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

Comparing Coro/Changes (file contents):
Revision 1.459 by root, Sat Nov 22 15:48:43 2008 UTC vs.
Revision 1.604 by root, Fri Dec 7 14:21:09 2012 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)). 6TODO: __GCC_HAVE_DWARF2_CFI_ASM
7 7
8 TODO: Coro/Intro 8TODO: add peerhost/peerport from IO::Socket::INET etc.
9
10 - use experimental fiber implementation on native windows
11 perls.
12 - use sizeof (void *) as multiplication factor for stack sizes,
13 to accomodate the totally braindamaged microsoft 64 bit "os".
14 - implement 5.17 compatibility by almost blindly applying a
15 good-looking patch by Father Chrysostomos.
16 - move stack management functions into libcoro 3.
17 - support magic values as timed_io_once args.
18 - recommend AnyEvent 7+ or EV 4+, also require EV
19 version 4 or newer for Coro::EV.
20
216.10 Tue Oct 9 01:14:27 CEST 2012
22 - updated ecb.h, it had a typo that caused it to not compile on many
23 big endian systems (reported by many people).
24 - disable memory fences in ecb.h to improve portability.
25
266.09 Sat Oct 6 23:25:02 CEST 2012
27 - Coro::EV I/O watchers were not interruptible by exceptions
28 (Coro::State::throw) (testcase by sten).
29 - ->throw now puts threads into the ready queue, as this seems to
30 be expected by existing code, and code that doesn't cope with spurious
31 wakeups needs fixing anyway.
32 - use fd -1 in mmap.
33 - cast I32 to int in error message printf.
34 - warn about broken so-called "hardened" kernels.
35
366.08 Fri Apr 13 12:05:47 CEST 2012
37 - be more aggressive about exiting like perl does - formerly,
38 exiting from the non-main thread would not execute END blocks.
39
406.07 Fri Nov 11 21:21:48 CET 2011
41 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
42 - update ecb.h.
43
446.06 Mon Aug 8 23:59:48 CEST 2011
45 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
46 backend, which is an order of magnitude slower. unfortunately, cygwins
47 pthread implementation isn't very complete either, so allocate the stack
48 twice just to be sure.
49 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
50 in its ithreaded perl - best recompile cygwin and use the 'w'indows
51 backend for much better performance. also disable ithreads for
52 even better performance...).
53
546.05 Thu Aug 4 21:36:36 CEST 2011
55 - blush, condvar values would not be propagated from send to recv anymore
56 (reported by Chip Salzenberg).
57 - use exponential increase for the readline buffer length in
58 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
59
606.04 Wed Aug 3 17:15:45 CEST 2011
61 - use even more efficient and more compatible condvars for
62 compatibility to AnyEvent 6.x :)
63 - more inconsequential ecb.h updates.
64
656.03 Wed Aug 3 11:41:30 CEST 2011
66 - change how Coro patches AnyEvent condvars for compatibility to
67 AnyEvent 6.x.
68 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
69 hasn't been included.
70
716.02 Wed Jul 13 04:35:19 CEST 2011
72 - "improve portability to Gentoo" - gentoo manages to put perl variables
73 in memory areas that are farther than 2gb apart, which the jit couldn't
74 handle and barfed. now it's just a bit slower on gentoo and similar
75 systems.
76
776.01 Sun Jul 3 12:31:14 CEST 2011
78 - workarounds are good, but the test for whether pthreads are used
79 was not good. thisone should be better.
80 - check differently whether gcc generates cfi instructions itself.
81
826.0 Wed Jun 29 19:43:35 CEST 2011
83 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
84 destroyed and cleaned up automatically (e.g. async { schedule }).
85 - implement a JIT compiler for part of the thread switch code,
86 which gives a 50% speed improvement on threaded perls, and
87 about 4% on non-thraeded perls (so threaded perls now finally
88 reach about half the speed of non-threaded perls).
89 - slightly modernise Coro::Intro, add section about rouse functions.
90 - avoid DEFSV and ERRSV, giving another 10% improvement
91 in thread switching.
92 - Coro::State->is_destroyed is now called is_zombie.
93 - implement a Coro->safe_cancel method that might fail, but
94 cancels in a "safer" way if it succeeds.
95 - add preliminary support for DEBUGGING perls.
96 - get rid of two hash-accesses when initialising a new Coro - this
97 speeds up coro creation by almost a factor of two.
98 - croak when a coro that is being cancelled tries to block
99 (e.g. while executing a guard block), instead of crashing or
100 deadlocking.
101 - use a more robust and also faster method to identify Coro::State
102 objects - speeds up everything a bit.
103 - implement Coro->cancel in XS for a 20% speed improvement, and to
104 be able to implement mutual cancellation.
105 - speed up context switches by a percent or two by more efficiently
106 allocating context stack entries.
107 - implement Coro->join and Coro->on_destroy in XS for a speedup and
108 a reduction in memory use.
109 - cancelling a coro while it itself is cancelling another coro is
110 now supported and working, instead of triggering an assertion.
111 - be a bit more crash-resistant when calling (buggy) on_destroy
112 callbacks (best effort).
113 - move on_destroy into the slf_frame, to allow extension slf
114 functions to have destructors.
115 - get rid if coro refcounting - simply crash in other interpreter
116 threads by nulling the pointers on clone.
117 - simplify warn/die hook handling when loading Coro - the convoluted
118 logic seems to be no longer neccessary.
119 - use libecb instead of our own home-grown gcc hacks.
120 - document alternatives to Coro::LWP. Please use them :)
121 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
122 sigaltstack. Really. wine suffers from it, erlang suffers from it,
123 and it's known since at least 2006.
124
1255.372 Wed Feb 23 06:14:30 CET 2011
126 - apparently mingw doesn't provide a working gettimeofday, try to
127 work around that by relying on Time::HiRes (indirectly brought to
128 my attention by Max Maischein).
129 - fix some portability issues when Time::HiRes was used.
130
1315.371 Mon Feb 21 14:36:08 CET 2011
132 - backport to windows process emulation code again.
133
1345.37 Sat Feb 19 07:49:44 CET 2011
135 - add a big "Coro thread life cycle" section to "man Coro".
136 - try a tentative workaround against the breakage that 5.13 has
137 introduced without depreciation period. sigh.
138 - no longer use Time::HiRes if gettimeofday is available, which
139 saves quite a lot of memory.
140
1415.36 Sun Feb 13 05:33:41 CET 2011
142 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
143 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
144 method call.
145 - undocument Coro::Timer::sleep and obsolete whole module.
146 - optimise Coro::Timer::timeout memory and cpu usage.
147 - slightly updated Coro::Intro for recent changes.
148 - do not initialise PL_dirty anymore.
149
1505.25 Thu Nov 11 01:08:39 CET 2010
151 - try a different approach on netbsd - netbsd 5 finally has marginally
152 working pthreads, but still broken ucontext/sigaltstack.
153 - openbsd 4.8 finally got their act together, Coro works out of the box
154 with asm, setjmp and pthreads (no change, just informational).
155
1565.24 Sat Oct 23 11:27:12 CEST 2010
157 - port to the EV 4.0 API.
158 - work around bugs in mingw32, making strawberry perl work
159 out of the box.
160 - correctly modify Coro::AIO function prototypes
161 so that they reflect the "no optional parameters" rule.
162 - "ported" libcoro to C++.
163
1645.23 Mon May 17 18:50:42 CEST 2010
165 - be more resistant to ordering changes when initialising
166 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
167 Waldorf).
168 - document that perl 5.12 deliberately removed support for cloning.
169
1705.22 Wed Apr 14 03:55:35 CEST 2010
171 - correctly return udnef on errors in Coro::Handle::read/write
172 (testcase by Marc Mims).
173 - convert Coro::Util into a "perl compatibility wrapper" - the functions
174 are less useful now, but are drop-in replacements for existing
175 functions, listing better alternatives in the documentation. This also
176 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
177 with Coro::Util::inet_aton.
178 - do not override $Coro::idle unconditionally in Coro.pm, as other
179 modules could have provided their own idle coro already
180 (for exmaple, Coro::AnyEvent).
181 - fix Coro::Util::gethost* functions.
182 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
183
1845.21 Wed Dec 16 07:19:51 CET 2009
185 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
186 together.
187 - add some examples on how to combine other event loops with Coro in
188 Coro::AnyEvent, and how to run it (and not to block). Seems to be
189 the most common source of confusion.
190 - try to catch people naively blocking in an event callback.
191 - work around the perl filehandle bug issue in conjunction with
192 older common::sense (as indirectly pointed out by ZSystem).
193 - clarify the "not from signal handlers" section.
194
1955.2 Sun Oct 4 14:54:24 CEST 2009
196 - Coro::Storable destroyed the prototypes of the functions it wrapped.
197 - export rouse_cb and rouse_wait by default now.
198 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
199 - new method $state->swap_sv.
200 - added section on "windows process emulation" to the manpage,
201 after a not-so-fruitful (nor-friendly) "discussion" with chip
202 salzenberg (discussion implies arguments, but his only arguments
203 were ad-hominems, one wonders why he started it in the first
204 place). I hope this explains it well enough for him to understand,
205 and maybe well enough for others to understand.
206 - use common::sense everywhere now.
207 - idle callbacks are no longer supported, use idle coros instead.
208 - print a thread listing when a deadlock is detected.
209
2105.17 Sat Aug 22 23:09:31 CEST 2009
211 - work around a bug in the perl debugger causing crashes
212 when running under the debugger by marking _pool_handler as nodebug.
213 - speed up Coro::async considerably.
214 - try some hacks to get netbsd to work "more often" - their broken
215 setjmp/longjmp, ucontext *and* phtreads are really hard on Coro.
216 - convert Coro to AE 5.0 API.
217
2185.162 Tue Jul 28 04:04:03 CEST 2009
219 - perl 5.8.2 is now minimum requirement.
220 - skip t/19_handle.t on broken windows perls.
221
2225.161 Wed Jul 22 04:47:38 CEST 2009
223 - Coro::AnyEvent::poll could have a different prototype when EV was
224 used as backend (analyzed by Tatsuhiko Miyagawa).
225 - Coro::AnyEvent errornously initialised the event loop when loaded,
226 not on demand.
227 - try to workaround rare */t/01_unblock.t failures.
228
2295.16 Tue Jul 21 01:44:37 CEST 2009
230 - Coro::AnyEvent failed to hook into the event loop
231 when no threads had been readied between detecting
232 the event loop and actually running it.
233 - considerably speed up Coro::Select by taking avdantage
234 of AnyEvent > 4.8 and some other optimisations.
235 - implement paragraph readline mode in Coro::Handle
236 (based on patches by Zsbán Ambrus).
237 - replace WSAEINPROGRESS by WSAEWOULDBLOCK (reported
238 and analyzed by Yasuhiro MATSUMOTO).
239 - clarified libcoro license and copyright.
240 - someone stole my EXTRA_META!!!
241 - implement Coro::Select::patch_pp_sselect and it's brother,
242 for hardcode select overriding.
243
2445.151 Mon Jul 6 05:41:57 CEST 2009
245 - backport to windows process emulation code again (patch by
246 Yasuhiro MATSUMOTO).
247 - slightly update Coro::MakeMaker.
248
2495.15 Tue Jun 30 10:28:06 CEST 2009
250 - deprecate Coro::Socket, document how to get ipv6 support via
251 AnyEvent::Socket instead.
252 - implement signal->wait ($cb) interface, similar to semaphores.
253 - work around SvOK not supporting getmagic, so we have to getmagic
254 to test for undef :/ (reported by Matthias Waldorf).
255 - load Coro::AnyEvent in all modules using AnyEvent.
256 - work around perl corrupting our internal data structures,
257 reported by Tokuhiro Matsuno.
258 - enable per-coroutine real and cpu time gathering
259 (Coro::State::enable_times).
260
2615.14 Wed Jun 24 01:37:48 CEST 2009
262 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
263 as this seems to have been a difficult concept (poll, sleep, idle,
264 idle_upto).
265 - add Coro::AnyEvent::readable/writable functions.
266 - clarify Coro::EV/Event/AnyEvent manpages.
267 - free per-thread global scalars in the thread calling ->cancel, to
268 avoid crashes when $_, $@ etc., are magical but some of those
269 had already been freed.
270 - "unexperimentalise" the callback interface for Coro::Semaphore.
271 - speed up ready queue management/context switching by using a linked
272 list instead of an array (~5-10%).
273 - implement "watch" command in Coro::Debug shells.
274 - for fun, implement time-slicing as an example in the manpage.
275 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
276 Coro::Event, use the normal AnyEvent handling instead of dieing
277 (the same is true for Coro::Handle).
278 - properly document Coro::EV::timed_io_once.
279 - avoid unneccessary ->cancel calls in Coro::Handle.
280 - maybe make it work on mingw32 with win32 backend
281 (based on patch by Yasuhiro Matsumoto).
282
2835.132 Fri May 29 09:00:39 CEST 2009
284 - do not keep a reference to the argument itself in
285 Coro::Semaphore::guard, as it could change later.
286 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
287 should support a prepare callback.
288
2895.131 Mon Mar 16 23:20:37 CET 2009
290 - implement and document Coro->suspend, Coro->resume.
291 - fix Coro::Select implementation to not (often) close
292 the passed file descriptors (testcase provided by pippijn).
293
2945.13 Mon Dec 15 21:51:42 CET 2008
295 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
296 - don't set diehook to C<undef> but instead to NULL, to avoid
297 spurious warnings.
298 - fix a lot of bugs in Coro::SemaphoreSet.
299 - Coro::SemaphoreSet will less often create a semaphore needlessly.
300 - add Coro::SemaphoreSet::count and wait methods.
301 - take advantage of the new Guard module.
302 - deprecate Coro::guard.
303 - try to fix the dreaded 01_unblock tests once more. I hate it when
304 testsuites need more fixing than the code they are supposed to test.
305 - croak in more cases when a required callback isn't resolvable.
306 - fix some minor issues in Coro::State->call/eval.
307 - use current coroutine context instead of a temporary one
308 when temporarily switching to another coroutine.
309 - do not call C-level on_destroy handlers during global destruction,
310 to avoid needless segfaults.
311
3125.12 Sun Dec 7 13:30:38 CET 2008
313 - add default config for MirOS, which seems to be a bug-to-bug
314 compatible fork of openbsd ("world domination by releasing
315 openbsd cvs before the openbsd folks do it" or so :).
316 - free_padlist did destroy the names pad, not good, but didn't
317 seem to bother perl - this could fix issues such as eval ""
318 inside a function called from multiple coroutines.
319 - use a different method to detect destruction time.
320 - be more careful when freeing padlists just before global
321 destruction.
322 - fixed and expanded the call/cc example.
323 - renamed _terminate to _coro_run.
324 - new method Coro::Channel->shutdown.
325 - try pthreads on openbsd <4.4 (broken sigaltstack, will
326 pthreads fare better?).
327 - be less picky about destroying "a" running coroutine.
328
3295.11 Tue Nov 25 21:49:05 CET 2008
330 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
331 working code with failed assertions, introducing more bugs than
332 it fixes, requiring elaborate workarounds :(
333
3345.1 Mon Nov 24 08:54:59 CET 2008
335 - wrote a small introductory tutorial - Coro::Intro.
9 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API. 336 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
10 - Coro::Select did errornously dup the file descriptors 337 - Coro::Select did errornously dup the file descriptors
11 and didn't work with all AnyEvent backends. 338 and didn't work with all AnyEvent backends.
339 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
340 LWP data transfers.
12 - disassociate c contexts from coro objects - this is agruably more 341 - disassociate c contexts from coro objects - this is agruably more
13 correct, but mostly allows sharing of cctxs between coro and state 342 correct, but mostly allows sharing of cctxs between coro and state
14 objects, for added memory savings and speed increases. 343 objects, for added memory savings and speed increases.
15 - bumped $Coro::POOL_RSS up to 32kb by default. 344 - bumped $Coro::POOL_RSS up to 32kb by default.
16 - no longer set the optype to OP_CUSTOM, as B::* understandably 345 - no longer set the optype to OP_CUSTOM, as B::* understandably
17 doesn't like this very much (and we *are* a type of entersub). 346 doesn't like this very much (and we *are* a type of entersub).
18 - implement state cloning, just to prove that call/cc can be done. 347 - implement state cloning, just to prove that call/cc can be done.
348 - automatically load Coro::AnyEvent in Coro::Handle.
349 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
350 do not have this method (and don't need it either).
351 - speed up generic anyevent methods in Coro::Handle by using rouse
352 callbacks.
353 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
354 others. It also makes the debugger happier, as you can trace
355 through the idle threads now.
356 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
357 - eg/event was pretty broken.
19 - better 5.8.6 compatibility. 358 - better 5.8.6 compatibility.
20 359
215.0 Thu Nov 20 10:35:05 CET 2008 3605.0 Thu Nov 20 10:35:05 CET 2008
22 - NEW ARCHITECTURE: use the latest 4.x version if you experience 361 - NEW ARCHITECTURE: use the latest 4.x version if you experience
23 stability issues. 362 stability issues.

Diff Legend

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