1 |
Revision history for Perl extension Coro. |
2 |
|
3 |
TODO: should explore PerlIO::coroaio (perl leaks like hell). |
4 |
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 |
but very lacking in others (higher level ipc)). |
7 |
|
8 |
TODO: testcancel |
9 |
TODO: guard as SLF - to avoid deadlocks. |
10 |
- avoid deadlock in Coro::Channel if maxsize == 1 (Richard Hundt). |
11 |
|
12 |
5.0 |
13 |
- NEW ARCHITECTURE: use the latest 4.x version if you experience |
14 |
stability issues. |
15 |
- bump API version to 7 - all dependents must be recompiled. |
16 |
- removed timed_down and timed_guard functions - they were not being |
17 |
used anyways and should be replaced by a more generic mechanism - |
18 |
and were annoying to support anyways :) |
19 |
- removed SemaphoreSet's waiter method - use sem instead. |
20 |
- Coro::Semaphore->adjust didn't correctly wake up enough waiters. |
21 |
- async_pool did free a scalar value twice |
22 |
("Attempt to unreference..."). |
23 |
- fix a longstanding bug where calling terminate on a coro that |
24 |
was waiting for a semaphore that was just becoming available |
25 |
would cause a deadlock (semaphore would get into a state where |
26 |
it was available but waiters were waiting(. |
27 |
- perl's process emulation is now not even theoretically supported |
28 |
anymore. |
29 |
- use named constants in Coro::Channel (Richard Hundt). |
30 |
- directly patch the entersub opcode calling SLF functions (cede, |
31 |
transfer and so on). this does speed up context switching, but |
32 |
more importanly, it frees us from the hardcoded behaviour of |
33 |
entersub, so we might actually be able to return something from |
34 |
those functions and atcually create new ones. |
35 |
- take advantage of __builtin_frame_address on gcc. |
36 |
- expose THX in coroapi (not sure whether this was a wise decision, |
37 |
as "threaded" perls are running at half speed anyways). |
38 |
- implement execute_slf (schedule-like-function) interface that makes |
39 |
it possible to implement schedule-like-functions in XS. |
40 |
- use new SLF interface to massively speed up Coro::EV by roughly a |
41 |
factor of two. |
42 |
- used new SLF interface to massively speed up Coro::Semaphore by a |
43 |
factor of three. |
44 |
- used new SLF interface to speed up Coro::AIO by roughly a factor of |
45 |
four and reduce its memory usage considerably. |
46 |
- implement Coro::SemaphoreSet purely in terms of Coro::Semaphore, |
47 |
for a nice speedup and vastly more correct behaviour. Also implement |
48 |
a new method "sem" to get at the underlying semaphore object. |
49 |
- forgot to include Coro/libcoro/README in the dist for all these years. |
50 |
- work around a freebsd pthreads bug (manual testcancel is required as |
51 |
pthread_cond_wait isn't a cancellation point on freebsd). |
52 |
|
53 |
4.913 Sat Nov 15 07:58:28 CET 2008 |
54 |
- async_pool did free a scalar value twice |
55 |
("Attempt to unreference..."). |
56 |
|
57 |
4.912 Thu Nov 13 18:31:23 CET 2008 |
58 |
- minor cleanups. |
59 |
- use much larger stacks on linux and perl < 5.8.8. |
60 |
- Coro::Debug::new_unix_server did not unlink the socket |
61 |
when destroyed. |
62 |
|
63 |
4.911 Tue Nov 11 04:26:17 CET 2008 |
64 |
- "port" to threaded perls. |
65 |
|
66 |
4.91 Mon Nov 10 05:36:38 CET 2008 |
67 |
- the ->throw exception object no longer leaks. |
68 |
- creating a new cctx leaked a scopestack entry (memleak). |
69 |
- new coroutines didn't get created with a zero flags field |
70 |
(unknown impact). |
71 |
- calling ->throw on a not-yet-started coroutine should now work |
72 |
instead of being ignored. |
73 |
- ->throw is now supported on Coro::State objects. |
74 |
- clean up cctx creation code a bit. |
75 |
- entersub is actually a UNOP, not a LOGOP (not a bugfix). |
76 |
|
77 |
4.9 Sat Nov 8 17:45:27 CET 2008 |
78 |
- (libcoro) did not preserve rbp with CORO_ASM (we are getting there). |
79 |
- (libcoro) no longer leak threads in the experimental pthread backend, |
80 |
also speed it up considerably. |
81 |
- (libcoro) do not rely on makecontext passing void *'s unscathed. |
82 |
- fix compiletime dependencies on libcoro in the Makefile. |
83 |
- cctx_count wasn't always updated properly. |
84 |
- Coro::State::cctx_stacksize wasn't applied correctly. |
85 |
- new function Coro::State::cctx_max_idle. |
86 |
- the default max number of idle C contexts is now 4. |
87 |
- (libcoro) try harder to get _setjmp/_longjmp. |
88 |
- (libcoro) cleanup and extend the libcoro API to officially |
89 |
allow the creation of empty source contexts. |
90 |
- very experimental workaround for the totally broken netbsd platform. |
91 |
- tried to hack around openbsd bugs. |
92 |
|
93 |
4.804 Wed Nov 5 16:36:00 CET 2008 |
94 |
- Coro::Debug::new_unix_server would not create a non-blocking listening |
95 |
socket, sometimes causing freezes. |
96 |
- (libcoro) fix misaligned stack points for setjmp and assembly |
97 |
methods, which can cause crashes on x86/x86_64 with a sufficiently |
98 |
aggressive compiler. |
99 |
- new function: Coro::Debug::new_tcp_server. |
100 |
- move ->throw into the Coro class because it only works on coro objects. |
101 |
|
102 |
4.803 Mon Nov 3 17:16:12 CET 2008 |
103 |
- (libcoro) use a global asm statement to become independent of gcc |
104 |
otpimisations for CORO_ASM (thanks to pippijn for the idea). |
105 |
- try to workaround yet another broken bsd, this time dragonfly. |
106 |
|
107 |
4.802 Thu Oct 30 10:56:12 CET 2008 |
108 |
- support -fno-omit-frame-pointer on x86 with the assembly method. |
109 |
- tune 01_unblock.t tests a bit. |
110 |
|
111 |
4.801 Wed Oct 22 18:33:37 CEST 2008 |
112 |
- improve readline speed for very long "lines". |
113 |
- backport to 5.8.8. |
114 |
|
115 |
4.8 Thu Oct 2 13:34:40 CEST 2008 |
116 |
- new function Coro::AIO::aio_wait. |
117 |
- Coro.:AIO and Coro::BDB now "use Coro::AnyEvent". |
118 |
- greatly speed up and reduce memory usage of Coro::AIO requests. |
119 |
- implement some other µ-optimisations. |
120 |
|
121 |
4.749 Mon Sep 29 14:40:12 CEST 2008 |
122 |
- port to slow and broken pseudo-threaded perls. (courtesy pippijn). |
123 |
|
124 |
4.748 Sat Sep 27 14:03:19 CEST 2008 |
125 |
- implement, but do not document, PerlIO::cede(granularity). |
126 |
- Coro::Storable forgot to wrap Storable::pstore. |
127 |
- work around the multitude of leaks and memory corruption |
128 |
bugs in PerlIO::via by using our own C-level perliol. As a side |
129 |
effect, Coro::Storable is now much, much, much faster. |
130 |
|
131 |
4.747 Tue Sep 23 01:59:41 CEST 2008 |
132 |
- fix a per-cv memleak (one empty array was leaked per |
133 |
code reference). |
134 |
- avoid a crash in coro->call|rss when the coroutine was already |
135 |
destroyed (most noticably when using Coro::Debug::ps :) |
136 |
- also protect *Storable::FILE. |
137 |
- push up default storable granularity to 20ms. |
138 |
|
139 |
4.746 Sun Sep 21 03:22:20 CEST 2008 |
140 |
- be more insistent on locking Storable against reentrancy |
141 |
in Coro::Storable. |
142 |
- move swap_def?v and throw to Coro::State, as documented. |
143 |
|
144 |
4.745 Thu Jul 24 00:14:38 CEST 2008 |
145 |
- remove debugging code related to MgPV_nolen_const, also try to |
146 |
make it compile with perl 5.8.6 (yes, apple apparently loves |
147 |
outdated software). Reported by John S. |
148 |
|
149 |
4.744 Tue Jul 8 22:06:35 CEST 2008 |
150 |
- correctly provide default DIE/WARN handlers as documented. |
151 |
- also overwrite PL_vtbl_sigelem.svt_clear, even though current |
152 |
implementations inside perl work fine for us. |
153 |
|
154 |
4.743 Mon Jun 16 00:21:57 CEST 2008 |
155 |
- when using Coro::EV without running EV::loop it could |
156 |
result in busy-waiting for events, this has been fixed. |
157 |
- reduce codesize and improve performance by using EV_DEFAULT_UC. |
158 |
|
159 |
4.742 Sat May 31 14:10:21 CEST 2008 |
160 |
- implement a workaround for (some) perl <5.8.8 versions. |
161 |
- require EV 3.3+. |
162 |
|
163 |
4.741 Fri May 30 23:33:09 CEST 2008 |
164 |
- tell netbsd how utterly broken their imitation of an OS is |
165 |
and refuse to build by default if pthreads are in use. |
166 |
- switch to "s" method on all bsd's by default, as their ucontext |
167 |
stuff seems just too broken. |
168 |
- fix a bug in Coro::Select. |
169 |
|
170 |
4.74 Thu May 29 20:05:31 CEST 2008 |
171 |
- do not test Coro::LWP for lack of dependencies. |
172 |
- do not test Coro::Select for lack of working perls. |
173 |
|
174 |
4.73 Thu May 29 2008 |
175 |
- fix a bug in Coro::EV which would cause it to block despite |
176 |
there being runnable coroutines. |
177 |
- sprinkle "no warnings" freely over everything, also suppress |
178 |
warnings for some other modules. |
179 |
- fix typo in WSAEWOULDBLOCK. |
180 |
|
181 |
4.72 Sun May 25 05:14:36 CEST 2008 |
182 |
- tweak META.yaml a bit, unfortunately, there is no documented way |
183 |
to have optional dependencies with CPAN. doh :( |
184 |
- avoid running some tests on windows because they would fail due to |
185 |
perl bug (broken fork, broken pipes...). |
186 |
- work around perl on windows bugs where perl returns undocumented |
187 |
error codes for sysread, syswrite etc. by taking advantage |
188 |
of AnyEvent's workaround for that problem. |
189 |
- use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around |
190 |
a common perl implementation bug on windows. |
191 |
- use unix domain sockets in testsuite to work around |
192 |
common perl implementation bugs on widows (they are emulated by |
193 |
tcp sockets on windows. ugh.) |
194 |
|
195 |
4.71 Sat May 24 20:01:27 CEST 2008 |
196 |
- fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). |
197 |
- take advantage of async name resolution of AnyEvent::Util. |
198 |
- work around brutal inet_aton override in Coro::LWP. |
199 |
- take advantage of the readyhook in Coro::EV, for smoother |
200 |
scheduling. |
201 |
|
202 |
4.7 Sun May 11 00:32:19 CEST 2008 |
203 |
- completely reworked the Coro manpage. |
204 |
- added Coro::AnyEvent, generic event loop integration. |
205 |
- implement cancel, ready and kill commands in Coro::Debug. |
206 |
- document find_coro in Coro::Debug. |
207 |
- incompatible change: rename has_stack to has_cctx. |
208 |
- Coro::AIO and Coro::BDB no longer force event model detection, |
209 |
use AnyEvent::AIO and AnyEvent::BDB. |
210 |
|
211 |
4.6 Sat Apr 26 10:05:14 CEST 2008 |
212 |
- INCOMPATIBLE CHANGE: sub/code attributes are no longer supported |
213 |
by the Coro module. It was a mistake to have it in the first place. |
214 |
- (experimental) support for activestate perl 5.10 (method "w"). |
215 |
- (experimental) support for strawberry perl (method "a"). |
216 |
- coro_sigelem_set did not return a value although it had to, |
217 |
actual impact unknown. |
218 |
|
219 |
4.51 Mon Apr 14 13:28:27 CEST 2008 |
220 |
- make it compile again on 5.8. |
221 |
|
222 |
4.50 Thu Apr 10 09:43:17 CEST 2008 |
223 |
- I did it twice! (see 4.49). |
224 |
|
225 |
4.49 Sun Apr 6 21:23:31 CEST 2008 |
226 |
- grr, instead of compiling the recent changes on 5.10 only they |
227 |
were compiled on 5.8 only. |
228 |
|
229 |
4.48 Sun Apr 6 20:36:46 CEST 2008 |
230 |
- allow coroutine switches during eval's under 5.10.x, as it seems |
231 |
the parser is a per-interpreter option now, so this is safe |
232 |
(this might fix the odd crashing bug). |
233 |
- drop support for 5.9.x versions: they are dead, jim. |
234 |
|
235 |
4.47 Sun Apr 6 00:37:52 CEST 2008 |
236 |
- force cctx allocation on API calls: we know we need to force one |
237 |
and gcc actually manages to confuse our heuristic nowadays, |
238 |
leading to crashes and worse. |
239 |
- document force_cctx. |
240 |
|
241 |
4.46 Fri Apr 4 22:05:43 CEST 2008 |
242 |
- upgrade libcoro, resulting in pthread-backend (which was only created |
243 |
to fulfill the rules of the programming languages shootout). |
244 |
|
245 |
4.45 Thu Mar 13 11:55:36 CET 2008 |
246 |
- fix a file leaking bug in eg/mhyttpd that would allow |
247 |
downloading of any file (reported by oesi). |
248 |
- fix deadlock bug in Coro::Channel (reported by Richard Hundt) |
249 |
(also add testcase). |
250 |
- support Broadcast option in Coro::Socket (patch by Richard Hundt, |
251 |
apparently having loads of fun with that). |
252 |
- implement $state->swap_defsv, ->swap_defav and document ->throw. |
253 |
|
254 |
4.4 Wed Feb 13 16:44:29 CET 2008 |
255 |
- only assume .cfi pseudo isns exist on GNU/Linux. |
256 |
- add get_timed method to Coro::Channels. |
257 |
- fixed Coro::Debug manpage. |
258 |
- perl 5.11 compatbility improvement based on patch by |
259 |
Andreas König. |
260 |
|
261 |
4.37 Sun Jan 20 11:25:23 CET 2008 |
262 |
- soften the check for an active parser for perl 5.10 (avoiding |
263 |
"Coro::State::transfer called while parsing" in many cases). |
264 |
|
265 |
4.36 Sun Jan 13 10:53:56 CET 2008 |
266 |
- reset diehook when terminating from an async_pool as to not |
267 |
trigger any __DIE__ handlers. |
268 |
|
269 |
4.35 Sun Jan 13 04:14:13 CET 2008 |
270 |
- "bt" debug command now displays any exceptions |
271 |
from longmess and also skips the innermost |
272 |
stackframes, giving more useufl output. |
273 |
- allow backtraces in coroutines blocked in Coro::EV, |
274 |
at a <1% speed hit. |
275 |
- handle localising of $SIG{__WARN__} and __DIE__ |
276 |
properly (with a proper amount of dirty hacking). |
277 |
|
278 |
4.34 Sat Dec 22 17:49:53 CET 2007 |
279 |
- upgrade to EV version 2.0 API. |
280 |
|
281 |
4.33 Mon Dec 17 08:36:12 CET 2007 |
282 |
- make Coro::AIO etc. loadable in the absence of EV. |
283 |
|
284 |
4.32 Mon Dec 17 07:46:02 CET 2007 |
285 |
- majorly improved Coro::Handle's performance with Coro::EV. |
286 |
- implemented the until now mythical Coro::BDB module. |
287 |
- specialcase EV in Coro::AIO and Coro::BDB for extra speed. |
288 |
|
289 |
4.31 Wed Dec 5 12:32:39 CET 2007 |
290 |
- remove warn statement form Coro::Util. |
291 |
|
292 |
4.3 Tue Dec 4 20:33:14 CET 2007 |
293 |
- calls to the idle function could cause stack corruption |
294 |
when the stack changed. |
295 |
- do no longer rely on the presence of EV::DNS if EV is used |
296 |
(because it is gone), but instead take avdantage of EV::ADNS |
297 |
if available. |
298 |
- add ($) prototypes to all functions in Coro::Storable. |
299 |
- use a conventional (safer) idle callback in Coro::EV. |
300 |
- do accept quickly in Coro::Debug to avoid endless loops. |
301 |
|
302 |
4.22 Fri Nov 30 16:04:04 CET 2007 |
303 |
- really use optimised versions for Event and EV in Coro::Util |
304 |
and Coro::Handle. |
305 |
|
306 |
4.21 Sun Nov 25 10:48:59 CET 2007 |
307 |
- fix a spurious memory read. |
308 |
- Coro::EV no longer keeps the eventloop "alive". |
309 |
|
310 |
4.2 Fri Nov 9 20:47:05 CET 2007 |
311 |
- enable/disable tracing from a new coroutine, not a pooled one. |
312 |
- fix a memleak in Coro::Event. |
313 |
- removed killall call from fork_eval. |
314 |
- made sure store_fd is already loaded so that fork_eval |
315 |
does not have to parse autoload in each subprocess. |
316 |
- only use assembly method if -O switch is in $Config{optimize}. |
317 |
- add (optional) Coro::EV module, so far the best event loop module |
318 |
directly supported by Coro. |
319 |
- if the event model is EV, use EV::DNS to resolve |
320 |
stuff in Coro::Util. |
321 |
- don't get confused by multiple event notifications in Coro::Handle. |
322 |
- initial support for EV (libevent interface). |
323 |
- require Event and EV using configure_requires, to force their existance. |
324 |
|
325 |
4.13 Wed Oct 24 07:26:45 CEST 2007 |
326 |
- add Coro::Storable::blocking_thaw. |
327 |
- use a vastly more complicated technique to localise |
328 |
$SIG{__WARN/DIE__} that also works on perls <= 5.8.8. |
329 |
- use a coroutine for the idle callback Coro::Event, |
330 |
instead of running Event in the current coroutine context. |
331 |
This also catches recursive invocations. |
332 |
- actually report fork errors in gethostbyname and inet_aton. |
333 |
|
334 |
4.11 Thu Oct 11 02:40:24 CEST 2007 |
335 |
- port to threaded perls. |
336 |
|
337 |
4.1 Thu Oct 11 02:38:16 CEST 2007 |
338 |
- incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now |
339 |
be local to each coro (see Coro::State). |
340 |
- incompatible change: for very deep reasons, cede and cede_notself |
341 |
cannot return anything, so nothing will be returned. |
342 |
- possibly bring back 5.10 compatibility (untested). |
343 |
- work around stupid (and wrong) warning on 5.10 :(. |
344 |
- overlay the saved state over the context stack. This saves |
345 |
a few hundred bytes per coroutine on average and also |
346 |
speeds up context switching a bit. |
347 |
- further tune default stack sizes. |
348 |
- (more) correctly calculate stack usage in coro_rss. |
349 |
- Coro::Storable::blocking_* did not properly lock |
350 |
resulting in races between coroutines. |
351 |
- added Coro::Storable::guard. |
352 |
- stopping to trace a coroutine could destroy the cctx of |
353 |
an unrelated coroutine. |
354 |
- explain the relationship between Perl and C coroutines in |
355 |
more detail in Coro::State. |
356 |
- Coro::Util::inet_aton did not short-circuit dotted quad forms, |
357 |
causing a fork per resolve. This also affected Coro::Socket. |
358 |
- switch to a separate stack in $coro->call/eval to avoid |
359 |
invalidating pointers. |
360 |
|
361 |
4.03 Sat Oct 6 21:24:00 CEST 2007 |
362 |
- added Coro::throw method. |
363 |
- minor code cleanups. |
364 |
|
365 |
4.02 Sat Oct 6 02:36:47 CEST 2007 |
366 |
- fix a very minor per-coroutine memleak (a single codereference). |
367 |
- fixed a bug where the currently in-use c context would be freed |
368 |
prematurely (can happen only when programs change the stacksize |
369 |
or use tracing). |
370 |
- tracing can no longer keep a coro alive after it terminated. |
371 |
- do static branch prediction in the common path for gcc. gives |
372 |
about 2-5% speed improvement here. |
373 |
|
374 |
4.01 Fri Oct 5 22:10:49 CEST 2007 |
375 |
- instead of recreating *a* standard output handle we simply |
376 |
use STDOUT, which is faster and hopefully more robust. |
377 |
|
378 |
4.0 Fri Oct 5 12:56:00 CEST 2007 |
379 |
- incompatibly changed Coro::Storable::freeze. |
380 |
- major new feature: added Coro::Debug, for interactive coroutine |
381 |
debugging, tracing and much more. |
382 |
- major bug fix: unbelievable, but true: $_, $/ and many other |
383 |
"saved" variables actually weren't being saved. This has been fixed, |
384 |
of course, while increasing performance while losing all the save |
385 |
flags. |
386 |
- save flags are gone, and all the api functions dealing with them. |
387 |
- added Coro::Semaphore::adjust. |
388 |
- added Coro::Util::fork_eval. |
389 |
- added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}. |
390 |
- added Coro::killall. |
391 |
- reduce initial stack sizes to allow for "micro-coroutines". |
392 |
- better async_pool resource management, moved parts of async_pool |
393 |
handling to XS (major speed improvement). |
394 |
- actually croak before modifying important data structures. |
395 |
- refuse to transfer while compiling. |
396 |
- possibly support eval EXPR better now. |
397 |
- enable assembly per default on linux+bsd x86+amd64. |
398 |
- all internal members were renamed _something for easier subclassing. |
399 |
- many minor tweaks. |
400 |
|
401 |
3.63 Wed May 16 14:10:06 CEST 2007 |
402 |
- implement handcoded assembly for x86/amd64 SVR ABI. |
403 |
|
404 |
3.62 Fri Apr 27 21:36:06 CEST 2007 |
405 |
- upgrade libcoro (which might set unwind info correctly). |
406 |
- change default on linux to setjmp/longjmp. |
407 |
|
408 |
3.61 Thu Apr 19 12:36:18 CEST 2007 |
409 |
- Coro::Storable caused an endless loop when thawing invalid |
410 |
pst's sometimes. |
411 |
- use a Semaphore in Coro::Storable, as Storable doesn't |
412 |
seem to be reentrant (although it is documented to |
413 |
be threadsafe...). |
414 |
- fix Coro::Signal to bring back the original unreliable |
415 |
but stateful semantics. |
416 |
- fixed a lot of typos in Coro.pm (patch submitted by David |
417 |
Steinbrunner, which applied flawlessly). |
418 |
|
419 |
3.6 Sat Apr 14 17:13:31 CEST 2007 |
420 |
- added some bugfixes to get eg/myhttpd working again. |
421 |
- added Coro::Storable for often-cede'ing freeze/thaw. |
422 |
- try to do a clean exit when a coroutine calls exit |
423 |
(EXPERIMENTAL). |
424 |
- got rid of indirect call through _coro_init. |
425 |
- updated the partly antique examples in eg/ to |
426 |
work again and be a bit less magic, too. |
427 |
- fixed Coro::Signal semantics to work as documented again. |
428 |
|
429 |
3.55 Sun Mar 25 01:20:47 CET 2007 |
430 |
- add SAVE_DEFFH to save the default filehandle and enable |
431 |
it by default. |
432 |
- finally move socket-operations from Coro::Socket to Coro::Handle |
433 |
to be able to unblock foreign sockets. |
434 |
- add Coro::State::save_also and guarded_save. |
435 |
- add count accessor to Coro::Semaphore. |
436 |
- add Coro::State::cctx_stacksize. |
437 |
- just for the fun of it, do not rely on implicit context, |
438 |
which can dramatically improve performance, but people |
439 |
using windows-process-emulation perls do not care much about |
440 |
performance. |
441 |
|
442 |
3.51 Sun Mar 4 14:18:04 CET 2007 |
443 |
- fixed a problem when you weakened references to Coro::State's |
444 |
(patch by Yuval Kogman). |
445 |
|
446 |
3.501 Wed Feb 28 12:44:07 CET 2007 |
447 |
- rename some global symbols as macosx from hell redefines |
448 |
them without asking. |
449 |
|
450 |
3.5 Tue Feb 13 20:22:53 CET 2007 |
451 |
- do AnyEvent model detection earlier, avoiding problems |
452 |
caused by first using AnyEvent and later Coro::Event. |
453 |
- implement and document Coro::Event event objects. |
454 |
- fix a potential problem in Coro::Event causing crashes. |
455 |
- initialise PL_comppad when creating a new coroutine, |
456 |
avoids crashes on early coro destruction. |
457 |
|
458 |
3.41 Mon Jan 22 19:19:49 CET 2007 |
459 |
- readline on Coro::Handle did not support undefined $/, |
460 |
nor did it deliver partial lines on EOF or error. |
461 |
- implement malloc fallback for stack allocation because |
462 |
stupid broken idiotic OSX has a stupid broken |
463 |
idiotic fits-the-whole-os mmap "implementation" and |
464 |
my dick feels longer if Coro is portable even to |
465 |
obsolete platforms. |
466 |
|
467 |
3.4 Fri Jan 19 21:52:54 CET 2007 |
468 |
- remove t/09_timer.t, as it isn't really testing much |
469 |
but was rather flaky in practise. |
470 |
- async_pool coro would keep arguments and callback alive until |
471 |
it was reused. |
472 |
- cancellation of a coroutine could cause spurious idle calls |
473 |
in cede_notself. |
474 |
|
475 |
3.3 Sat Jan 6 03:45:00 CET 2007 |
476 |
- implement $coro->on_destroy. |
477 |
- Coro::Event blocking semantics have been changed, |
478 |
documented and - hopefully - improved. |
479 |
- fix nice adding, not subtracting, from priority. |
480 |
- fix ->prio and api_is_ready (patch by Mark Hinds). |
481 |
- fixed an assert ("... == prev__cctx->idle_te") |
482 |
that could errronously trigger. |
483 |
- fix various large and small memleaks. |
484 |
- use a (hopefully) more stable cancel implementation |
485 |
that immediately frees the coroutine data. |
486 |
- cede/cede_notself return a status now. |
487 |
- added Coro::guard function. |
488 |
- added a global coroutine pool for jobs (on my machine, |
489 |
I can create and execute 48k simple coros/s with async, |
490 |
and 128k coros with async_pool). |
491 |
- Coro::AIO now uses the coroutine priority as io priority. |
492 |
|
493 |
3.2 Fri Dec 22 05:07:09 CET 2006 |
494 |
- improve portability to slightly older perls. |
495 |
- use cleaner coroutine destruction. |
496 |
- simplify configuration for users. |
497 |
- optionally (unrecommended) prefer perl functions over |
498 |
their coro replacements. |
499 |
|
500 |
3.11 Tue Dec 5 13:11:24 CET 2006 |
501 |
- fixed some bogus assert's, but as perl.h disables assert even |
502 |
without NDEBUG (thank you very much), not too many people should |
503 |
notice (that did include myself). Andreas König noticed, though :) |
504 |
- do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have |
505 |
it. Also noticed by Andreas König :) |
506 |
- save/restore tainted status. |
507 |
- verified to pass the testsuite on my 5.9.5. |
508 |
|
509 |
3.1 Mon Dec 4 23:03:40 CET 2006 |
510 |
- INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow). |
511 |
- incompatible change: transfer flags are now per-state. |
512 |
- give a better error message on deadlock. |
513 |
- document Coro::nready. |
514 |
- enhanced testsuite. |
515 |
|
516 |
3.01 Sun Dec 3 23:47:42 CET 2006 |
517 |
- forgot to include Coro::Timer. |
518 |
|
519 |
3.0 Sun Dec 3 22:57:25 CET 2006 |
520 |
- the "FINALLY COMPLETELY STABLE" release (coming soon: |
521 |
the "FAMOUS LAST WORDS" release). |
522 |
- implement a new stack sharing algorithm, which uses a stack |
523 |
pool (size currently hardcoded). |
524 |
- make stack sharing mandatory (it no longer uses a heuristic). |
525 |
- eval/die no longer cause weird problems under heavy use. |
526 |
- Coro::Event could cause livelocks if it was used but |
527 |
no Coro::Event watchers were used. |
528 |
- Coro::Event now uses asynccheck as prepare does not |
529 |
check for changed watchers. |
530 |
- Coro::Event allows multiple waiting coros and will wake up one |
531 |
per event. |
532 |
- Coro::Event should be cleaner and more efficient now. |
533 |
- new utility function Coro::unblock_sub. |
534 |
- document the sad fact that Event is no longer reentrant. |
535 |
- putting a coroutine into the ready queue twice could under |
536 |
some circumstances lead to stack corruption. |
537 |
- minor incompatibility: subclassing Coro::State is supported |
538 |
directly now without going through a _coro_state member. |
539 |
- state/coro switching is much faster now. |
540 |
- very minor optimisations and code/documentation cleanup. |
541 |
- avoid problems due to compiler inlining. |
542 |
- removed timers from Coro::Timer -> use AnyEvent instead. |
543 |
- replaced Coro::idle coroutine by (cleaner) idle handler. |
544 |
- updated to newest libcoro. |
545 |
- implement enhanced support for valgrind. |
546 |
- implement is_ready and return value for ready. |
547 |
- removed Coro::Cont, it was a misnomer (it's generators), |
548 |
and it was somewhat annoying to get it right. Will come back |
549 |
if somebody asks for it :) |
550 |
- many ->wait methods and Coro::Event could return |
551 |
spuriously without the event having happened. |
552 |
|
553 |
2.5 Tue Nov 7 12:22:33 CET 2006 |
554 |
- made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket |
555 |
use AnyEvent, moved them to Coro/. |
556 |
- added Coro::LWP which contains all the uglyness required to |
557 |
make LWP non-blocking. |
558 |
- should work with perl 5.9.x now (Andreas König made me do it). |
559 |
- fixed another bug in Coro::Select when the timeout was undef. |
560 |
- reuse PL_start_env for all coros, saves some memory per coroutine. |
561 |
- manage PL_top_env differently, hopefully to avoid panic: top_env. |
562 |
- timeout argument was not properly used in Coro::Socket. |
563 |
- allow limited forms of subclassing in Coro::Handle/Coro::Socket. |
564 |
- emulate undocumented(!) functionality of IO::Socket required |
565 |
by LWP(!!). |
566 |
- updated eg/lwp to work with newer lwp's. |
567 |
- remove "FATAL: uncaught exception" prefix. Coroutines that die |
568 |
kill the whole process, just as exceptions in the main "coroutine" |
569 |
did already. |
570 |
|
571 |
2.1 Wed Nov 1 23:01:13 CET 2006 |
572 |
- fix a long-standing bug in Coro::Select where select with |
573 |
zero timeout would instead change the current default filehandle. |
574 |
- use a simpler and hopefully more robust way to clone padlists |
575 |
(uses less memory and a perl function instead of our own). |
576 |
- coro can now create a stack guard area on many architectures. |
577 |
- Coro::AIO properly reexports additional functions from IO::AIO. |
578 |
- updated libcoro with a workaround for OS X, |
579 |
pach and testing by Michael Schwern. |
580 |
|
581 |
2.0 Tue Oct 24 05:47:17 CEST 2006 |
582 |
- support additional aio requests in Coro::AIO. |
583 |
|
584 |
1.9 (never properly released due to a glitch) |
585 |
|
586 |
1.8 Thu Feb 2 00:59:06 CET 2006 |
587 |
- applied suggested patch by SAMV to avoid problems during stupid |
588 |
mark & sweep gc run. |
589 |
- applied patch by Scott Walters for 5.9.3 compatibility. |
590 |
|
591 |
1.7 Tue Dec 27 01:41:58 CET 2005 |
592 |
- added Coro::AIO, a thin wrapper around IO::AIO. |
593 |
- improved Makefile.PL explanations. |
594 |
|
595 |
1.6 Mon Dec 12 21:30:05 CET 2005 |
596 |
- additionally save PM_curpm might fix as-of-yet |
597 |
unseen problems with regex matches being attributed to the |
598 |
wrong package. |
599 |
- add t/10_bugs.t, which currently checks against imho broken |
600 |
perls that use 0.26MB of stack space per Perl_magic_get |
601 |
invocation (newer linuxes) as opposed to the 0.0002MB perl |
602 |
normally uses. |
603 |
- make stacksize configurable for the ultimate debian experience. |
604 |
|
605 |
1.51 Mon Dec 12 18:48:36 CET 2005 |
606 |
- remove debugging warn() accidentally left in Coro::Select. |
607 |
|
608 |
1.5 Tue Nov 29 13:32:44 CET 2005 |
609 |
- use Coro::Event inside Coro::Select to avoid spurious deadlocks. |
610 |
- fix Coro::Select 'select'. |
611 |
- strict'ify some modules. |
612 |
- libcoro errornously restored the SIGUSR2 handler to SIGUSR1. |
613 |
- use XSLoader in selected modules. |
614 |
- remove some 5.6 compatibility cruft. |
615 |
|
616 |
1.4 Tue Sep 6 00:11:05 CEST 2005 |
617 |
- libcoro did not take into account the trampoline on amd64, |
618 |
when 'l'inux method was used. |
619 |
|
620 |
1.31 Tue Aug 30 23:31:33 CEST 2005 |
621 |
- some portability fixes/workarounds. |
622 |
|
623 |
1.3 Sat Aug 20 03:08:56 CEST 2005 |
624 |
- no code changes. module seems to work fine. |
625 |
- improve Coro::State docs, remove reference to nonexisting |
626 |
Coro::State::flush method. |
627 |
- no longer autodetect windows, present it as an option instead. |
628 |
|
629 |
1.2 Mon May 16 02:00:55 CEST 2005 |
630 |
- included libcoro.c earlier in State.xs, might improve portability. |
631 |
- use a faster and possibly more stable Coro::Cont implementation. |
632 |
- accept x86_64 in addition to amd64 for optimized linux-amd64 support. |
633 |
- fix bugs Coro::SemaphoreSet that could cause locks never to be freed. |
634 |
- fix bugs in CoroAPI.h and document it in Coro::MakeMaker. |
635 |
|
636 |
1.11 Thu Mar 3 18:00:52 CET 2005 |
637 |
- change of contact address. |
638 |
|
639 |
1.1 Tue Feb 22 20:51:16 CET 2005 |
640 |
- support [l]inux method on amd64. |
641 |
- allow some context switches while compiling/eval'ing. |
642 |
|
643 |
1.0 Tue Aug 10 03:47:27 CEST 2004 |
644 |
- Event 0.89 fixes the bug again. |
645 |
- fix the remaining(??) memleak. I didn't free the cache-AV for |
646 |
padlists, so one AV per created anon-sub leaked. |
647 |
|
648 |
0.97 Fri May 14 15:23:32 CEST 2004 |
649 |
- remove stupid gcc extension that was an experiment only anyway. |
650 |
|
651 |
0.96 Thu May 13 18:09:29 CEST 2004 |
652 |
- tell people if they have yet another broken Event variant. |
653 |
- made Coro to work _better_ with threads (it seems that, as long |
654 |
as you use Coros in one thread only everything will work fine. |
655 |
Looking at the perl source the problem might be missing locking |
656 |
between Coro and some perl internal routines. That's not easily |
657 |
fixable). |
658 |
- (hopefully) got rid of the static xnext inside transfer. |
659 |
|
660 |
0.95 Sun Feb 15 03:20:28 CET 2004 |
661 |
- removed Coro::State::flush, implemented a new and much |
662 |
faster caching mechanism by attaching magic to the CV. |
663 |
- a memleak with (real) closures remains. |
664 |
- dramatically reduced size of stacks to 65536 (linux/x86), after |
665 |
observing that even myhttpd never had more than a kilobyte |
666 |
of stack in use (stack is still only physically allocated in |
667 |
pagesize increments for systems suporting mmap). |
668 |
- fix for a "cygwin-64" architecture added, proposed by Stefan Traby. |
669 |
- fix for threaded perl. |
670 |
|
671 |
0.9 Sun Nov 23 04:32:39 CET 2003 |
672 |
- fix a few very hard to track down but (of course) stupid |
673 |
memory leaks. |
674 |
- stringify version number to avoid locale problems :(. |
675 |
- perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters, |
676 |
hopefully ignoring it "just works". |
677 |
|
678 |
0.8 Wed Nov 5 19:38:40 CET 2003 |
679 |
- port to perl5.005_03, but only 5.8.x+ is supported! |
680 |
- honor LocalAddr even without LocalPort in Coro::Socket. |
681 |
|
682 |
0.7 Tue May 27 03:12:38 CEST 2003 |
683 |
- the version jump indicates some level of testing, not gobs |
684 |
of new features. |
685 |
- uh... I found the prompt function in ExtUtils::MakeMaker. |
686 |
Highly correct stuff, that is... |
687 |
- fixed(?) a bug with die's in coroutines causing "panic: top_env". |
688 |
the fix is not well understood by the author, so beware :(. |
689 |
|
690 |
0.652 Thu May 8 02:54:46 CEST 2003 |
691 |
- Applied patch by Slaven Rezic to set default to "s" on FreeBSD, |
692 |
cause version 4 doesn't have ucontext. |
693 |
- Benjamin Reed reported that setjmp works fine on darwin, |
694 |
so preselect it. |
695 |
|
696 |
0.651 Sat Mar 29 15:00:23 CET 2003 |
697 |
- fix a bug in Coro::Handle where some bytes could get lost |
698 |
on reading, reported by jason@nichego.net. |
699 |
|
700 |
0.65 Sun Mar 23 00:08:26 CET 2003 |
701 |
- added a README.linux-glibc. |
702 |
- new module Coro::Select. |
703 |
- also save/restore PL_comppad, fixes segfaults with 5.8.1. |
704 |
(I never claimed that I know what I am doing ;). |
705 |
- set default to setjmp/longjmp on non-x86-linux. |
706 |
|
707 |
0.6 Thu Nov 21 11:09:06 CET 2002 |
708 |
- made lazy context switching the default. |
709 |
- fixed the bug where SemaphoreSet::guard |
710 |
was the same as timed_guard and timed_guard was missing. |
711 |
- fix a memleak in Coro::Timer. |
712 |
- reclassified context sharing as not experimental. |
713 |
|
714 |
0.534 Sun Apr 14 03:05:12 CEST 2002 |
715 |
- fixed backspace => tab in header parsing (myhttpd). |
716 |
- added eg/readline, for elmex. |
717 |
- Coro::Event now calls ->start on first call instead of |
718 |
->again, which created wrong timeouts for the first ->next. |
719 |
- fixed a bug where Coro::Socket returned a closed socket instead |
720 |
of nothing on ECONNREFUSED and similar error conditions. |
721 |
|
722 |
0.533 Mon Feb 18 18:49:40 CET 2002 |
723 |
- adapt to finally working Event-0.85. |
724 |
- fixed buggy event initialization (trapped under windows, but how |
725 |
could it possibly have worked anywhere?) |
726 |
|
727 |
0.532 Wed Jan 16 02:45:32 CET 2002 |
728 |
- removed the reference to pp_entersub, might work on |
729 |
(some) win32 perls now (testsuite works under cygwin, |
730 |
without Event, which freezes). |
731 |
|
732 |
0.531 Mon Dec 10 22:18:44 CET 2001 |
733 |
- Fixed a bug in SemaphoreSet::timed_down. |
734 |
|
735 |
0.53 Tue Nov 27 21:11:13 CET 2001 |
736 |
- some tests for Event. |
737 |
- slightly different internal architecture to get refcnt into the object |
738 |
and not the reference. might fix a few bugs, certainly did introduce |
739 |
new ones. |
740 |
- Coro::Timer (independent of Event). |
741 |
- new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet. |
742 |
|
743 |
0.52 Tue Nov 6 21:36:18 CET 2001 |
744 |
- ported to cygwin (trivial). |
745 |
patch by Gerrit P. Haase <gp@familiehaase.de>. |
746 |
- small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov. |
747 |
|
748 |
0.51 Thu Nov 1 20:39:01 CET 2001 |
749 |
- terminate/cancel now work properly (otherwise termination |
750 |
could cause a "next coroutine is not and contains not..." |
751 |
error. |
752 |
- added Coro::Socket::shutdown. |
753 |
- Coro::Event::loop no is the same as Event::loop. |
754 |
- implemented terminate with args + join. |
755 |
|
756 |
0.5 Fri Sep 28 16:15:35 CEST 2001 |
757 |
- fixed "print" on a Coro::Handle. The print method worked. |
758 |
- small tweaks (seem to reduce memory consumption a lot) |
759 |
in various modules. |
760 |
- splendid use of "no warnings" scattered throughout. |
761 |
- added Coro::Handle::rbuf, fixed Coro::Handle::fh. |
762 |
|
763 |
0.49 Sun Sep 16 02:42:45 CEST 2001 |
764 |
- changed some method calls to function calls for speed |
765 |
inside Coro::Handle. |
766 |
- make Coro::Handle use an array instead of a hash for |
767 |
speed reasons. |
768 |
- IRIX mystery solved: it's SGI's NT, after all: "standard, huh?". |
769 |
sjlj and ucontext should now work. |
770 |
- IRIX-specific port for libcoro. |
771 |
- swapped order of accept results to match IO::Socket. |
772 |
- changed getsock/peername to sock/peername to match IO::Socket. |
773 |
- fixed a bug that caused segfault when returning to main task |
774 |
under some circumstances. |
775 |
- other bugfixes. |
776 |
|
777 |
0.45 Sun Sep 2 02:54:01 CEST 2001 |
778 |
- new method Coro::Handle::timeout. |
779 |
- corrected speling of set...name to get...name in Coro::Socket. |
780 |
- Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle. |
781 |
- Coro::Handle now supports fileno and FILENO. |
782 |
- added eg/myhttpd, a web server. |
783 |
- Coro::Socket now accepts numeric ports as well as "port(number)" |
784 |
syntax. |
785 |
- moved some scheduling primitives into xs code. more to come. |
786 |
- new simple priority system. |
787 |
- implemented Coro::Event using XS, almost four times faster! |
788 |
- small memory corruption problem fixed (boy that was difficult). |
789 |
|
790 |
0.13 Wed Aug 8 16:53:07 CEST 2001 |
791 |
- new method Coro::Handle::readline. |
792 |
- added eg/lwp and Coro::Handle::autoflush, to make LWP work. |
793 |
(see eg/lwp on how to make LWP non-blocking). |
794 |
- renamed Coro::Socket::new_inet to new. |
795 |
- added Coro::Util, some utility functions. |
796 |
- die/eval now works better (still get annoying (but true ;) |
797 |
"Callback called exit" messages). |
798 |
- Coro::Handle now supports timeout, and Coro::Socket Timeout. |
799 |
- much better Coro::Socket support. |
800 |
|
801 |
0.12 Fri Jul 27 04:19:01 CEST 2001 |
802 |
- do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined. |
803 |
- very experimental stack sharing algorithm. not 100% safe but |
804 |
should work well in practise ;) |
805 |
- added Coro::RWLock. |
806 |
- Coro::Specific now works. |
807 |
|
808 |
0.11 Tue Jul 24 22:49:21 CEST 2001 |
809 |
- added specialized hack for newer and older linux versions (fast). |
810 |
- renamed Coro::Event::IO to Coro::Handle. |
811 |
- new module Coro::Socket. |
812 |
|
813 |
0.1 Tue Jul 24 01:47:53 CEST 2001 |
814 |
- release candidate 3. A rather new internal structure :( |
815 |
- the great renaming: Damian Conway gave me a suitable replacement |
816 |
for yield. |
817 |
- added Coro/libcoro, a portable coroutine implementation for C |
818 |
(not even perl-dependent!!!), which can be used by Coro::State. |
819 |
- renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in |
820 |
perl-5.6. |
821 |
- new function Coro::Event::idle. |
822 |
- the idle process is now overriden |
823 |
by default in Coro::Event. |
824 |
- Coro::Channel now enforces the size. |
825 |
- canceling events now works in all cases. |
826 |
- Coro state now include $_ and $@. |
827 |
- yet another bug workaround that I do not really understand :( |
828 |
- new module Coro::Event::IO (very undocumented) to do |
829 |
non-blocking i/o. |
830 |
- performance tuning. |
831 |
|
832 |
0.08 Thu Jul 19 06:13:25 CEST 2001 |
833 |
- release candidate 2 (new functionality). |
834 |
- add Coro::State::flush function. |
835 |
- transfer now supports a flags value (mostly for speed ATM). |
836 |
- might compile and work in the presence of threads now. |
837 |
- continuations have a saner syntax. |
838 |
- no more memleaks. |
839 |
|
840 |
0.07 Tue Jul 17 17:40:18 CEST 2001 |
841 |
- release candidate 1 ;) |
842 |
- slightly nicer code. |
843 |
- fixed a scheduling bug in Coro::Event. |
844 |
|
845 |
0.06 Tue Jul 17 04:23:24 CEST 2001 |
846 |
- ok, I found the showstopper - the same sub must not be |
847 |
re-entered in two different coroutines, otherwise => crash. I |
848 |
see no easy solution to this problem, except by walking the call |
849 |
chaing and saving/restoring the cv's, which is what I do now. |
850 |
- memory leaks still latent, especially at thread termination. |
851 |
- Coro::Event now works (haha). |
852 |
|
853 |
0.05 Sun Jul 15 17:32:20 CEST 2001 |
854 |
- fixed a few issues in Coro::Event. |
855 |
- I forgot to include Coro::Event in 0.04 :(:(:( |
856 |
|
857 |
0.04 Sun Jul 15 05:24:59 CEST 2001 |
858 |
- @_ is now properly localized. |
859 |
- Coro::State is now easier subclassable. |
860 |
- Coro::Cont now coroutine-aware. |
861 |
- Coro::Specific is a low-overhead module to create |
862 |
coroutine-specific vars. |
863 |
- Coro::Event provides a simple interface to Event. |
864 |
|
865 |
0.03 Fri Jul 13 14:51:52 CEST 2001 |
866 |
- transfer() now implemented in XS (beware). |
867 |
- new module Coro::Cont for really faked continuations. |
868 |
- big internal architecture changes: Coro::State is now |
869 |
really low-level and can thus be used to implement other |
870 |
interesting things, While "Coro::" implements a process-like |
871 |
model. Still crude and subject to change. |
872 |
- $_ and $@ are no longer being localized. |
873 |
|
874 |
0.02 Tue Jul 10 01:38:17 CEST 2001 |
875 |
- implemented "async" attribute. |
876 |
- $_ and $@ are now localized. |
877 |
- added Coro::Channel. |
878 |
- more testcases, still no docs. |
879 |
|
880 |
0.01 Tue Jul 3 02:18:41 CEST 2001 |
881 |
- original version; copied from Convert::Scalar. |
882 |
|