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

Comparing Coro/Changes (file contents):
Revision 1.554 by root, Thu Mar 31 04:24:19 2011 UTC vs.
Revision 1.663 by root, Sat Oct 29 19:12:45 2016 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: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6TODO: swap_sv, maybe add scope_swap_sv?
7TODO: croak when async_pool tries to run canceled thread?
6 8
9 - re-try JIT allocation without PROT_EXEC, to hopefully improve
10 portability.
11
126.511 Sun Jun 26 23:44:50 CEST 2016
13 - make it compile with threaded perls again.
14 - simplify and speed up __DIE__ and __WARN__ handling.
15
166.51 Sat Jun 25 20:57:20 CEST 2016
17 - 6.5 release didn't compile on most older perls due to
18 a logic error enabling experimental 5.24 code for all versions.
19 - since various broken or deficient patches float around
20 to work around the vtbl API breakage in perl, let's provide
21 our own workaround, which seems to be much less invasive
22 and more compatible than the approaches seen so far. YMMV.
23 - work around assertion failure in perl_destruct on 5.24
24 (analyzed by Dave Mitchell).
25
266.5 Wed Jun 22 22:23:50 CEST 2016
27 - swap_sv swapping order was not symmetrical, causing
28 wrong swaps when swap_sv was used multiple times on the
29 same sv in the same thread.
30 - swap_sv calls can now be undone by calling it again
31 with the same variables.
32 - swap_sv calls will now be undone in async_pool threads.
33 - split Coro::Semaphore::up/adjust into separate xs functions
34 for better error reporting, at slight codesize increase.
35 - (libcoro) arm assembly support, please test and report.
36 - adjust to PL_savestack changes in perl 5.24 (adapted from
37 the debian patch, which unfortunately gets it wrong).
38
396.49 Sat Oct 17 01:40:12 CEST 2015
40 - throwing an exception to a thread waiting in
41 Coro::Handle using Coro::EV did not stop the watchers,
42 causing the next call to fail (testcase by Martin Pritchard).
43 - bump minimum perl version to 5.10.
44
456.48 Sun Oct 4 19:03:51 CEST 2015
46 - fix memory leak when Coro::Handle uses Coro::EV internally
47 (testcase by Sten Sten).
48 - update code to libev 4 API (internally, Coro still used the
49 version 3 API calls).
50 - remove unused hv_sig.
51
526.47 Sat Jul 11 03:58:20 CEST 2015
53 - swap_svs - NVs can be in the SV head beginning with 5.20.
54
556.46 Tue Jun 30 14:36:30 CEST 2015
56 - restore portability to perls with windows fork emulation
57 (patch by Petr Písař).
58
596.45 Tue Jun 30 01:40:08 CEST 2015
60 - restore portability to perl 5.16 and below.
61
626.44 Tue Jun 30 00:41:54 CEST 2015
63 - avoid segfaulting (or worse) when tracing a canceled thread.
64 - the Event module silently broke it's public hook API, causing
65 failures on perls compiled with -Duselongdouble. Requiring
66 the latest version of Event because I don't know which release
67 changed this (it's not mentioned in the ChangeLog).
68 - allow xs level enterleave hooks via CoroAPI.
69
706.43 Thu Jun 4 15:38:14 CEST 2015
71 - use stability canary.
72 - port to stableperl-5.22.0-1.001.
73 - update libecb.
74
756.42 Wed Feb 11 20:29:52 CET 2015
76 - Coro::SemaphoreSet->try did not actually work (analyzed by
77 SATO Kentaro).
78 - upgrade libecb to be C11 compliant.
79
806.41 Sat Sep 6 22:08:46 CEST 2014
81 - restore portability to perl 5.8.x.
82 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
83 up some memory, but fixes "use feature" and similar modules
84 when used inside a Coro.
85 - allow zero as argument to Coro::Channel to mean the same thing
86 as no argument. this works with older versions as well,
87 but wasn't legal until now.
88 - slightly better c header file detection.
89
906.39 Mon Jun 2 00:00:08 CEST 2014
91 - work around more incompatible changes in 5.20.
92
936.38 Sun Jun 1 21:54:23 CEST 2014
94 - check that perl slots actually have enough space to hold
95 interpreter variables.
96 - untested port to perl 5.19 (64 bit tmps indices) (reported
97 by Andreas König).
98 - croak when cancel is called without a thread context.
99
1006.37 Tue Mar 4 13:27:33 CET 2014
101 - *sigh*, removed leftover debugging code from debugging a
102 perl bug, of all things.
103
1046.36 Tue Mar 4 07:11:59 CET 2014
105 - semaphores would not clear the destroy hook when interrupted
106 by ->throw, causing segfaults or worse.
107 - ->throw on a thread waiting for a semaphore did not acquire
108 the semaphore, but also didn't wake up other waiters,
109 possibly causing a deadlock.
110 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
111 to report a full stacktrace, which should help find out
112 where the actual call is.
113 - "atomically" destroy data in slf_destroy, because it is
114 the right thing to do, just in case.
115 - disable ecb.h libm dependency, to compile on ancient systems
116 or under adverse conditions.
117
1186.33 Mon Nov 18 11:26:27 CET 2013
119 - do not crash when freeing padlists with holes (in 5.18).
120 - tentative SVt_BIND 5.19 port/fix.
121
1226.32 Tue Nov 5 15:35:35 CET 2013
123 - use a new algorithm to derive padlists for perl 5.18. The old
124 one could lead to 0-pointer accesses inside perl (reported
125 by Darin McBride).
126
1276.31 Thu May 9 07:39:48 CEST 2013
128 - Coro::AIO requests would crash if the thread was ready'd
129 while the request was ongoing.
130
1316.29 Wed May 8 02:55:18 CEST 2013
132 - when an on_destroy handler destructs the coro currently being
133 destructed a perl scalar could be accessed after being freed,
134 likely causing a crash.
135
1366.28 Wed Mar 6 06:58:02 CET 2013
137 - clean remnants of existing __DIE__ and __WARN__ handlers so
138 they lose their magic and will not cause segfaults later
139 (testcase by Andrey Sagulin).
140 - improved Coro::State documentation a bit.
141 - Coro::Debug::command now flushes the output.
142 - add hack detection code for x32 abi, because the braindead slugs
143 who designed that made it look exactly like x86_64 without
144 providing proper compile time symbols to test for it. as a result,
145 this detection cannot work reliably.
146 - valgrind stack registering was broken.
147 - do not rely on Time::HiRes anymore in Coro::Debug.
148
1496.23 Fri Dec 7 23:36:37 CET 2012
150 - use experimental fiber implementation on native windows
151 perls.
152 - use sizeof (void *) as multiplication factor for stack sizes,
153 to accomodate the totally braindamaged microsoft 64 bit "os".
154 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
155 activeperl 5.16 crashes when PerlIO_define_layer is called due
156 to some bug in the perl dll, strawberry perl at least passes
157 the testsuite.
158 - implement Coro::Handle->peeraddr/host/port, for slightly
159 improved compatibility with LWP.
160 - implement 5.17 compatibility by almost blindly applying a
161 good-looking patch by Father Chrysostomos.
162 - move stack management functions into libcoro 3.
163 - libcoro version 3 "released".
164 - support magic values as timed_io_once args.
165 - recommend AnyEvent 7+ or EV 4+, also require EV
166 version 4 or newer for Coro::EV.
167
1686.10 Tue Oct 9 01:14:27 CEST 2012
169 - updated ecb.h, it had a typo that caused it to not compile on many
170 big endian systems (reported by many people).
171 - disable memory fences in ecb.h to improve portability.
172
1736.09 Sat Oct 6 23:25:02 CEST 2012
174 - Coro::EV I/O watchers were not interruptible by exceptions
175 (Coro::State::throw) (testcase by sten).
176 - ->throw now puts threads into the ready queue, as this seems to
177 be expected by existing code, and code that doesn't cope with spurious
178 wakeups needs fixing anyway.
179 - use fd -1 in mmap.
180 - cast I32 to int in error message printf.
181 - warn about broken so-called "hardened" kernels.
182
1836.08 Fri Apr 13 12:05:47 CEST 2012
184 - be more aggressive about exiting like perl does - formerly,
185 exiting from the non-main thread would not execute END blocks.
186
1876.07 Fri Nov 11 21:21:48 CET 2011
188 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
189 - update ecb.h.
190
1916.06 Mon Aug 8 23:59:48 CEST 2011
192 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
193 backend, which is an order of magnitude slower. unfortunately, cygwins
194 pthread implementation isn't very complete either, so allocate the stack
195 twice just to be sure.
196 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
197 in its ithreaded perl - best recompile cygwin and use the 'w'indows
198 backend for much better performance. also disable ithreads for
199 even better performance...).
200
2016.05 Thu Aug 4 21:36:36 CEST 2011
202 - blush, condvar values would not be propagated from send to recv anymore
203 (reported by Chip Salzenberg).
204 - use exponential increase for the readline buffer length in
205 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
206
2076.04 Wed Aug 3 17:15:45 CEST 2011
208 - use even more efficient and more compatible condvars for
209 compatibility to AnyEvent 6.x :)
210 - more inconsequential ecb.h updates.
211
2126.03 Wed Aug 3 11:41:30 CEST 2011
213 - change how Coro patches AnyEvent condvars for compatibility to
214 AnyEvent 6.x.
215 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
216 hasn't been included.
217
2186.02 Wed Jul 13 04:35:19 CEST 2011
219 - "improve portability to Gentoo" - gentoo manages to put perl variables
220 in memory areas that are farther than 2gb apart, which the jit couldn't
221 handle and barfed. now it's just a bit slower on gentoo and similar
222 systems.
223
2246.01 Sun Jul 3 12:31:14 CEST 2011
225 - workarounds are good, but the test for whether pthreads are used
226 was not good. this one should be better.
227 - check differently whether gcc generates cfi instructions itself.
228
2296.0 Wed Jun 29 19:43:35 CEST 2011
230 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
231 destroyed and cleaned up automatically (e.g. async { schedule }).
232 - implement a JIT compiler for part of the thread switch code,
233 which gives a 50% speed improvement on threaded perls, and
234 about 4% on non-threaded perls (so threaded perls now finally
235 reach about half the speed of non-threaded perls).
236 - slightly modernise Coro::Intro, add section about rouse functions.
237 - avoid DEFSV and ERRSV, giving another 10% improvement
238 in thread switching.
239 - Coro::State->is_destroyed is now called is_zombie.
240 - implement a Coro->safe_cancel method that might fail, but
241 cancels in a "safer" way if it succeeds.
242 - add preliminary support for DEBUGGING perls.
243 - get rid of two hash-accesses when initialising a new Coro - this
244 speeds up coro creation by almost a factor of two.
245 - croak when a coro that is being cancelled tries to block
246 (e.g. while executing a guard block), instead of crashing or
247 deadlocking.
248 - use a more robust and also faster method to identify Coro::State
249 objects - speeds up everything a bit.
250 - implement Coro->cancel in XS for a 20% speed improvement, and to
251 be able to implement mutual cancellation.
252 - speed up context switches by a percent or two by more efficiently
253 allocating context stack entries.
254 - implement Coro->join and Coro->on_destroy in XS for a speedup and
255 a reduction in memory use.
256 - cancelling a coro while it itself is cancelling another coro is
257 now supported and working, instead of triggering an assertion.
7 - be a bit more crash-resistant when calling (buggy) on_destroy 258 - be a bit more crash-resistant when calling (buggy) on_destroy
8 callbacks (bets effort). 259 callbacks (best effort).
260 - move on_destroy into the slf_frame, to allow extension slf
261 functions to have destructors.
262 - get rid if coro refcounting - simply crash in other interpreter
263 threads by nulling the pointers on clone.
264 - simplify warn/die hook handling when loading Coro - the convoluted
265 logic seems to be no longer neccessary.
266 - use libecb instead of our own home-grown gcc hacks.
267 - document alternatives to Coro::LWP. Please use them :)
268 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
269 sigaltstack. Really. wine suffers from it, erlang suffers from it,
270 and it's known since at least 2006.
9 271
105.372 Wed Feb 23 06:14:30 CET 2011 2725.372 Wed Feb 23 06:14:30 CET 2011
11 - apparently mingw doesn't provide a working gettimeofday, try to 273 - apparently mingw doesn't provide a working gettimeofday, try to
12 work around that by relying on Time::HiRes (indirectly brought to 274 work around that by relying on Time::HiRes (indirectly brought to
13 my attention by Max Maischein). 275 my attention by Max Maischein).
… …
331 - ->throw is now supported on Coro::State objects. 593 - ->throw is now supported on Coro::State objects.
332 - clean up cctx creation code a bit. 594 - clean up cctx creation code a bit.
333 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 595 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
334 596
3354.9 Sat Nov 8 17:45:27 CET 2008 5974.9 Sat Nov 8 17:45:27 CET 2008
336 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 598 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
337 - (libcoro) no longer leak threads in the experimental pthread backend, 599 - (libcoro) no longer leak threads in the experimental pthread backend,
338 also speed it up considerably. 600 also speed it up considerably.
339 - (libcoro) do not rely on makecontext passing void *'s unscathed. 601 - (libcoro) do not rely on makecontext passing void *'s unscathed.
340 - fix compiletime dependencies on libcoro in the Makefile. 602 - fix compiletime dependencies on libcoro in the Makefile.
341 - cctx_count wasn't always updated properly. 603 - cctx_count wasn't always updated properly.
… …
403 - remove debugging code related to MgPV_nolen_const, also try to 665 - remove debugging code related to MgPV_nolen_const, also try to
404 make it compile with perl 5.8.6 (yes, apple apparently loves 666 make it compile with perl 5.8.6 (yes, apple apparently loves
405 outdated software). Reported by John S. 667 outdated software). Reported by John S.
406 668
4074.744 Tue Jul 8 22:06:35 CEST 2008 6694.744 Tue Jul 8 22:06:35 CEST 2008
408 - correctly provide default DIE/WARN handlers as documented. 670 - correctly provide default DIE/WARN handlers as documented.
409 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 671 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
410 implementations inside perl work fine for us. 672 implementations inside perl work fine for us.
411 673
4124.743 Mon Jun 16 00:21:57 CEST 2008 6744.743 Mon Jun 16 00:21:57 CEST 2008
413 - when using Coro::EV without running EV::loop it could 675 - when using Coro::EV without running EV::loop it could
… …
435 - sprinkle "no warnings" freely over everything, also suppress 697 - sprinkle "no warnings" freely over everything, also suppress
436 warnings for some other modules. 698 warnings for some other modules.
437 - fix typo in WSAEWOULDBLOCK. 699 - fix typo in WSAEWOULDBLOCK.
438 700
4394.72 Sun May 25 05:14:36 CEST 2008 7014.72 Sun May 25 05:14:36 CEST 2008
440 - tweak META.yaml a bit, unfortunately, there is no documented way 702 - tweak META.yaml a bit, unfortunately, there is no documented way
441 to have optional dependencies with CPAN. doh :( 703 to have optional dependencies with CPAN. doh :(
442 - avoid running some tests on windows because they would fail due to 704 - avoid running some tests on windows because they would fail due to
443 perl bug (broken fork, broken pipes...). 705 perl bug (broken fork, broken pipes...).
444 - work around perl on windows bugs where perl returns undocumented 706 - work around perl on windows bugs where perl returns undocumented
445 error codes for sysread, syswrite etc. by taking advantage 707 error codes for sysread, syswrite etc. by taking advantage
… …
449 - use unix domain sockets in testsuite to work around 711 - use unix domain sockets in testsuite to work around
450 common perl implementation bugs on widows (they are emulated by 712 common perl implementation bugs on widows (they are emulated by
451 tcp sockets on windows. ugh.) 713 tcp sockets on windows. ugh.)
452 714
4534.71 Sat May 24 20:01:27 CEST 2008 7154.71 Sat May 24 20:01:27 CEST 2008
454 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 716 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
455 - take advantage of async name resolution of AnyEvent::Util. 717 - take advantage of async name resolution of AnyEvent::Util.
456 - work around brutal inet_aton override in Coro::LWP. 718 - work around brutal inet_aton override in Coro::LWP.
457 - take advantage of the readyhook in Coro::EV, for smoother 719 - take advantage of the readyhook in Coro::EV, for smoother
458 scheduling. 720 scheduling.
459 721
4604.7 Sun May 11 00:32:19 CEST 2008 7224.7 Sun May 11 00:32:19 CEST 2008
461 - completely reworked the Coro manpage. 723 - completely reworked the Coro manpage.
462 - added Coro::AnyEvent, generic event loop integration. 724 - added Coro::AnyEvent, generic event loop integration.
463 - implement cancel, ready and kill commands in Coro::Debug. 725 - implement cancel, ready and kill commands in Coro::Debug.
464 - document find_coro in Coro::Debug. 726 - document find_coro in Coro::Debug.
465 - incompatible change: rename has_stack to has_cctx. 727 - incompatible change: rename has_stack to has_cctx.
466 - Coro::AIO and Coro::BDB no longer force event model detection, 728 - Coro::AIO and Coro::BDB no longer force event model detection,
467 use AnyEvent::AIO and AnyEvent::BDB. 729 use AnyEvent::AIO and AnyEvent::BDB.
… …
591 853
5924.11 Thu Oct 11 02:40:24 CEST 2007 8544.11 Thu Oct 11 02:40:24 CEST 2007
593 - port to threaded perls. 855 - port to threaded perls.
594 856
5954.1 Thu Oct 11 02:38:16 CEST 2007 8574.1 Thu Oct 11 02:38:16 CEST 2007
596 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 858 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
597 be local to each coro (see Coro::State). 859 be local to each coro (see Coro::State).
598 - incompatible change: for very deep reasons, cede and cede_notself 860 - incompatible change: for very deep reasons, cede and cede_notself
599 cannot return anything, so nothing will be returned. 861 cannot return anything, so nothing will be returned.
600 - possibly bring back 5.10 compatibility (untested). 862 - possibly bring back 5.10 compatibility (untested).
601 - work around stupid (and wrong) warning on 5.10 :(. 863 - work around stupid (and wrong) warning on 5.10 :(.
… …
673 but stateful semantics. 935 but stateful semantics.
674 - fixed a lot of typos in Coro.pm (patch submitted by David 936 - fixed a lot of typos in Coro.pm (patch submitted by David
675 Steinbrunner, which applied flawlessly). 937 Steinbrunner, which applied flawlessly).
676 938
6773.6 Sat Apr 14 17:13:31 CEST 2007 9393.6 Sat Apr 14 17:13:31 CEST 2007
678 - added some bugfixes to get eg/myhttpd working again. 940 - added some bugfixes to get eg/myhttpd working again.
679 - added Coro::Storable for often-cede'ing freeze/thaw. 941 - added Coro::Storable for often-cede'ing freeze/thaw.
680 - try to do a clean exit when a coroutine calls exit 942 - try to do a clean exit when a coroutine calls exit
681 (EXPERIMENTAL). 943 (EXPERIMENTAL).
682 - got rid of indirect call through _coro_init. 944 - got rid of indirect call through _coro_init.
683 - updated the partly antique examples in eg/ to 945 - updated the partly antique examples in eg/ to

Diff Legend

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