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

Comparing Coro/Changes (file contents):
Revision 1.521 by root, Sat Aug 22 22:36:23 2009 UTC vs.
Revision 1.639 by root, Thu Jun 4 22:58:28 2015 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
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7TODO: unready_all
8TODO: myhttpd header parsing
9TODO: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6TODO: swap_sv, maybe add scope_swap_sv, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
8TODO: croak when async_pool tries to run canceled thread?
9
10 - port to stableperl-5.22.0-1.001.
11
126.43 Thu Jun 4 15:38:14 CEST 2015
13 - use stability canary.
14 - update libecb.
15
166.42 Wed Feb 11 20:29:52 CET 2015
17 - Coro::SemaphoreSet->try did not actually work (analyzed by
18 SATO Kentaro).
19 - upgrade libecb to be C11 compliant.
20
216.41 Sat Sep 6 22:08:46 CEST 2014
22 - restore portability to perl 5.8.x.
23 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
24 up some memory, but fixes "use feature" and similar modules
25 when used inside a Coro.
26 - allow zero as argument to Coro::Channel to mean the same thing
27 as no argument. this works with older versions as well,
28 but wasn't legal until now.
29 - slightly better c header file detection.
30
316.39 Mon Jun 2 00:00:08 CEST 2014
32 - work around more incompatible changes in 5.20.
33
346.38 Sun Jun 1 21:54:23 CEST 2014
35 - check that perl slots actually have enough space to hold
36 interpreter variables.
37 - untested port to perl 5.19 (64 bit tmps indices) (reported
38 by Andreas König).
39 - croak when cancel is called without a thread context.
40
416.37 Tue Mar 4 13:27:33 CET 2014
42 - *sigh*, removed leftover debugging code from debugging a
43 perl bug, of all things.
44
456.36 Tue Mar 4 07:11:59 CET 2014
46 - semaphores would not clear the destroy hook when interrupted
47 by ->throw, causing segfaults or worse.
48 - ->throw on a thread waiting for a semaphore did not acquire
49 the semaphore, but also didn't wake up other waiters,
50 possibly causing a deadlock.
51 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
52 to report a full stacktrace, which should help find out
53 where the actual call is.
54 - "atomically" destroy data in slf_destroy, because it is
55 the right thing to do, just in case.
56 - disable ecb.h libm dependency, to compile on ancient systems
57 or under adverse conditions.
58
596.33 Mon Nov 18 11:26:27 CET 2013
60 - do not crash when freeing padlists with holes (in 5.18).
61 - tentative SVt_BIND 5.19 port/fix.
62
636.32 Tue Nov 5 15:35:35 CET 2013
64 - use a new algorithm to derive padlists for perl 5.18. The old
65 one could lead to 0-pointer accesses inside perl (reported
66 by Darin McBride).
67
686.31 Thu May 9 07:39:48 CEST 2013
69 - Coro::AIO requests would crash if the thread was ready'd
70 while the request was ongoing.
71
726.29 Wed May 8 02:55:18 CEST 2013
73 - when an on_destroy handler destructs the coro currently being
74 destructed a perl scalar could be accessed after being freed,
75 likely causing a crash.
76
776.28 Wed Mar 6 06:58:02 CET 2013
78 - clean remnants of existing __DIE__ and __WARN__ handlers so
79 they lose their magic and will not cause segfaults later
80 (testcase by Andrey Sagulin).
81 - improved Coro::State documentation a bit.
82 - Coro::Debug::command now flushes the output.
83 - add hack detection code for x32 abi, because the braindead slugs
84 who designed that made it look exactly like x86_64 without
85 providing proper compile time symbols to test for it. as a result,
86 this detection cannot work reliably.
87 - valgrind stack registering was broken.
88 - do not rely on Time::HiRes anymore in Coro::Debug.
89
906.23 Fri Dec 7 23:36:37 CET 2012
91 - use experimental fiber implementation on native windows
92 perls.
93 - use sizeof (void *) as multiplication factor for stack sizes,
94 to accomodate the totally braindamaged microsoft 64 bit "os".
95 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
96 activeperl 5.16 crashes when PerlIO_define_layer is called due
97 to some bug in the perl dll, strawberry perl at least passes
98 the testsuite.
99 - implement Coro::Handle->peeraddr/host/port, for slightly
100 improved compatibility with LWP.
101 - implement 5.17 compatibility by almost blindly applying a
102 good-looking patch by Father Chrysostomos.
103 - move stack management functions into libcoro 3.
104 - libcoro version 3 "released".
105 - support magic values as timed_io_once args.
106 - recommend AnyEvent 7+ or EV 4+, also require EV
107 version 4 or newer for Coro::EV.
108
1096.10 Tue Oct 9 01:14:27 CEST 2012
110 - updated ecb.h, it had a typo that caused it to not compile on many
111 big endian systems (reported by many people).
112 - disable memory fences in ecb.h to improve portability.
113
1146.09 Sat Oct 6 23:25:02 CEST 2012
115 - Coro::EV I/O watchers were not interruptible by exceptions
116 (Coro::State::throw) (testcase by sten).
117 - ->throw now puts threads into the ready queue, as this seems to
118 be expected by existing code, and code that doesn't cope with spurious
119 wakeups needs fixing anyway.
120 - use fd -1 in mmap.
121 - cast I32 to int in error message printf.
122 - warn about broken so-called "hardened" kernels.
123
1246.08 Fri Apr 13 12:05:47 CEST 2012
125 - be more aggressive about exiting like perl does - formerly,
126 exiting from the non-main thread would not execute END blocks.
127
1286.07 Fri Nov 11 21:21:48 CET 2011
129 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
130 - update ecb.h.
131
1326.06 Mon Aug 8 23:59:48 CEST 2011
133 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
134 backend, which is an order of magnitude slower. unfortunately, cygwins
135 pthread implementation isn't very complete either, so allocate the stack
136 twice just to be sure.
137 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
138 in its ithreaded perl - best recompile cygwin and use the 'w'indows
139 backend for much better performance. also disable ithreads for
140 even better performance...).
141
1426.05 Thu Aug 4 21:36:36 CEST 2011
143 - blush, condvar values would not be propagated from send to recv anymore
144 (reported by Chip Salzenberg).
145 - use exponential increase for the readline buffer length in
146 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
147
1486.04 Wed Aug 3 17:15:45 CEST 2011
149 - use even more efficient and more compatible condvars for
150 compatibility to AnyEvent 6.x :)
151 - more inconsequential ecb.h updates.
152
1536.03 Wed Aug 3 11:41:30 CEST 2011
154 - change how Coro patches AnyEvent condvars for compatibility to
155 AnyEvent 6.x.
156 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
157 hasn't been included.
158
1596.02 Wed Jul 13 04:35:19 CEST 2011
160 - "improve portability to Gentoo" - gentoo manages to put perl variables
161 in memory areas that are farther than 2gb apart, which the jit couldn't
162 handle and barfed. now it's just a bit slower on gentoo and similar
163 systems.
164
1656.01 Sun Jul 3 12:31:14 CEST 2011
166 - workarounds are good, but the test for whether pthreads are used
167 was not good. this one should be better.
168 - check differently whether gcc generates cfi instructions itself.
169
1706.0 Wed Jun 29 19:43:35 CEST 2011
171 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
172 destroyed and cleaned up automatically (e.g. async { schedule }).
173 - implement a JIT compiler for part of the thread switch code,
174 which gives a 50% speed improvement on threaded perls, and
175 about 4% on non-threaded perls (so threaded perls now finally
176 reach about half the speed of non-threaded perls).
177 - slightly modernise Coro::Intro, add section about rouse functions.
178 - avoid DEFSV and ERRSV, giving another 10% improvement
179 in thread switching.
180 - Coro::State->is_destroyed is now called is_zombie.
181 - implement a Coro->safe_cancel method that might fail, but
182 cancels in a "safer" way if it succeeds.
183 - add preliminary support for DEBUGGING perls.
184 - get rid of two hash-accesses when initialising a new Coro - this
185 speeds up coro creation by almost a factor of two.
186 - croak when a coro that is being cancelled tries to block
187 (e.g. while executing a guard block), instead of crashing or
188 deadlocking.
189 - use a more robust and also faster method to identify Coro::State
190 objects - speeds up everything a bit.
191 - implement Coro->cancel in XS for a 20% speed improvement, and to
192 be able to implement mutual cancellation.
193 - speed up context switches by a percent or two by more efficiently
194 allocating context stack entries.
195 - implement Coro->join and Coro->on_destroy in XS for a speedup and
196 a reduction in memory use.
197 - cancelling a coro while it itself is cancelling another coro is
198 now supported and working, instead of triggering an assertion.
199 - be a bit more crash-resistant when calling (buggy) on_destroy
200 callbacks (best effort).
201 - move on_destroy into the slf_frame, to allow extension slf
202 functions to have destructors.
203 - get rid if coro refcounting - simply crash in other interpreter
204 threads by nulling the pointers on clone.
205 - simplify warn/die hook handling when loading Coro - the convoluted
206 logic seems to be no longer neccessary.
207 - use libecb instead of our own home-grown gcc hacks.
208 - document alternatives to Coro::LWP. Please use them :)
209 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
210 sigaltstack. Really. wine suffers from it, erlang suffers from it,
211 and it's known since at least 2006.
212
2135.372 Wed Feb 23 06:14:30 CET 2011
214 - apparently mingw doesn't provide a working gettimeofday, try to
215 work around that by relying on Time::HiRes (indirectly brought to
216 my attention by Max Maischein).
217 - fix some portability issues when Time::HiRes was used.
218
2195.371 Mon Feb 21 14:36:08 CET 2011
220 - backport to windows process emulation code again.
221
2225.37 Sat Feb 19 07:49:44 CET 2011
223 - add a big "Coro thread life cycle" section to "man Coro".
224 - try a tentative workaround against the breakage that 5.13 has
225 introduced without depreciation period. sigh.
226 - no longer use Time::HiRes if gettimeofday is available, which
227 saves quite a lot of memory.
228
2295.36 Sun Feb 13 05:33:41 CET 2011
230 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
231 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
232 method call.
233 - undocument Coro::Timer::sleep and obsolete whole module.
234 - optimise Coro::Timer::timeout memory and cpu usage.
235 - slightly updated Coro::Intro for recent changes.
236 - do not initialise PL_dirty anymore.
237
2385.25 Thu Nov 11 01:08:39 CET 2010
239 - try a different approach on netbsd - netbsd 5 finally has marginally
240 working pthreads, but still broken ucontext/sigaltstack.
241 - openbsd 4.8 finally got their act together, Coro works out of the box
242 with asm, setjmp and pthreads (no change, just informational).
243
2445.24 Sat Oct 23 11:27:12 CEST 2010
245 - port to the EV 4.0 API.
246 - work around bugs in mingw32, making strawberry perl work
247 out of the box.
248 - correctly modify Coro::AIO function prototypes
249 so that they reflect the "no optional parameters" rule.
250 - "ported" libcoro to C++.
251
2525.23 Mon May 17 18:50:42 CEST 2010
253 - be more resistant to ordering changes when initialising
254 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
255 Waldorf).
256 - document that perl 5.12 deliberately removed support for cloning.
257
2585.22 Wed Apr 14 03:55:35 CEST 2010
259 - correctly return udnef on errors in Coro::Handle::read/write
260 (testcase by Marc Mims).
261 - convert Coro::Util into a "perl compatibility wrapper" - the functions
262 are less useful now, but are drop-in replacements for existing
263 functions, listing better alternatives in the documentation. This also
264 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
265 with Coro::Util::inet_aton.
266 - do not override $Coro::idle unconditionally in Coro.pm, as other
267 modules could have provided their own idle coro already
268 (for exmaple, Coro::AnyEvent).
269 - fix Coro::Util::gethost* functions.
270 - Coro::Timer corretcly exports it's symbols (reported by Hideki Yamamura).
271
2725.21 Wed Dec 16 07:19:51 CET 2009
273 - automatically load Coro::AnyEvent when AnyEvent and Coro are used
274 together.
275 - add some examples on how to combine other event loops with Coro in
276 Coro::AnyEvent, and how to run it (and not to block). Seems to be
277 the most common source of confusion.
278 - try to catch people naively blocking in an event callback.
279 - work around the perl filehandle bug issue in conjunction with
280 older common::sense (as indirectly pointed out by ZSystem).
281 - clarify the "not from signal handlers" section.
282
2835.2 Sun Oct 4 14:54:24 CEST 2009
284 - Coro::Storable destroyed the prototypes of the functions it wrapped.
285 - export rouse_cb and rouse_wait by default now.
286 - fix various prototype mismatches in Coro::AnyEvent and Coro::Handle.
287 - new method $state->swap_sv.
288 - added section on "windows process emulation" to the manpage,
289 after a not-so-fruitful (nor-friendly) "discussion" with chip
290 salzenberg (discussion implies arguments, but his only arguments
291 were ad-hominems, one wonders why he started it in the first
292 place). I hope this explains it well enough for him to understand,
293 and maybe well enough for others to understand.
294 - use common::sense everywhere now.
295 - idle callbacks are no longer supported, use idle coros instead.
296 - print a thread listing when a deadlock is detected.
10 297
115.17 Sat Aug 22 23:09:31 CEST 2009 2985.17 Sat Aug 22 23:09:31 CEST 2009
12 - work around a bug in the perl debugger causing crashes 299 - work around a bug in the perl debugger causing crashes
13 when running under the debugger by marking _pool_handler as nodebug. 300 when running under the debugger by marking _pool_handler as nodebug.
14 - speed up Coro::async considerably. 301 - speed up Coro::async considerably.
… …
247 - ->throw is now supported on Coro::State objects. 534 - ->throw is now supported on Coro::State objects.
248 - clean up cctx creation code a bit. 535 - clean up cctx creation code a bit.
249 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 536 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
250 537
2514.9 Sat Nov 8 17:45:27 CET 2008 5384.9 Sat Nov 8 17:45:27 CET 2008
252 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 539 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
253 - (libcoro) no longer leak threads in the experimental pthread backend, 540 - (libcoro) no longer leak threads in the experimental pthread backend,
254 also speed it up considerably. 541 also speed it up considerably.
255 - (libcoro) do not rely on makecontext passing void *'s unscathed. 542 - (libcoro) do not rely on makecontext passing void *'s unscathed.
256 - fix compiletime dependencies on libcoro in the Makefile. 543 - fix compiletime dependencies on libcoro in the Makefile.
257 - cctx_count wasn't always updated properly. 544 - cctx_count wasn't always updated properly.
… …
319 - remove debugging code related to MgPV_nolen_const, also try to 606 - remove debugging code related to MgPV_nolen_const, also try to
320 make it compile with perl 5.8.6 (yes, apple apparently loves 607 make it compile with perl 5.8.6 (yes, apple apparently loves
321 outdated software). Reported by John S. 608 outdated software). Reported by John S.
322 609
3234.744 Tue Jul 8 22:06:35 CEST 2008 6104.744 Tue Jul 8 22:06:35 CEST 2008
324 - correctly provide default DIE/WARN handlers as documented. 611 - correctly provide default DIE/WARN handlers as documented.
325 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 612 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
326 implementations inside perl work fine for us. 613 implementations inside perl work fine for us.
327 614
3284.743 Mon Jun 16 00:21:57 CEST 2008 6154.743 Mon Jun 16 00:21:57 CEST 2008
329 - when using Coro::EV without running EV::loop it could 616 - when using Coro::EV without running EV::loop it could
… …
351 - sprinkle "no warnings" freely over everything, also suppress 638 - sprinkle "no warnings" freely over everything, also suppress
352 warnings for some other modules. 639 warnings for some other modules.
353 - fix typo in WSAEWOULDBLOCK. 640 - fix typo in WSAEWOULDBLOCK.
354 641
3554.72 Sun May 25 05:14:36 CEST 2008 6424.72 Sun May 25 05:14:36 CEST 2008
356 - tweak META.yaml a bit, unfortunately, there is no documented way 643 - tweak META.yaml a bit, unfortunately, there is no documented way
357 to have optional dependencies with CPAN. doh :( 644 to have optional dependencies with CPAN. doh :(
358 - avoid running some tests on windows because they would fail due to 645 - avoid running some tests on windows because they would fail due to
359 perl bug (broken fork, broken pipes...). 646 perl bug (broken fork, broken pipes...).
360 - work around perl on windows bugs where perl returns undocumented 647 - work around perl on windows bugs where perl returns undocumented
361 error codes for sysread, syswrite etc. by taking advantage 648 error codes for sysread, syswrite etc. by taking advantage
… …
365 - use unix domain sockets in testsuite to work around 652 - use unix domain sockets in testsuite to work around
366 common perl implementation bugs on widows (they are emulated by 653 common perl implementation bugs on widows (they are emulated by
367 tcp sockets on windows. ugh.) 654 tcp sockets on windows. ugh.)
368 655
3694.71 Sat May 24 20:01:27 CEST 2008 6564.71 Sat May 24 20:01:27 CEST 2008
370 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 657 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
371 - take advantage of async name resolution of AnyEvent::Util. 658 - take advantage of async name resolution of AnyEvent::Util.
372 - work around brutal inet_aton override in Coro::LWP. 659 - work around brutal inet_aton override in Coro::LWP.
373 - take advantage of the readyhook in Coro::EV, for smoother 660 - take advantage of the readyhook in Coro::EV, for smoother
374 scheduling. 661 scheduling.
375 662
3764.7 Sun May 11 00:32:19 CEST 2008 6634.7 Sun May 11 00:32:19 CEST 2008
377 - completely reworked the Coro manpage. 664 - completely reworked the Coro manpage.
378 - added Coro::AnyEvent, generic event loop integration. 665 - added Coro::AnyEvent, generic event loop integration.
379 - implement cancel, ready and kill commands in Coro::Debug. 666 - implement cancel, ready and kill commands in Coro::Debug.
380 - document find_coro in Coro::Debug. 667 - document find_coro in Coro::Debug.
381 - incompatible change: rename has_stack to has_cctx. 668 - incompatible change: rename has_stack to has_cctx.
382 - Coro::AIO and Coro::BDB no longer force event model detection, 669 - Coro::AIO and Coro::BDB no longer force event model detection,
383 use AnyEvent::AIO and AnyEvent::BDB. 670 use AnyEvent::AIO and AnyEvent::BDB.
… …
507 794
5084.11 Thu Oct 11 02:40:24 CEST 2007 7954.11 Thu Oct 11 02:40:24 CEST 2007
509 - port to threaded perls. 796 - port to threaded perls.
510 797
5114.1 Thu Oct 11 02:38:16 CEST 2007 7984.1 Thu Oct 11 02:38:16 CEST 2007
512 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 799 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
513 be local to each coro (see Coro::State). 800 be local to each coro (see Coro::State).
514 - incompatible change: for very deep reasons, cede and cede_notself 801 - incompatible change: for very deep reasons, cede and cede_notself
515 cannot return anything, so nothing will be returned. 802 cannot return anything, so nothing will be returned.
516 - possibly bring back 5.10 compatibility (untested). 803 - possibly bring back 5.10 compatibility (untested).
517 - work around stupid (and wrong) warning on 5.10 :(. 804 - work around stupid (and wrong) warning on 5.10 :(.
… …
589 but stateful semantics. 876 but stateful semantics.
590 - fixed a lot of typos in Coro.pm (patch submitted by David 877 - fixed a lot of typos in Coro.pm (patch submitted by David
591 Steinbrunner, which applied flawlessly). 878 Steinbrunner, which applied flawlessly).
592 879
5933.6 Sat Apr 14 17:13:31 CEST 2007 8803.6 Sat Apr 14 17:13:31 CEST 2007
594 - added some bugfixes to get eg/myhttpd working again. 881 - added some bugfixes to get eg/myhttpd working again.
595 - added Coro::Storable for often-cede'ing freeze/thaw. 882 - added Coro::Storable for often-cede'ing freeze/thaw.
596 - try to do a clean exit when a coroutine calls exit 883 - try to do a clean exit when a coroutine calls exit
597 (EXPERIMENTAL). 884 (EXPERIMENTAL).
598 - got rid of indirect call through _coro_init. 885 - got rid of indirect call through _coro_init.
599 - updated the partly antique examples in eg/ to 886 - updated the partly antique examples in eg/ to

Diff Legend

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