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

Comparing Coro/Changes (file contents):
Revision 1.392 by root, Fri Nov 7 20:12:26 2008 UTC vs.
Revision 1.551 by root, Sat Feb 19 06:51:22 2011 UTC

1Revision history for Perl extension Coro. 1Revision history for Perl extension Coro.
2 2
3TODO: should explore PerlIO::coroaio (perl leaks like hell). 3TODO: should explore PerlIO::coroaio (perl leaks like hell).
4TODO: maybe implement a default message channel, very much like Erlang's 4TODO: channel->maxsize(newsize)?
5 actor model (which is cool in a lot of important aspects (failures!), 5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx
6 but very lacking in others (higher level ipc)).
7 6
85.0 75.37 Sat Feb 19 07:49:44 CET 2011
8 - add a big "Coro thread life cycle" section to "man Coro".
9 - try a tentative workaround against the breakage that 5.13 has
10 introduced without depreciation period. sigh.
11 - no longer use Time::HiRes if gettimeofday is available, which
12 saves quite a lot of memory.
13
145.36 Sun Feb 13 05:33:41 CET 2011
15 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
16 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
17 method call.
18 - undocument Coro::Timer::sleep and obsolete whole module.
19 - optimise Coro::Timer::timeout memory and cpu usage.
20 - slightly updated Coro::Intro for recent changes.
21 - do not initialise PL_dirty anymore.
22
235.25 Thu Nov 11 01:08:39 CET 2010
24 - try a different approach on netbsd - netbsd 5 finally has marginally
25 working pthreads, but still broken ucontext/sigaltstack.
26 - openbsd 4.8 finally got their act together, Coro works out of the box
27 with asm, setjmp and pthreads (no change, just informational).
28
295.24 Sat Oct 23 11:27:12 CEST 2010
30 - port to the EV 4.0 API.
31 - work around bugs in mingw32, making strawberry perl work
32 out of the box.
33 - correctly modify Coro::AIO function prototypes
34 so that they reflect the "no optional parameters" rule.
35 - "ported" libcoro to C++.
36
375.23 Mon May 17 18:50:42 CEST 2010
38 - be more resistant to ordering changes when initialising
39 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
40 Waldorf).
41 - document that perl 5.12 deliberately removed support for cloning.
42
435.22 Wed Apr 14 03:55:35 CEST 2010
44 - correctly return udnef on errors in Coro::Handle::read/write
45 (testcase by Marc Mims).
46 - convert Coro::Util into a "perl compatibility wrapper" - the functions
47 are less useful now, but are drop-in replacements for existing
48 functions, listing better alternatives in the documentation. This also
49 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
50 with Coro::Util::inet_aton.
51 - do not override $Coro::idle unconditionally in Coro.pm, as other
52 modules could have provided their own idle coro already
53 (for exmaple, Coro::AnyEvent).
54 - fix Coro::Util::gethost* functions.
55 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
56
575.21 Wed Dec 16 07:19:51 CET 2009
58 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
59 together.
60 - add some examples on how to combine other event loops with Coro in
61 Coro::AnyEvent, and how to run it (and not to block). Seems to be
62 the most common source of confusion.
63 - try to catch people naively blocking in an event callback.
64 - work around the perl filehandle bug issue in conjunction with
65 older common::sense (as indirectly pointed out by ZSystem).
66 - clarify the "not from signal handlers" section.
67
685.2 Sun Oct 4 14:54:24 CEST 2009
69 - Coro::Storable destroyed the prototypes of the functions it wrapped.
70 - export rouse_cb and rouse_wait by default now.
71 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
72 - new method $state->swap_sv.
73 - added section on "windows process emulation" to the manpage,
74 after a not-so-fruitful (nor-friendly) "discussion" with chip
75 salzenberg (discussion implies arguments, but his only arguments
76 were ad-hominems, one wonders why he started it in the first
77 place). I hope this explains it well enough for him to understand,
78 and maybe well enough for others to understand.
79 - use common::sense everywhere now.
80 - idle callbacks are no longer supported, use idle coros instead.
81 - print a thread listing when a deadlock is detected.
82
835.17 Sat Aug 22 23:09:31 CEST 2009
84 - work around a bug in the perl debugger causing crashes
85 when running under the debugger by marking _pool_handler as nodebug.
86 - speed up Coro::async considerably.
87 - try some hacks to get netbsd to work "more often" - their broken
88 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
89 - convert Coro to AE 5.0 API.
90
915.162 Tue Jul 28 04:04:03 CEST 2009
92 - perl 5.8.2 is now minimum requirement.
93 - skip t/19_handle.t on broken windows perls.
94
955.161 Wed Jul 22 04:47:38 CEST 2009
96 - Coro::AnyEvent::poll could have a different prototype when EV was
97 used as backend (analyzed by Tatsuhiko Miyagawa).
98 - Coro::AnyEvent errornously initialised the event loop when loaded,
99 not on demand.
100 - try to workaround rare */t/01_unblock.t failures.
101
1025.16 Tue Jul 21 01:44:37 CEST 2009
103 - Coro::AnyEvent failed to hook into the event loop
104 when no threads had been readied between detecting
105 the event loop and actually running it.
106 - considerably speed up Coro::Select by taking avdantage
107 of AnyEvent > 4.8 and some other optimisations.
108 - implement paragraph readline mode in Coro::Handle
109 (based on patches by Zsbán Ambrus).
110 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
111 and analyzed by Yasuhiro MATSUMOTO).
112 - clarified libcoro license and copyright.
113 - someone stole my EXTRA_META!!!
114 - implement Coro::Select::patch_pp_sselect and it's brother,
115 for hardcode select overriding.
116
1175.151 Mon Jul 6 05:41:57 CEST 2009
118 - backport to windows process emulation code again (patch by
119 Yasuhiro MATSUMOTO).
120 - slightly update Coro::MakeMaker.
121
1225.15 Tue Jun 30 10:28:06 CEST 2009
123 - deprecate Coro::Socket, document how to get ipv6 support via
124 AnyEvent::Socket instead.
125 - implement signal->wait ($cb) interface, similar to semaphores.
126 - work around SvOK not supporting getmagic, so we have to getmagic
127 to test for undef :/ (reported by Matthias Waldorf).
128 - load Coro::AnyEvent in all modules using AnyEvent.
129 - work around perl corrupting our internal data structures,
130 reported by Tokuhiro Matsuno.
131 - enable per-coroutine real and cpu time gathering
132 (Coro::State::enable_times).
133
1345.14 Wed Jun 24 01:37:48 CEST 2009
135 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
136 as this seems to have been a difficult concept (poll, sleep, idle,
137 idle_upto).
138 - add Coro::AnyEvent::readable/writable functions.
139 - clarify Coro::EV/Event/AnyEvent manpages.
140 - free per-thread global scalars in the thread calling ->cancel, to
141 avoid crashes when $_, $@ etc., are magical but some of those
142 had already been freed.
143 - "unexperimentalise" the callback interface for Coro::Semaphore.
144 - speed up ready queue management/context switching by using a linked
145 list instead of an array (~5-10%).
146 - implement "watch" command in Coro::Debug shells.
147 - for fun, implement time-slicing as an example in the manpage.
148 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
149 Coro::Event, use the normal AnyEvent handling instead of dieing
150 (the same is true for Coro::Handle).
151 - properly document Coro::EV::timed_io_once.
152 - avoid unneccessary ->cancel calls in Coro::Handle.
153 - maybe make it work on mingw32 with win32 backend
154 (based on patch by Yasuhiro Matsumoto).
155
1565.132 Fri May 29 09:00:39 CEST 2009
157 - do not keep a reference to the argument itself in
158 Coro::Semaphore::guard, as it could change later.
159 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
160 should support a prepare callback.
161
1625.131 Mon Mar 16 23:20:37 CET 2009
163 - implement and document Coro->suspend, Coro->resume.
164 - fix Coro::Select implementation to not (often) close
165 the passed file descriptors (testcase provided by pippijn).
166
1675.13 Mon Dec 15 21:51:42 CET 2008
168 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
169 - don't set diehook to C<undef> but instead to NULL, to avoid
170 spurious warnings.
171 - fix a lot of bugs in Coro::SemaphoreSet.
172 - Coro::SemaphoreSet will less often create a semaphore needlessly.
173 - add Coro::SemaphoreSet::count and wait methods.
174 - take advantage of the new Guard module.
175 - deprecate Coro::guard.
176 - try to fix the dreaded 01_unblock tests once more. I hate it when
177 testsuites need more fixing than the code they are supposed to test.
178 - croak in more cases when a required callback isn't resolvable.
179 - fix some minor issues in Coro::State->call/eval.
180 - use current coroutine context instead of a temporary one
181 when temporarily switching to another coroutine.
182 - do not call C-level on_destroy handlers during global destruction,
183 to avoid needless segfaults.
184
1855.12 Sun Dec 7 13:30:38 CET 2008
186 - add default config for MirOS, which seems to be a bug-to-bug
187 compatible fork of openbsd ("world domination by releasing
188 openbsd cvs before the openbsd folks do it" or so :).
189 - free_padlist did destroy the names pad, not good, but didn't
190 seem to bother perl - this could fix issues such as eval ""
191 inside a function called from multiple coroutines.
192 - use a different method to detect destruction time.
193 - be more careful when freeing padlists just before global
194 destruction.
195 - fixed and expanded the call/cc example.
196 - renamed _terminate to _coro_run.
197 - new method Coro::Channel->shutdown.
198 - try pthreads on openbsd <4.4 (broken sigaltstack, will
199 pthreads fare better?).
200 - be less picky about destroying "a" running coroutine.
201
2025.11 Tue Nov 25 21:49:05 CET 2008
203 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
204 working code with failed assertions, introducing more bugs than
205 it fixes, requiring elaborate workarounds :(
206
2075.1 Mon Nov 24 08:54:59 CET 2008
208 - wrote a small introductory tutorial - Coro::Intro.
209 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
210 - Coro::Select did errornously dup the file descriptors
211 and didn't work with all AnyEvent backends.
212 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
213 LWP data transfers.
214 - disassociate c contexts from coro objects - this is agruably more
215 correct, but mostly allows sharing of cctxs between coro and state
216 objects, for added memory savings and speed increases.
217 - bumped $Coro::POOL_RSS up to 32kb by default.
218 - no longer set the optype to OP_CUSTOM, as B::* understandably
219 doesn't like this very much (and we *are* a type of entersub).
220 - implement state cloning, just to prove that call/cc can be done.
221 - automatically load Coro::AnyEvent in Coro::Handle.
222 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
223 do not have this method (and don't need it either).
224 - speed up generic anyevent methods in Coro::Handle by using rouse
225 callbacks.
226 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
227 others. It also makes the debugger happier, as you can trace
228 through the idle threads now.
229 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
230 - eg/event was pretty broken.
231 - better 5.8.6 compatibility.
232
2335.0 Thu Nov 20 10:35:05 CET 2008
234 - NEW ARCHITECTURE: use the latest 4.x version if you experience
235 stability issues.
236 - bump API version to 7 - all dependents must be recompiled.
237 - removed timed_* functions - they were not being used anyways
238 and should be replaced by a more generic mechanism -
239 and were annoying to support anyways :)
240 - removed SemaphoreSet's waiter method - use sem method instead.
241 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
242 - async_pool did free a scalar value twice
243 ("Attempt to unreference...").
244 - fix a longstanding bug where calling terminate on a coro that
245 was waiting for a semaphore that was just becoming available
246 would cause a deadlock (semaphore would get into a state where
247 it was available but waiters were still blocked).
248 - calling throw on a coroutine that is waiting for a semaphore will
249 no longer make it acquire the semaphore (and thus leak a count).
250 - perl's process emulation is now not even theoretically supported
251 anymore.
252 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
253 conversion of callback-style to blocking-style.
254 - new methods $coro->schedule_to and ->cede_to, to specifically
255 schedule or cede to a specific coroutine.
256 - new function Coro::Semaphore::wait.
257 - use named constants in Coro::Channel (Richard Hundt).
258 - directly patch the entersub opcode calling SLF functions (cede,
259 transfer and so on). this does speed up context switching, but
260 more importanly, it frees us from the hardcoded behaviour of
261 entersub, so we might actually be able to return something from
262 those functions and atcually create new ones.
263 - take advantage of __builtin_frame_address on gcc.
264 - expose THX in coroapi (not sure whether this was a wise decision,
265 as "threaded" perls are running at half speed anyways).
266 - implement execute_slf (schedule-like-function) interface that makes
267 it possible to implement schedule-like-functions in XS.
268 - use new SLF interface to massively speed up Coro::EV by roughly a
269 factor of two.
270 - used new SLF interface to massively speed up Coro::Semaphore by a
271 factor of three.
272 - used new SLF interface to speed up Coro::AIO by roughly a factor of
273 four and reduce its memory usage considerably.
274 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
275 for a nice speedup and vastly more correct behaviour. Also implement
276 a new method "sem" to get at the underlying semaphore object.
277 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
278 (in comparison) 20-40% speedup.
279 - used new SLF interface to reimplement Coro::Signal gaining
280 some unknown (because I was too lazy), but certain, speedup, and also
281 making signals reliable for the first time.
282 - used new SLF interface and other optimisations to speed up async_pool
283 by a factor of two. It also doesn't rely on perl's exception mechanism
284 to exit anymore. The overhead for terminating an async_pool, coro over
285 a normal async is now very small.
286 - sped up coroutine creation/destruction by 40%.
287 - forgot to include Coro/libcoro/README in the dist for all these years.
288 - work around a freebsd pthreads bug (manual testcancel is required as
289 pthread_cond_wait isn't a cancellation point on freebsd).
290 - use new rouse functions to speed up and simplify Coro::BDB.
291 - make "prefer perl native functions" work with threaded perls.
292 - condense Coro::Debug ps output, hint at v and w flags.
293 - (libcoro) lots of minor cleanups and portability improvements.
294
2954.914 Wed Nov 19 12:54:18 CET 2008
296 - fix a disastrous bug in the readline optimisation
297 introduced in 4.801.
298
2994.913 Sat Nov 15 07:58:28 CET 2008
300 - async_pool did free a scalar value twice
301 ("Attempt to unreference...").
302
3034.912 Thu Nov 13 18:31:23 CET 2008
304 - minor cleanups.
305 - use much larger stacks on linux and perl < 5.8.8.
306 - Coro::Debug::new_unix_server did not unlink the socket
307 when destroyed.
308
3094.911 Tue Nov 11 04:26:17 CET 2008
310 - "port" to threaded perls.
311
3124.91 Mon Nov 10 05:36:38 CET 2008
313 - the ->throw exception object no longer leaks.
314 - creating a new cctx leaked a scopestack entry (memleak).
315 - new coroutines didn't get created with a zero flags field
316 (unknown impact).
317 - calling ->throw on a not-yet-started coroutine should now work
318 instead of being ignored.
319 - ->throw is now supported on Coro::State objects.
320 - clean up cctx creation code a bit.
321 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
322
3234.9 Sat Nov 8 17:45:27 CET 2008
9 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 324 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
10 - (libcoro) no longer leak threads in the experimental pthread backend, 325 - (libcoro) no longer leak threads in the experimental pthread backend,
11 also speed it up considerably. 326 also speed it up considerably.
327 - (libcoro) do not rely on makecontext passing void *'s unscathed.
12 - fix compiletime dependencies on libcoro in the Makefile. 328 - fix compiletime dependencies on libcoro in the Makefile.
13 - cctx_count wasn't always updated properly. 329 - cctx_count wasn't always updated properly.
330 - Coro::State::cctx_stacksize wasn't applied correctly.
331 - new function Coro::State::cctx_max_idle.
332 - the default max number of idle C contexts is now 4.
333 - (libcoro) try harder to get _setjmp/_longjmp.
14 - (libcoro) cleanup and extend the libcoro API to officialy 334 - (libcoro) cleanup and extend the libcoro API to officially
15 allow creation of empty source contexts. 335 allow the creation of empty source contexts.
336 - very experimental workaround for the totally broken netbsd platform.
337 - tried to hack around openbsd bugs.
16 338
174.804 Wed Nov 5 16:36:00 CET 2008 3394.804 Wed Nov 5 16:36:00 CET 2008
18 - Coro::Debug::new_unix_server would not create a non-blocking listening 340 - Coro::Debug::new_unix_server would not create a non-blocking listening
19 socket, sometimes causing freezes. 341 socket, sometimes causing freezes.
20 - (libcoro) fix misaligned stack points for setjmp and assembly 342 - (libcoro) fix misaligned stack points for setjmp and assembly

Diff Legend

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