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

Comparing Coro/Changes (file contents):
Revision 1.179 by root, Mon Dec 11 20:54:45 2006 UTC vs.
Revision 1.364 by root, Mon Jul 21 03:34:04 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::via::CoroCede and PerlIO::via::CoroAIO.
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)).
4 7
8 - remove debugging code related to MgPV_nolen_const, also try to
9 make it compile with perl 5.8.6 (yes, apple apparently loves
10 outdated software). Reported by John S.
11
124.744 Tue Jul 8 22:06:35 CEST 2008
13 - correctly provide default DIE/WARN handlers as documented.
14 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
15 implementations inside perl work fine for us.
16
174.743 Mon Jun 16 00:21:57 CEST 2008
18 - when using Coro::EV without running EV::loop it could
19 result in busy-waiting for events, this has been fixed.
20 - reduce codesize and improve performance by using EV_DEFAULT_UC.
21
224.742 Sat May 31 14:10:21 CEST 2008
23 - implement a workaround for (some) perl <5.8.8 versions.
24 - require EV 3.3+.
25
264.741 Fri May 30 23:33:09 CEST 2008
27 - tell netbsd how utterly broken their imitation of an OS is
28 and refuse to build by default if pthreads are in use.
29 - switch to "s" method on all bsd's by default, as their ucontext
30 stuff seems just too broken.
31 - fix a bug in Coro::Select.
32
334.74 Thu May 29 20:05:31 CEST 2008
34 - do not test Coro::LWP for lack of dependencies.
35 - do not test Coro::Select for lack of working perls.
36
374.73 Thu May 29 2008
38 - fix a bug in Coro::EV which would cause it to block despite
39 there being runnable coroutines.
40 - sprinkle "no warnings" freely over everything, also suppress
41 warnings for some other modules.
42 - fix typo in WSAEWOULDBLOCK.
43
444.72 Sun May 25 05:14:36 CEST 2008
45 - tweak META.yaml a bit, unfortunately, there is no documented way
46 to have optional dependencies with CPAN. doh :(
47 - avoid running some tests on windows because they would fail due to
48 perl bug (broken fork, broken pipes...).
49 - work around perl on windows bugs where perl returns undocumented
50 error codes for sysread, syswrite etc. by taking advantage
51 of AnyEvent's workaround for that problem.
52 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
53 a common perl implementation bug on windows.
54 - use unix domain sockets in testsuite to work around
55 common perl implementation bugs on widows (they are emulated by
56 tcp sockets on windows. ugh.)
57
584.71 Sat May 24 20:01:27 CEST 2008
59 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
60 - take advantage of async name resolution of AnyEvent::Util.
61 - work around brutal inet_aton override in Coro::LWP.
62 - take advantage of the readyhook in Coro::EV, for smoother
63 scheduling.
64
654.7 Sun May 11 00:32:19 CEST 2008
66 - completely reworked the Coro manpage.
67 - added Coro::AnyEvent, generic event loop integration.
68 - implement cancel, ready and kill commands in Coro::Debug.
69 - document find_coro in Coro::Debug.
70 - incompatible change: rename has_stack to has_cctx.
71 - Coro::AIO and Coro::BDB no longer force event model detection,
72 use AnyEvent::AIO and AnyEvent::BDB.
73
744.6 Sat Apr 26 10:05:14 CEST 2008
75 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
76 by the Coro module. It was a mistake to have it in the first place.
77 - (experimental) support for activestate perl 5.10 (method "w").
78 - (experimental) support for strawberry perl (method "a").
79 - coro_sigelem_set did not return a value although it had to,
80 actual impact unknown.
81
824.51 Mon Apr 14 13:28:27 CEST 2008
83 - make it compile again on 5.8.
84
854.50 Thu Apr 10 09:43:17 CEST 2008
86 - I did it twice! (see 4.49).
87
884.49 Sun Apr 6 21:23:31 CEST 2008
89 - grr, instead of compiling the recent changes on 5.10 only they
90 were compiled on 5.8 only.
91
924.48 Sun Apr 6 20:36:46 CEST 2008
93 - allow coroutine switches during eval's under 5.10.x, as it seems
94 the parser is a per-interpreter option now, so this is safe
95 (this might fix the odd crashing bug).
96 - drop support for 5.9.x versions: they are dead, jim.
97
984.47 Sun Apr 6 00:37:52 CEST 2008
99 - force cctx allocation on API calls: we know we need to force one
100 and gcc actually manages to confuse our heuristic nowadays,
101 leading to crashes and worse.
102 - document force_cctx.
103
1044.46 Fri Apr 4 22:05:43 CEST 2008
105 - upgrade libcoro, resulting in pthread-backend (which was only created
106 to fulfill the rules of the programming languages shootout).
107
1084.45 Thu Mar 13 11:55:36 CET 2008
109 - fix a file leaking bug in eg/mhyttpd that would allow
110 downloading of any file (reported by oesi).
111 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
112 (also add testcase).
113 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
114 apparently having loads of fun with that).
115 - implement $state->swap_defsv, ->swap_defav and document ->throw.
116
1174.4 Wed Feb 13 16:44:29 CET 2008
118 - only assume .cfi pseudo isns exist on GNU/Linux.
119 - add get_timed method to Coro::Channels.
120 - fixed Coro::Debug manpage.
121 - perl 5.11 compatbility improvement based on patch by
122 Andreas König.
123
1244.37 Sun Jan 20 11:25:23 CET 2008
125 - soften the check for an active parser for perl 5.10 (avoiding
126 "Coro::State::transfer called while parsing" in many cases).
127
1284.36 Sun Jan 13 10:53:56 CET 2008
129 - reset diehook when terminating from an async_pool as to not
130 trigger any __DIE__ handlers.
131
1324.35 Sun Jan 13 04:14:13 CET 2008
133 - "bt" debug command now displays any exceptions
134 from longmess and also skips the innermost
135 stackframes, giving more useufl output.
136 - allow backtraces in coroutines blocked in Coro::EV,
137 at a <1% speed hit.
138 - handle localising of $SIG{__WARN__} and __DIE__
139 properly (with a proper amount of dirty hacking).
140
1414.34 Sat Dec 22 17:49:53 CET 2007
142 - upgrade to EV version 2.0 API.
143
1444.33 Mon Dec 17 08:36:12 CET 2007
145 - make Coro::AIO etc. loadable in the absence of EV.
146
1474.32 Mon Dec 17 07:46:02 CET 2007
148 - majorly improved Coro::Handle's performance with Coro::EV.
149 - implemented the until now mythical Coro::BDB module.
150 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
151
1524.31 Wed Dec 5 12:32:39 CET 2007
153 - remove warn statement form Coro::Util.
154
1554.3 Tue Dec 4 20:33:14 CET 2007
156 - calls to the idle function could cause stack corruption
157 when the stack changed.
158 - do no longer rely on the presence of EV::DNS if EV is used
159 (because it is gone), but instead take avdantage of EV::ADNS
160 if available.
161 - add ($) prototypes to all functions in Coro::Storable.
162 - use a conventional (safer) idle callback in Coro::EV.
163 - do accept quickly in Coro::Debug to avoid endless loops.
164
1654.22 Fri Nov 30 16:04:04 CET 2007
166 - really use optimised versions for Event and EV in Coro::Util
167 and Coro::Handle.
168
1694.21 Sun Nov 25 10:48:59 CET 2007
170 - fix a spurious memory read.
171 - Coro::EV no longer keeps the eventloop "alive".
172
1734.2 Fri Nov 9 20:47:05 CET 2007
174 - enable/disable tracing from a new coroutine, not a pooled one.
175 - fix a memleak in Coro::Event.
176 - removed killall call from fork_eval.
177 - made sure store_fd is already loaded so that fork_eval
178 does not have to parse autoload in each subprocess.
179 - only use assembly method if -O switch is in $Config{optimize}.
180 - add (optional) Coro::EV module, so far the best event loop module
181 directly supported by Coro.
182 - if the event model is EV, use EV::DNS to resolve
183 stuff in Coro::Util.
184 - don't get confused by multiple event notifications in Coro::Handle.
185 - initial support for EV (libevent interface).
186 - require Event and EV using configure_requires, to force their existance.
187
1884.13 Wed Oct 24 07:26:45 CEST 2007
189 - add Coro::Storable::blocking_thaw.
190 - use a vastly more complicated technique to localise
191 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
192 - use a coroutine for the idle callback Coro::Event,
193 instead of running Event in the current coroutine context.
194 This also catches recursive invocations.
195 - actually report fork errors in gethostbyname and inet_aton.
196
1974.11 Thu Oct 11 02:40:24 CEST 2007
198 - port to threaded perls.
199
2004.1 Thu Oct 11 02:38:16 CEST 2007
201 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
202 be local to each coro (see Coro::State).
203 - incompatible change: for very deep reasons, cede and cede_notself
204 cannot return anything, so nothing will be returned.
205 - possibly bring back 5.10 compatibility (untested).
206 - work around stupid (and wrong) warning on 5.10 :(.
207 - overlay the saved state over the context stack. This saves
208 a few hundred bytes per coroutine on average and also
209 speeds up context switching a bit.
210 - further tune default stack sizes.
211 - (more) correctly calculate stack usage in coro_rss.
212 - Coro::Storable::blocking_* did not properly lock
213 resulting in races between coroutines.
214 - added Coro::Storable::guard.
215 - stopping to trace a coroutine could destroy the cctx of
216 an unrelated coroutine.
217 - explain the relationship between Perl and C coroutines in
218 more detail in Coro::State.
219 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
220 causing a fork per resolve. This also affected Coro::Socket.
221 - switch to a separate stack in $coro->call/eval to avoid
222 invalidating pointers.
223
2244.03 Sat Oct 6 21:24:00 CEST 2007
225 - added Coro::throw method.
226 - minor code cleanups.
227
2284.02 Sat Oct 6 02:36:47 CEST 2007
229 - fix a very minor per-coroutine memleak (a single codereference).
230 - fixed a bug where the currently in-use c context would be freed
231 prematurely (can happen only when programs change the stacksize
232 or use tracing).
233 - tracing can no longer keep a coro alive after it terminated.
234 - do static branch prediction in the common path for gcc. gives
235 about 2-5% speed improvement here.
236
2374.01 Fri Oct 5 22:10:49 CEST 2007
238 - instead of recreating *a* standard output handle we simply
239 use STDOUT, which is faster and hopefully more robust.
240
2414.0 Fri Oct 5 12:56:00 CEST 2007
242 - incompatibly changed Coro::Storable::freeze.
243 - major new feature: added Coro::Debug, for interactive coroutine
244 debugging, tracing and much more.
245 - major bug fix: unbelievable, but true: $_, $/ and many other
246 "saved" variables actually weren't being saved. This has been fixed,
247 of course, while increasing performance while losing all the save
248 flags.
249 - save flags are gone, and all the api functions dealing with them.
250 - added Coro::Semaphore::adjust.
251 - added Coro::Util::fork_eval.
252 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
253 - added Coro::killall.
254 - reduce initial stack sizes to allow for "micro-coroutines".
255 - better async_pool resource management, moved parts of async_pool
256 handling to XS (major speed improvement).
257 - actually croak before modifying important data structures.
258 - refuse to transfer while compiling.
259 - possibly support eval EXPR better now.
260 - enable assembly per default on linux+bsd x86+amd64.
261 - all internal members were renamed _something for easier subclassing.
262 - many minor tweaks.
263
2643.63 Wed May 16 14:10:06 CEST 2007
265 - implement handcoded assembly for x86/amd64 SVR ABI.
266
2673.62 Fri Apr 27 21:36:06 CEST 2007
268 - upgrade libcoro (which might set unwind info correctly).
269 - change default on linux to setjmp/longjmp.
270
2713.61 Thu Apr 19 12:36:18 CEST 2007
272 - Coro::Storable caused an endless loop when thawing invalid
273 pst's sometimes.
274 - use a Semaphore in Coro::Storable, as Storable doesn't
275 seem to be reentrant (although it is documented to
276 be threadsafe...).
277 - fix Coro::Signal to bring back the original unreliable
278 but stateful semantics.
279 - fixed a lot of typos in Coro.pm (patch submitted by David
280 Steinbrunner, which applied flawlessly).
281
2823.6 Sat Apr 14 17:13:31 CEST 2007
283 - added some bugfixes to get eg/myhttpd working again.
284 - added Coro::Storable for often-cede'ing freeze/thaw.
285 - try to do a clean exit when a coroutine calls exit
286 (EXPERIMENTAL).
287 - got rid of indirect call through _coro_init.
288 - updated the partly antique examples in eg/ to
289 work again and be a bit less magic, too.
290 - fixed Coro::Signal semantics to work as documented again.
291
2923.55 Sun Mar 25 01:20:47 CET 2007
293 - add SAVE_DEFFH to save the default filehandle and enable
294 it by default.
295 - finally move socket-operations from Coro::Socket to Coro::Handle
296 to be able to unblock foreign sockets.
297 - add Coro::State::save_also and guarded_save.
298 - add count accessor to Coro::Semaphore.
299 - add Coro::State::cctx_stacksize.
300 - just for the fun of it, do not rely on implicit context,
301 which can dramatically improve performance, but people
302 using windows-process-emulation perls do not care much about
303 performance.
304
3053.51 Sun Mar 4 14:18:04 CET 2007
306 - fixed a problem when you weakened references to Coro::State's
307 (patch by Yuval Kogman).
308
3093.501 Wed Feb 28 12:44:07 CET 2007
310 - rename some global symbols as macosx from hell redefines
311 them without asking.
312
3133.5 Tue Feb 13 20:22:53 CET 2007
314 - do AnyEvent model detection earlier, avoiding problems
315 caused by first using AnyEvent and later Coro::Event.
316 - implement and document Coro::Event event objects.
317 - fix a potential problem in Coro::Event causing crashes.
318 - initialise PL_comppad when creating a new coroutine,
319 avoids crashes on early coro destruction.
320
3213.41 Mon Jan 22 19:19:49 CET 2007
322 - readline on Coro::Handle did not support undefined $/,
323 nor did it deliver partial lines on EOF or error.
324 - implement malloc fallback for stack allocation because
325 stupid broken idiotic OSX has a stupid broken
326 idiotic fits-the-whole-os mmap "implementation" and
327 my dick feels longer if Coro is portable even to
328 obsolete platforms.
329
3303.4 Fri Jan 19 21:52:54 CET 2007
331 - remove t/09_timer.t, as it isn't really testing much
332 but was rather flaky in practise.
333 - async_pool coro would keep arguments and callback alive until
334 it was reused.
335 - cancellation of a coroutine could cause spurious idle calls
336 in cede_notself.
337
3383.3 Sat Jan 6 03:45:00 CET 2007
339 - implement $coro->on_destroy.
340 - Coro::Event blocking semantics have been changed,
341 documented and - hopefully - improved.
342 - fix nice adding, not subtracting, from priority.
343 - fix ->prio and api_is_ready (patch by Mark Hinds).
344 - fixed an assert ("... == prev__cctx->idle_te")
345 that could errronously trigger.
346 - fix various large and small memleaks.
347 - use a (hopefully) more stable cancel implementation
348 that immediately frees the coroutine data.
349 - cede/cede_notself return a status now.
350 - added Coro::guard function.
351 - added a global coroutine pool for jobs (on my machine,
352 I can create and execute 48k simple coros/s with async,
353 and 128k coros with async_pool).
354 - Coro::AIO now uses the coroutine priority as io priority.
355
3563.2 Fri Dec 22 05:07:09 CET 2006
5 - improve portability to slightly older perls. 357 - improve portability to slightly older perls.
358 - use cleaner coroutine destruction.
359 - simplify configuration for users.
360 - optionally (unrecommended) prefer perl functions over
361 their coro replacements.
6 362
73.11 Tue Dec 5 13:11:24 CET 2006 3633.11 Tue Dec 5 13:11:24 CET 2006
8 - fixed some bogus assert's, but as perl.h disables assert even 364 - fixed some bogus assert's, but as perl.h disables assert even
9 without NDEBUG (thank you very much), not too many people should 365 without NDEBUG (thank you very much), not too many people should
10 notice (that did include myself). Andreas König noticed, though :) 366 notice (that did include myself). Andreas König noticed, though :)

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines