not create a non-blocking listening
82 socket, sometimes causing freezes.
83 - (libcoro) fix misaligned stack points for setjmp and assembly
84 methods, which can cause crashes on x86/x86_64 with a sufficiently
85 aggressive compiler.
86 - new function: Coro::Debug::new_tcp_server.
87 - move ->throw into the Coro class because it only works on coro objects.
88
894.803 Mon Nov 3 17:16:12 CET 2008
90 - (libcoro) use a global asm statement to become independent of gcc
91 otpimisations for CORO_ASM (thanks to pippijn for the idea).
92 - try to workaround yet another broken bsd, this time dragonfly.
93
944.802 Thu Oct 30 10:56:12 CET 2008
95 - support -fno-omit-frame-pointer on x86 with the assembly method.
96 - tune 01_unblock.t tests a bit.
97
984.801 Wed Oct 22 18:33:37 CEST 2008
99 - improve readline speed for very long "lines".
100 - backport to 5.8.8.
101
1024.8 Thu Oct 2 13:34:40 CEST 2008
103 - new function Coro::AIO::aio_wait.
104 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
105 - greatly speed up and reduce memory usage of Coro::AIO requests.
106 - implement some other µ-optimisations.
107
1084.749 Mon Sep 29 14:40:12 CEST 2008
109 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
110
1114.748 Sat Sep 27 14:03:19 CEST 2008
112 - implement, but do not document, PerlIO::cede(granularity).
113 - Coro::Storable forgot to wrap Storable::pstore.
114 - work around the multitude of leaks and memory corruption
115 bugs in PerlIO::via by using our own C-level perliol. As a side
116 effect, Coro::Storable is now much, much, much faster.
117
1184.747 Tue Sep 23 01:59:41 CEST 2008
119 - fix a per-cv memleak (one empty array was leaked per
120 code reference).
121 - avoid a crash in coro->call|rss when the coroutine was already
122 destroyed (most noticably when using Coro::Debug::ps :)
123 - also protect *Storable::FILE.
124 - push up default storable granularity to 20ms.
125
1264.746 Sun Sep 21 03:22:20 CEST 2008
127 - be more insistent on locking Storable against reentrancy
128 in Coro::Storable.
129 - move swap_def?v and throw to Coro::State, as documented.
130
1314.745 Thu Jul 24 00:14:38 CEST 2008
132 - remove debugging code related to MgPV_nolen_const, also try to
133 make it compile with perl 5.8.6 (yes, apple apparently loves
134 outdated software). Reported by John S.
135
1364.744 Tue Jul 8 22:06:35 CEST 2008
137 - correctly provide default DIE/WARN handlers as documented.
138 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
139 implementations inside perl work fine for us.
140
1414.743 Mon Jun 16 00:21:57 CEST 2008
142 - when using Coro::EV without running EV::loop it could
143 result in busy-waiting for events, this has been fixed.
144 - reduce codesize and improve performance by using EV_DEFAULT_UC.
145
1464.742 Sat May 31 14:10:21 CEST 2008
147 - implement a workaround for (some) perl <5.8.8 versions.
148 - require EV 3.3+.
149
1504.741 Fri May 30 23:33:09 CEST 2008
151 - tell netbsd how utterly broken their imitation of an OS is
152 and refuse to build by default if pthreads are in use.
153 - switch to "s" method on all bsd's by default, as their ucontext
154 stuff seems just too broken.
155 - fix a bug in Coro::Select.
156
1574.74 Thu May 29 20:05:31 CEST 2008
158 - do not test Coro::LWP for lack of dependencies.
159 - do not test Coro::Select for lack of working perls.
160
1614.73 Thu May 29 2008
162 - fix a bug in Coro::EV which would cause it to block despite
163 there being runnable coroutines.
164 - sprinkle "no warnings" freely over everything, also suppress
165 warnings for some other modules.
166 - fix typo in WSAEWOULDBLOCK.
167
1684.72 Sun May 25 05:14:36 CEST 2008
169 - tweak META.yaml a bit, unfortunately, there is no documented way
170 to have optional dependencies with CPAN. doh :(
171 - avoid running some tests on windows because they would fail due to
172 perl bug (broken fork, broken pipes...).
173 - work around perl on windows bugs where perl returns undocumented
174 error codes for sysread, syswrite etc. by taking advantage
175 of AnyEvent's workaround for that problem.
176 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
177 a common perl implementation bug on windows.
178 - use unix domain sockets in testsuite to work around
179 common perl implementation bugs on widows (they are emulated by
180 tcp sockets on windows. ugh.)
181
1824.71 Sat May 24 20:01:27 CEST 2008
183 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
184 - take advantage of async name resolution of AnyEvent::Util.
185 - work around brutal inet_aton override in Coro::LWP.
186 - take advantage of the readyhook in Coro::EV, for smoother
187 scheduling.
188
1894.7 Sun May 11 00:32:19 CEST 2008
190 - completely reworked the Coro manpage.
191 - added Coro::AnyEvent, generic event loop integration.
192 - implement cancel, ready and kill commands in Coro::Debug.
193 - document find_coro in Coro::Debug.
194 - incompatible change: rename has_stack to has_cctx.
195 - Coro::AIO and Coro::BDB no longer force event model detection,
196 use AnyEvent::AIO and AnyEvent::BDB.
197
1984.6 Sat Apr 26 10:05:14 CEST 2008
199 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
200 by the Coro module. It was a mistake to have it in the first place.
201 - (experimental) support for activestate perl 5.10 (method "w").
202 - (experimental) support for strawberry perl (method "a").
203 - coro_sigelem_set did not return a value although it had to,
204 actual impact unknown.
205
2064.51 Mon Apr 14 13:28:27 CEST 2008
207 - make it compile again on 5.8.
208
2094.50 Thu Apr 10 09:43:17 CEST 2008
210 - I did it twice! (see 4.49).
211
2124.49 Sun Apr 6 21:23:31 CEST 2008
213 - grr, instead of compiling the recent changes on 5.10 only they
214 were compiled on 5.8 only.
215
2164.48 Sun Apr 6 20:36:46 CEST 2008
217 - allow coroutine switches during eval's under 5.10.x, as it seems
218 the parser is a per-interpreter option now, so this is safe
219 (this might fix the odd crashing bug).
220 - drop support for 5.9.x versions: they are dead, jim.
221
2224.47 Sun Apr 6 00:37:52 CEST 2008
223 - force cctx allocation on API calls: we know we need to force one
224 and gcc actually manages to confuse our heuristic nowadays,
225 leading to crashes and worse.
226 - document force_cctx.
227
2284.46 Fri Apr 4 22:05:43 CEST 2008
229 - upgrade libcoro, resulting in pthread-backend (which was only created
230 to fulfill the rules of the programming languages shootout).
231
2324.45 Thu Mar 13 11:55:36 CET 2008
233 - fix a file leaking bug in eg/mhyttpd that would allow
234 downloading of any file (reported by oesi).
235 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
236 (also add testcase).
237 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
238 apparently having loads of fun with that).
239 - implement $state->swap_defsv, ->swap_defav and document ->throw.
240
2414.4 Wed Feb 13 16:44:29 CET 2008
242 - only assume .cfi pseudo isns exist on GNU/Linux.
243 - add get_timed method to Coro::Channels.
244 - fixed Coro::Debug manpage.
245 - perl 5.11 compatbility improvement based on patch by
246 Andreas König.
247
2484.37 Sun Jan 20 11:25:23 CET 2008
249 - soften the check for an active parser for perl 5.10 (avoiding
250 "Coro::State::transfer called while parsing" in many cases).
251
2524.36 Sun Jan 13 10:53:56 CET 2008
253 - reset diehook when terminating from an async_pool as to not
254 trigger any __DIE__ handlers.
255
2564.35 Sun Jan 13 04:14:13 CET 2008
257 - "bt" debug command now displays any exceptions
258 from longmess and also skips the innermost
259 stackframes, giving more useufl output.
260 - allow backtraces in coroutines blocked in Coro::EV,
261 at a <1% speed hit.
262 - handle localising of $SIG{__WARN__} and __DIE__
263 properly (with a proper amount of dirty hacking).
264
2654.34 Sat Dec 22 17:49:53 CET 2007
266 - upgrade to EV version 2.0 API.
267
2684.33 Mon Dec 17 08:36:12 CET 2007
269 - make Coro::AIO etc. loadable in the absence of EV.
270
2714.32 Mon Dec 17 07:46:02 CET 2007
272 - majorly improved Coro::Handle's performance with Coro::EV.
273 - implemented the until now mythical Coro::BDB module.
274 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
275
2764.31 Wed Dec 5 12:32:39 CET 2007
277 - remove warn statement form Coro::Util.
278
2794.3 Tue Dec 4 20:33:14 CET 2007
280 - calls to the idle function could cause stack corruption
281 when the stack changed.
282 - do no longer rely on the presence of EV::DNS if EV is used
283 (because it is gone), but instead take avdantage of EV::ADNS
284 if available.
285 - add ($) prototypes to all functions in Coro::Storable.
286 - use a conventional (safer) idle callback in Coro::EV.
287 - do accept quickly in Coro::Debug to avoid endless loops.
288
2894.22 Fri Nov 30 16:04:04 CET 2007
290 - really use optimised versions for Event and EV in Coro::Util
291 and Coro::Handle.
292
2934.21 Sun Nov 25 10:48:59 CET 2007
294 - fix a spurious memory read.
295 - Coro::EV no longer keeps the eventloop "alive".
296
2974.2 Fri Nov 9 20:47:05 CET 2007
298 - enable/disable tracing from a new coroutine, not a pooled one.
299 - fix a memleak in Coro::Event.
300 - removed killall call from fork_eval.
301 - made sure store_fd is already loaded so that fork_eval
302 does not have to parse autoload in each subprocess.
303 - only use assembly method if -O switch is in $Config{optimize}.
304 - add (optional) Coro::EV module, so far the best event loop module
305 directly supported by Coro.
306 - if the event model is EV, use EV::DNS to resolve
307 stuff in Coro::Util.
308 - don't get confused by multiple event notifications in Coro::Handle.
309 - initial support for EV (libevent interface).
310 - require Event and EV using configure_requires, to force their existance.
311
3124.13 Wed Oct 24 07:26:45 CEST 2007
313 - add Coro::Storable::blocking_thaw.
314 - use a vastly more complicated technique to localise
315 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
316 - use a coroutine for the idle callback Coro::Event,
317 instead of running Event in the current coroutine context.
318 This also catches recursive invocations.
319 - actually report fork errors in gethostbyname and inet_aton.
320
3214.11 Thu Oct 11 02:40:24 CEST 2007
322 - port to threaded perls.
323
3244.1 Thu Oct 11 02:38:16 CEST 2007
325 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
326 be local to each coro (see Coro::State).
327 - incompatible change: for very deep reasons, cede and cede_notself
328 cannot return anything, so nothing will be returned.
329 - possibly bring back 5.10 compatibility (untested).
330 - work around stupid (and wrong) warning on 5.10 :(.
331 - overlay the saved state over the context stack. This saves
332 a few hundred bytes per coroutine on average and also
333 speeds up context switching a bit.
334 - further tune default stack sizes.
335 - (more) correctly calculate stack usage in coro_rss.
336 - Coro::Storable::blocking_* did not properly lock
337 resulting in races between coroutines.
338 - added Coro::Storable::guard.
339 - stopping to trace a coroutine could destroy the cctx of
340 an unrelated coroutine.
341 - explain the relationship between Perl and C coroutines in
342 more detail in Coro::State.
343 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
344 causing a fork per resolve. This also affected Coro::Socket.
345 - switch to a separate stack in $coro->call/eval to avoid
346 invalidating pointers.
347
3484.03 Sat Oct 6 21:24:00 CEST 2007
349 - added Coro::throw method.
350 - minor code cleanups.
351
3524.02 Sat Oct 6 02:36:47 CEST 2007
353 - fix a very minor per-coroutine memleak (a single codereference).
354 - fixed a bug where the currently in-use c context would be freed
355 prematurely (can happen only when programs change the stacksize
356 or use tracing).
357 - tracing can no longer keep a coro alive after it terminated.
358 - do static branch prediction in the common path for gcc. gives
359 about 2-5% speed improvement here.
360
3614.01 Fri Oct 5 22:10:49 CEST 2007
362 - instead of recreating *a* standard output handle we simply
363 use STDOUT, which is faster and hopefully more robust.
364
3654.0 Fri Oct 5 12:56:00 CEST 2007
366 - incompatibly changed Coro::Storable::freeze.
367 - major new feature: added Coro::Debug, for interactive coroutine
368 debugging, tracing and much more.
369 - major bug fix: unbelievable, but true: $_, $/ and many other
370 "saved" variables actually weren't being saved. This has been fixed,
371 of course, while increasing performance while losing all the save
372 flags.
373 - save flags are gone, and all the api functions dealing with them.
374 - added Coro::Semaphore::adjust.
375 - added Coro::Util::fork_eval.
376 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
377 - added Coro::killall.
378 - reduce initial stack sizes to allow for "micro-coroutines".
379 - better async_pool resource management, moved parts of async_pool
380 handling to XS (major speed improvement).
381 - actually croak before modifying important data structures.
382 - refuse to transfer while compiling.
383 - possibly support eval EXPR better now.
384 - enable assembly per default on linux+bsd x86+amd64.
385 - all internal members were renamed _something for easier subclassing.
386 - many minor tweaks.
387
3883.63 Wed May 16 14:10:06 CEST 2007
389 - implement handcoded assembly for x86/amd64 SVR ABI.
390
3913.62 Fri Apr 27 21:36:06 CEST 2007
392 - upgrade libcoro (which might set unwind info correctly).
393 - change default on linux to setjmp/longjmp.
394
3953.61 Thu Apr 19 12:36:18 CEST 2007
8 - Coro::Storable caused an endless loop when thawing invalid 396 - Coro::Storable caused an endless loop when thawing invalid
9 pst's sometimes. 397 pst's sometimes.
10 - use a Semaphore in Coro::Storable, as Storable doesn't 398 - use a Semaphore in Coro::Storable, as Storable doesn't
11 seem to be reentrant (although it is documented to 399 seem to be reentrant (although it is documented to
12 be threadsafe...). 400 be threadsafe...).
13 - fix Coro::Signal to bring back the original unreliable 401 - fix Coro::Signal to bring back the original unreliable
14 but stateful semantics. 402 but stateful semantics.
403 - fixed a lot of typos in Coro.pm (patch submitted by David
404 Steinbrunner, which applied flawlessly).
15 405
163.6 Sat Apr 14 17:13:31 CEST 2007 4063.6 Sat Apr 14 17:13:31 CEST 2007
17 - added some bugfixes to get eg/myhttpd working again. 407 - added some bugfixes to get eg/myhttpd working again.
18 - added Coro::Storable for often-cede'ing freeze/thaw. 408 - added Coro::Storable for often-cede'ing freeze/thaw.
19 - try to do a clean exit when a coroutine calls exit 409 - try to do a clean exit when a coroutine calls exit

Diff Legend

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