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

Comparing Coro/Changes (file contents):
Revision 1.580 by root, Wed Aug 3 09:43:03 2011 UTC vs.
Revision 1.658 by root, Wed Jun 22 20:37:27 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?
8
9 - work around assertion failure in perl_destruct on 5.24
10 (analyzed by Dave Mitchell).
11
126.5 Wed Jun 22 22:23:50 CEST 2016
13 - swap_sv swapping order was not symmetrical, causing
14 wrong swaps when swap_sv was used multiple times on the
15 same sv in the same thread.
16 - swap_sv calls can now be undone by calling it again
17 with the same variables.
18 - swap_sv calls will now be undone in async_pool threads.
19 - split Coro::Semaphore::up/adjust into separate xs functions
20 for better error reporting, at slight codesize increase.
21 - (libcoro) arm assembly support, please test and report.
22 - adjust to PL_savestack changes in perl 5.24 (adapted from
23 the debian patch, which unfortunately gets it wrong).
24
256.49 Sat Oct 17 01:40:12 CEST 2015
26 - throwing an exception to a thread waiting in
27 Coro::Handle using Coro::EV did not stop the watchers,
28 causing the next call to fail (testcase by Martin Pritchard).
29 - bump minimum perl version to 5.10.
30
316.48 Sun Oct 4 19:03:51 CEST 2015
32 - fix memory leak when Coro::Handle uses Coro::EV internally
33 (testcase by Sten Sten).
34 - update code to libev 4 API (internally, Coro still used the
35 version 3 API calls).
36 - remove unused hv_sig.
37
386.47 Sat Jul 11 03:58:20 CEST 2015
39 - swap_svs - NVs can be in the SV head beginning with 5.20.
40
416.46 Tue Jun 30 14:36:30 CEST 2015
42 - restore portability to perls with windows fork emulation
43 (patch by Petr Písař).
44
456.45 Tue Jun 30 01:40:08 CEST 2015
46 - restore portability to perl 5.16 and below.
47
486.44 Tue Jun 30 00:41:54 CEST 2015
49 - avoid segfaulting (or worse) when tracing a canceled thread.
50 - the Event module silently broke it's public hook API, causing
51 failures on perls compiled with -Duselongdouble. Requiring
52 the latest version of Event because I don't know which release
53 changed this (it's not mentioned in the ChangeLog).
54 - allow xs level enterleave hooks via CoroAPI.
55
566.43 Thu Jun 4 15:38:14 CEST 2015
57 - use stability canary.
58 - port to stableperl-5.22.0-1.001.
59 - update libecb.
60
616.42 Wed Feb 11 20:29:52 CET 2015
62 - Coro::SemaphoreSet->try did not actually work (analyzed by
63 SATO Kentaro).
64 - upgrade libecb to be C11 compliant.
65
666.41 Sat Sep 6 22:08:46 CEST 2014
67 - restore portability to perl 5.8.x.
68 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
69 up some memory, but fixes "use feature" and similar modules
70 when used inside a Coro.
71 - allow zero as argument to Coro::Channel to mean the same thing
72 as no argument. this works with older versions as well,
73 but wasn't legal until now.
74 - slightly better c header file detection.
75
766.39 Mon Jun 2 00:00:08 CEST 2014
77 - work around more incompatible changes in 5.20.
78
796.38 Sun Jun 1 21:54:23 CEST 2014
80 - check that perl slots actually have enough space to hold
81 interpreter variables.
82 - untested port to perl 5.19 (64 bit tmps indices) (reported
83 by Andreas König).
84 - croak when cancel is called without a thread context.
85
866.37 Tue Mar 4 13:27:33 CET 2014
87 - *sigh*, removed leftover debugging code from debugging a
88 perl bug, of all things.
89
906.36 Tue Mar 4 07:11:59 CET 2014
91 - semaphores would not clear the destroy hook when interrupted
92 by ->throw, causing segfaults or worse.
93 - ->throw on a thread waiting for a semaphore did not acquire
94 the semaphore, but also didn't wake up other waiters,
95 possibly causing a deadlock.
96 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
97 to report a full stacktrace, which should help find out
98 where the actual call is.
99 - "atomically" destroy data in slf_destroy, because it is
100 the right thing to do, just in case.
101 - disable ecb.h libm dependency, to compile on ancient systems
102 or under adverse conditions.
103
1046.33 Mon Nov 18 11:26:27 CET 2013
105 - do not crash when freeing padlists with holes (in 5.18).
106 - tentative SVt_BIND 5.19 port/fix.
107
1086.32 Tue Nov 5 15:35:35 CET 2013
109 - use a new algorithm to derive padlists for perl 5.18. The old
110 one could lead to 0-pointer accesses inside perl (reported
111 by Darin McBride).
112
1136.31 Thu May 9 07:39:48 CEST 2013
114 - Coro::AIO requests would crash if the thread was ready'd
115 while the request was ongoing.
116
1176.29 Wed May 8 02:55:18 CEST 2013
118 - when an on_destroy handler destructs the coro currently being
119 destructed a perl scalar could be accessed after being freed,
120 likely causing a crash.
121
1226.28 Wed Mar 6 06:58:02 CET 2013
123 - clean remnants of existing __DIE__ and __WARN__ handlers so
124 they lose their magic and will not cause segfaults later
125 (testcase by Andrey Sagulin).
126 - improved Coro::State documentation a bit.
127 - Coro::Debug::command now flushes the output.
128 - add hack detection code for x32 abi, because the braindead slugs
129 who designed that made it look exactly like x86_64 without
130 providing proper compile time symbols to test for it. as a result,
131 this detection cannot work reliably.
132 - valgrind stack registering was broken.
133 - do not rely on Time::HiRes anymore in Coro::Debug.
134
1356.23 Fri Dec 7 23:36:37 CET 2012
136 - use experimental fiber implementation on native windows
137 perls.
138 - use sizeof (void *) as multiplication factor for stack sizes,
139 to accomodate the totally braindamaged microsoft 64 bit "os".
140 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
141 activeperl 5.16 crashes when PerlIO_define_layer is called due
142 to some bug in the perl dll, strawberry perl at least passes
143 the testsuite.
144 - implement Coro::Handle->peeraddr/host/port, for slightly
145 improved compatibility with LWP.
146 - implement 5.17 compatibility by almost blindly applying a
147 good-looking patch by Father Chrysostomos.
148 - move stack management functions into libcoro 3.
149 - libcoro version 3 "released".
150 - support magic values as timed_io_once args.
151 - recommend AnyEvent 7+ or EV 4+, also require EV
152 version 4 or newer for Coro::EV.
153
1546.10 Tue Oct 9 01:14:27 CEST 2012
155 - updated ecb.h, it had a typo that caused it to not compile on many
156 big endian systems (reported by many people).
157 - disable memory fences in ecb.h to improve portability.
158
1596.09 Sat Oct 6 23:25:02 CEST 2012
160 - Coro::EV I/O watchers were not interruptible by exceptions
161 (Coro::State::throw) (testcase by sten).
162 - ->throw now puts threads into the ready queue, as this seems to
163 be expected by existing code, and code that doesn't cope with spurious
164 wakeups needs fixing anyway.
165 - use fd -1 in mmap.
166 - cast I32 to int in error message printf.
167 - warn about broken so-called "hardened" kernels.
168
1696.08 Fri Apr 13 12:05:47 CEST 2012
170 - be more aggressive about exiting like perl does - formerly,
171 exiting from the non-main thread would not execute END blocks.
172
1736.07 Fri Nov 11 21:21:48 CET 2011
174 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
175 - update ecb.h.
176
1776.06 Mon Aug 8 23:59:48 CEST 2011
178 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
179 backend, which is an order of magnitude slower. unfortunately, cygwins
180 pthread implementation isn't very complete either, so allocate the stack
181 twice just to be sure.
182 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
183 in its ithreaded perl - best recompile cygwin and use the 'w'indows
184 backend for much better performance. also disable ithreads for
185 even better performance...).
186
1876.05 Thu Aug 4 21:36:36 CEST 2011
188 - blush, condvar values would not be propagated from send to recv anymore
189 (reported by Chip Salzenberg).
190 - use exponential increase for the readline buffer length in
191 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
192
1936.04 Wed Aug 3 17:15:45 CEST 2011
194 - use even more efficient and more compatible condvars for
195 compatibility to AnyEvent 6.x :)
196 - more inconsequential ecb.h updates.
6 197
76.03 Wed Aug 3 11:41:30 CEST 2011 1986.03 Wed Aug 3 11:41:30 CEST 2011
8 - change how Coro patches AnyEvent condvars for compatibility to 199 - change how Coro patches AnyEvent condvars for compatibility to
9 AnyEvent 6.x. 200 AnyEvent 6.x.
10 - update ecb.h, to no longer include <pthread.h> in case WinNT.h 201 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
16 handle and barfed. now it's just a bit slower on gentoo and similar 207 handle and barfed. now it's just a bit slower on gentoo and similar
17 systems. 208 systems.
18 209
196.01 Sun Jul 3 12:31:14 CEST 2011 2106.01 Sun Jul 3 12:31:14 CEST 2011
20 - workarounds are good, but the test for whether pthreads are used 211 - workarounds are good, but the test for whether pthreads are used
21 was not good. thisone should be better. 212 was not good. this one should be better.
22 - check differently whether gcc generates cfi instructions itself. 213 - check differently whether gcc generates cfi instructions itself.
23 214
246.0 Wed Jun 29 19:43:35 CEST 2011 2156.0 Wed Jun 29 19:43:35 CEST 2011
25 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be 216 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
26 destroyed and cleaned up automatically (e.g. async { schedule }). 217 destroyed and cleaned up automatically (e.g. async { schedule }).
27 - implement a JIT compiler for part of the thread switch code, 218 - implement a JIT compiler for part of the thread switch code,
28 which gives a 50% speed improvement on threaded perls, and 219 which gives a 50% speed improvement on threaded perls, and
29 about 4% on non-thraeded perls (so threaded perls now finally 220 about 4% on non-threaded perls (so threaded perls now finally
30 reach about half the speed of non-threaded perls). 221 reach about half the speed of non-threaded perls).
31 - slightly modernise Coro::Intro, add section about rouse functions. 222 - slightly modernise Coro::Intro, add section about rouse functions.
32 - avoid DEFSV and ERRSV, giving another 10% improvement 223 - avoid DEFSV and ERRSV, giving another 10% improvement
33 in thread switching. 224 in thread switching.
34 - Coro::State->is_destroyed is now called is_zombie. 225 - Coro::State->is_destroyed is now called is_zombie.
388 - ->throw is now supported on Coro::State objects. 579 - ->throw is now supported on Coro::State objects.
389 - clean up cctx creation code a bit. 580 - clean up cctx creation code a bit.
390 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 581 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
391 582
3924.9 Sat Nov 8 17:45:27 CET 2008 5834.9 Sat Nov 8 17:45:27 CET 2008
393 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 584 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
394 - (libcoro) no longer leak threads in the experimental pthread backend, 585 - (libcoro) no longer leak threads in the experimental pthread backend,
395 also speed it up considerably. 586 also speed it up considerably.
396 - (libcoro) do not rely on makecontext passing void *'s unscathed. 587 - (libcoro) do not rely on makecontext passing void *'s unscathed.
397 - fix compiletime dependencies on libcoro in the Makefile. 588 - fix compiletime dependencies on libcoro in the Makefile.
398 - cctx_count wasn't always updated properly. 589 - cctx_count wasn't always updated properly.
460 - remove debugging code related to MgPV_nolen_const, also try to 651 - remove debugging code related to MgPV_nolen_const, also try to
461 make it compile with perl 5.8.6 (yes, apple apparently loves 652 make it compile with perl 5.8.6 (yes, apple apparently loves
462 outdated software). Reported by John S. 653 outdated software). Reported by John S.
463 654
4644.744 Tue Jul 8 22:06:35 CEST 2008 6554.744 Tue Jul 8 22:06:35 CEST 2008
465 - correctly provide default DIE/WARN handlers as documented. 656 - correctly provide default DIE/WARN handlers as documented.
466 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 657 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
467 implementations inside perl work fine for us. 658 implementations inside perl work fine for us.
468 659
4694.743 Mon Jun 16 00:21:57 CEST 2008 6604.743 Mon Jun 16 00:21:57 CEST 2008
470 - when using Coro::EV without running EV::loop it could 661 - when using Coro::EV without running EV::loop it could
492 - sprinkle "no warnings" freely over everything, also suppress 683 - sprinkle "no warnings" freely over everything, also suppress
493 warnings for some other modules. 684 warnings for some other modules.
494 - fix typo in WSAEWOULDBLOCK. 685 - fix typo in WSAEWOULDBLOCK.
495 686
4964.72 Sun May 25 05:14:36 CEST 2008 6874.72 Sun May 25 05:14:36 CEST 2008
497 - tweak META.yaml a bit, unfortunately, there is no documented way 688 - tweak META.yaml a bit, unfortunately, there is no documented way
498 to have optional dependencies with CPAN. doh :( 689 to have optional dependencies with CPAN. doh :(
499 - avoid running some tests on windows because they would fail due to 690 - avoid running some tests on windows because they would fail due to
500 perl bug (broken fork, broken pipes...). 691 perl bug (broken fork, broken pipes...).
501 - work around perl on windows bugs where perl returns undocumented 692 - work around perl on windows bugs where perl returns undocumented
502 error codes for sysread, syswrite etc. by taking advantage 693 error codes for sysread, syswrite etc. by taking advantage
506 - use unix domain sockets in testsuite to work around 697 - use unix domain sockets in testsuite to work around
507 common perl implementation bugs on widows (they are emulated by 698 common perl implementation bugs on widows (they are emulated by
508 tcp sockets on windows. ugh.) 699 tcp sockets on windows. ugh.)
509 700
5104.71 Sat May 24 20:01:27 CEST 2008 7014.71 Sat May 24 20:01:27 CEST 2008
511 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 702 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
512 - take advantage of async name resolution of AnyEvent::Util. 703 - take advantage of async name resolution of AnyEvent::Util.
513 - work around brutal inet_aton override in Coro::LWP. 704 - work around brutal inet_aton override in Coro::LWP.
514 - take advantage of the readyhook in Coro::EV, for smoother 705 - take advantage of the readyhook in Coro::EV, for smoother
515 scheduling. 706 scheduling.
516 707
5174.7 Sun May 11 00:32:19 CEST 2008 7084.7 Sun May 11 00:32:19 CEST 2008
518 - completely reworked the Coro manpage. 709 - completely reworked the Coro manpage.
519 - added Coro::AnyEvent, generic event loop integration. 710 - added Coro::AnyEvent, generic event loop integration.
520 - implement cancel, ready and kill commands in Coro::Debug. 711 - implement cancel, ready and kill commands in Coro::Debug.
521 - document find_coro in Coro::Debug. 712 - document find_coro in Coro::Debug.
522 - incompatible change: rename has_stack to has_cctx. 713 - incompatible change: rename has_stack to has_cctx.
523 - Coro::AIO and Coro::BDB no longer force event model detection, 714 - Coro::AIO and Coro::BDB no longer force event model detection,
524 use AnyEvent::AIO and AnyEvent::BDB. 715 use AnyEvent::AIO and AnyEvent::BDB.
648 839
6494.11 Thu Oct 11 02:40:24 CEST 2007 8404.11 Thu Oct 11 02:40:24 CEST 2007
650 - port to threaded perls. 841 - port to threaded perls.
651 842
6524.1 Thu Oct 11 02:38:16 CEST 2007 8434.1 Thu Oct 11 02:38:16 CEST 2007
653 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 844 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
654 be local to each coro (see Coro::State). 845 be local to each coro (see Coro::State).
655 - incompatible change: for very deep reasons, cede and cede_notself 846 - incompatible change: for very deep reasons, cede and cede_notself
656 cannot return anything, so nothing will be returned. 847 cannot return anything, so nothing will be returned.
657 - possibly bring back 5.10 compatibility (untested). 848 - possibly bring back 5.10 compatibility (untested).
658 - work around stupid (and wrong) warning on 5.10 :(. 849 - work around stupid (and wrong) warning on 5.10 :(.
730 but stateful semantics. 921 but stateful semantics.
731 - fixed a lot of typos in Coro.pm (patch submitted by David 922 - fixed a lot of typos in Coro.pm (patch submitted by David
732 Steinbrunner, which applied flawlessly). 923 Steinbrunner, which applied flawlessly).
733 924
7343.6 Sat Apr 14 17:13:31 CEST 2007 9253.6 Sat Apr 14 17:13:31 CEST 2007
735 - added some bugfixes to get eg/myhttpd working again. 926 - added some bugfixes to get eg/myhttpd working again.
736 - added Coro::Storable for often-cede'ing freeze/thaw. 927 - added Coro::Storable for often-cede'ing freeze/thaw.
737 - try to do a clean exit when a coroutine calls exit 928 - try to do a clean exit when a coroutine calls exit
738 (EXPERIMENTAL). 929 (EXPERIMENTAL).
739 - got rid of indirect call through _coro_init. 930 - got rid of indirect call through _coro_init.
740 - updated the partly antique examples in eg/ to 931 - updated the partly antique examples in eg/ to

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines