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

Comparing Coro/Changes (file contents):
Revision 1.562 by root, Tue May 10 19:55:48 2011 UTC vs.
Revision 1.637 by root, Tue Feb 10 17:50:40 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: 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, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
8TODO: croak when async_pool triesd to run canceled thread?
6 9
10 - Coro::SemaphoreSet->try did not actually work (analyzed by
11 SATO Kentaro).
12
136.41 Sat Sep 6 22:08:46 CEST 2014
14 - restore portability to perl 5.8.x.
15 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
16 up some memory, but fixes "use feature" and similar modules
17 when used inside a Coro.
18 - allow zero as argument to Coro::Channel to mean the same thing
19 as no argument. this works with older versions as well,
20 but wasn't legal until now.
21 - slightly better c header file detection.
22
236.39 Mon Jun 2 00:00:08 CEST 2014
24 - work around more incompatible changes in 5.20.
25
266.38 Sun Jun 1 21:54:23 CEST 2014
27 - check that perl slots actually have enough space to hold
28 interpreter variables.
29 - untested port to perl 5.19 (64 bit tmps indices) (reported
30 by Andreas König).
31 - croak when cancel is called without a thread context.
32
336.37 Tue Mar 4 13:27:33 CET 2014
34 - *sigh*, removed leftover debugging code from debugging a
35 perl bug, of all things.
36
376.36 Tue Mar 4 07:11:59 CET 2014
38 - semaphores would not clear the destroy hook when interrupted
39 by ->throw, causing segfaults or worse.
40 - ->throw on a thread waiting for a semaphore did not acquire
41 the semaphore, but also didn't wake up other waiters,
42 possibly causing a deadlock.
43 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
44 to report a full stacktrace, which should help find out
45 where the actual call is.
46 - "atomically" destroy data in slf_destroy, because it is
47 the right thing to do, just in case.
48 - disable ecb.h libm dependency, to compile on ancient systems
49 or under adverse conditions.
50
516.33 Mon Nov 18 11:26:27 CET 2013
52 - do not crash when freeing padlists with holes (in 5.18).
53 - tentative SVt_BIND 5.19 port/fix.
54
556.32 Tue Nov 5 15:35:35 CET 2013
56 - use a new algorithm to derive padlists for perl 5.18. The old
57 one could lead to 0-pointer accesses inside perl (reported
58 by Darin McBride).
59
606.31 Thu May 9 07:39:48 CEST 2013
61 - Coro::AIO requests would crash if the thread was ready'd
62 while the request was ongoing.
63
646.29 Wed May 8 02:55:18 CEST 2013
65 - when an on_destroy handler destructs the coro currently being
66 destructed a perl scalar could be accessed after being freed,
67 likely causing a crash.
68
696.28 Wed Mar 6 06:58:02 CET 2013
70 - clean remnants of existing __DIE__ and __WARN__ handlers so
71 they lose their magic and will not cause segfaults later
72 (testcase by Andrey Sagulin).
73 - improved Coro::State documentation a bit.
74 - Coro::Debug::command now flushes the output.
75 - add hack detection code for x32 abi, because the braindead slugs
76 who designed that made it look exactly like x86_64 without
77 providing proper compile time symbols to test for it. as a result,
78 this detection cannot work reliably.
79 - valgrind stack registering was broken.
80 - do not rely on Time::HiRes anymore in Coro::Debug.
81
826.23 Fri Dec 7 23:36:37 CET 2012
83 - use experimental fiber implementation on native windows
84 perls.
85 - use sizeof (void *) as multiplication factor for stack sizes,
86 to accomodate the totally braindamaged microsoft 64 bit "os".
87 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
88 activeperl 5.16 crashes when PerlIO_define_layer is called due
89 to some bug in the perl dll, strawberry perl at least passes
90 the testsuite.
91 - implement Coro::Handle->peeraddr/host/port, for slightly
92 improved compatibility with LWP.
93 - implement 5.17 compatibility by almost blindly applying a
94 good-looking patch by Father Chrysostomos.
95 - move stack management functions into libcoro 3.
96 - libcoro version 3 "released".
97 - support magic values as timed_io_once args.
98 - recommend AnyEvent 7+ or EV 4+, also require EV
99 version 4 or newer for Coro::EV.
100
1016.10 Tue Oct 9 01:14:27 CEST 2012
102 - updated ecb.h, it had a typo that caused it to not compile on many
103 big endian systems (reported by many people).
104 - disable memory fences in ecb.h to improve portability.
105
1066.09 Sat Oct 6 23:25:02 CEST 2012
107 - Coro::EV I/O watchers were not interruptible by exceptions
108 (Coro::State::throw) (testcase by sten).
109 - ->throw now puts threads into the ready queue, as this seems to
110 be expected by existing code, and code that doesn't cope with spurious
111 wakeups needs fixing anyway.
112 - use fd -1 in mmap.
113 - cast I32 to int in error message printf.
114 - warn about broken so-called "hardened" kernels.
115
1166.08 Fri Apr 13 12:05:47 CEST 2012
117 - be more aggressive about exiting like perl does - formerly,
118 exiting from the non-main thread would not execute END blocks.
119
1206.07 Fri Nov 11 21:21:48 CET 2011
121 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
122 - update ecb.h.
123
1246.06 Mon Aug 8 23:59:48 CEST 2011
125 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
126 backend, which is an order of magnitude slower. unfortunately, cygwins
127 pthread implementation isn't very complete either, so allocate the stack
128 twice just to be sure.
129 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
130 in its ithreaded perl - best recompile cygwin and use the 'w'indows
131 backend for much better performance. also disable ithreads for
132 even better performance...).
133
1346.05 Thu Aug 4 21:36:36 CEST 2011
135 - blush, condvar values would not be propagated from send to recv anymore
136 (reported by Chip Salzenberg).
137 - use exponential increase for the readline buffer length in
138 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
139
1406.04 Wed Aug 3 17:15:45 CEST 2011
141 - use even more efficient and more compatible condvars for
142 compatibility to AnyEvent 6.x :)
143 - more inconsequential ecb.h updates.
144
1456.03 Wed Aug 3 11:41:30 CEST 2011
146 - change how Coro patches AnyEvent condvars for compatibility to
147 AnyEvent 6.x.
148 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
149 hasn't been included.
150
1516.02 Wed Jul 13 04:35:19 CEST 2011
152 - "improve portability to Gentoo" - gentoo manages to put perl variables
153 in memory areas that are farther than 2gb apart, which the jit couldn't
154 handle and barfed. now it's just a bit slower on gentoo and similar
155 systems.
156
1576.01 Sun Jul 3 12:31:14 CEST 2011
158 - workarounds are good, but the test for whether pthreads are used
159 was not good. this one should be better.
160 - check differently whether gcc generates cfi instructions itself.
161
1626.0 Wed Jun 29 19:43:35 CEST 2011
163 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
164 destroyed and cleaned up automatically (e.g. async { schedule }).
165 - implement a JIT compiler for part of the thread switch code,
166 which gives a 50% speed improvement on threaded perls, and
167 about 4% on non-threaded perls (so threaded perls now finally
168 reach about half the speed of non-threaded perls).
169 - slightly modernise Coro::Intro, add section about rouse functions.
170 - avoid DEFSV and ERRSV, giving another 10% improvement
171 in thread switching.
172 - Coro::State->is_destroyed is now called is_zombie.
173 - implement a Coro->safe_cancel method that might fail, but
174 cancels in a "safer" way if it succeeds.
175 - add preliminary support for DEBUGGING perls.
176 - get rid of two hash-accesses when initialising a new Coro - this
177 speeds up coro creation by almost a factor of two.
7 - croak when a coro that is being cancelled tries to block 178 - croak when a coro that is being cancelled tries to block
8 (e.g. while executing a guard block), instead of crashing or 179 (e.g. while executing a guard block), instead of crashing or
9 deadlocking. 180 deadlocking.
10 - implement a Coro->safe_cancel method that might fail, but 181 - use a more robust and also faster method to identify Coro::State
11 cancels in a "safer" way if it succeeds. 182 objects - speeds up everything a bit.
12 - implement Coro->cancel in XS for a 20% speed improvement, and to 183 - implement Coro->cancel in XS for a 20% speed improvement, and to
13 be able to implement mutual cancellation. 184 be able to implement mutual cancellation.
14 - speed up context switches by a percent or two by more efficiently 185 - speed up context switches by a percent or two by more efficiently
15 allocating context stack entries. 186 allocating context stack entries.
187 - implement Coro->join and Coro->on_destroy in XS for a speedup and
188 a reduction in memory use.
16 - cancelling a coro while it itself is cancelling another coro is 189 - cancelling a coro while it itself is cancelling another coro is
17 now supported and working, instead of triggering an assertion. 190 now supported and working, instead of triggering an assertion.
18 - be a bit more crash-resistant when calling (buggy) on_destroy 191 - be a bit more crash-resistant when calling (buggy) on_destroy
19 callbacks (best effort). 192 callbacks (best effort).
20 - move on_destroy into the slf_frame, to allow extension slf 193 - move on_destroy into the slf_frame, to allow extension slf
21 functions to have destructors. 194 functions to have destructors.
195 - get rid if coro refcounting - simply crash in other interpreter
196 threads by nulling the pointers on clone.
197 - simplify warn/die hook handling when loading Coro - the convoluted
198 logic seems to be no longer neccessary.
199 - use libecb instead of our own home-grown gcc hacks.
200 - document alternatives to Coro::LWP. Please use them :)
201 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
202 sigaltstack. Really. wine suffers from it, erlang suffers from it,
203 and it's known since at least 2006.
22 204
235.372 Wed Feb 23 06:14:30 CET 2011 2055.372 Wed Feb 23 06:14:30 CET 2011
24 - apparently mingw doesn't provide a working gettimeofday, try to 206 - apparently mingw doesn't provide a working gettimeofday, try to
25 work around that by relying on Time::HiRes (indirectly brought to 207 work around that by relying on Time::HiRes (indirectly brought to
26 my attention by Max Maischein). 208 my attention by Max Maischein).
… …
344 - ->throw is now supported on Coro::State objects. 526 - ->throw is now supported on Coro::State objects.
345 - clean up cctx creation code a bit. 527 - clean up cctx creation code a bit.
346 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 528 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
347 529
3484.9 Sat Nov 8 17:45:27 CET 2008 5304.9 Sat Nov 8 17:45:27 CET 2008
349 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 531 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
350 - (libcoro) no longer leak threads in the experimental pthread backend, 532 - (libcoro) no longer leak threads in the experimental pthread backend,
351 also speed it up considerably. 533 also speed it up considerably.
352 - (libcoro) do not rely on makecontext passing void *'s unscathed. 534 - (libcoro) do not rely on makecontext passing void *'s unscathed.
353 - fix compiletime dependencies on libcoro in the Makefile. 535 - fix compiletime dependencies on libcoro in the Makefile.
354 - cctx_count wasn't always updated properly. 536 - cctx_count wasn't always updated properly.
… …
416 - remove debugging code related to MgPV_nolen_const, also try to 598 - remove debugging code related to MgPV_nolen_const, also try to
417 make it compile with perl 5.8.6 (yes, apple apparently loves 599 make it compile with perl 5.8.6 (yes, apple apparently loves
418 outdated software). Reported by John S. 600 outdated software). Reported by John S.
419 601
4204.744 Tue Jul 8 22:06:35 CEST 2008 6024.744 Tue Jul 8 22:06:35 CEST 2008
421 - correctly provide default DIE/WARN handlers as documented. 603 - correctly provide default DIE/WARN handlers as documented.
422 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 604 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
423 implementations inside perl work fine for us. 605 implementations inside perl work fine for us.
424 606
4254.743 Mon Jun 16 00:21:57 CEST 2008 6074.743 Mon Jun 16 00:21:57 CEST 2008
426 - when using Coro::EV without running EV::loop it could 608 - when using Coro::EV without running EV::loop it could
… …
448 - sprinkle "no warnings" freely over everything, also suppress 630 - sprinkle "no warnings" freely over everything, also suppress
449 warnings for some other modules. 631 warnings for some other modules.
450 - fix typo in WSAEWOULDBLOCK. 632 - fix typo in WSAEWOULDBLOCK.
451 633
4524.72 Sun May 25 05:14:36 CEST 2008 6344.72 Sun May 25 05:14:36 CEST 2008
453 - tweak META.yaml a bit, unfortunately, there is no documented way 635 - tweak META.yaml a bit, unfortunately, there is no documented way
454 to have optional dependencies with CPAN. doh :( 636 to have optional dependencies with CPAN. doh :(
455 - avoid running some tests on windows because they would fail due to 637 - avoid running some tests on windows because they would fail due to
456 perl bug (broken fork, broken pipes...). 638 perl bug (broken fork, broken pipes...).
457 - work around perl on windows bugs where perl returns undocumented 639 - work around perl on windows bugs where perl returns undocumented
458 error codes for sysread, syswrite etc. by taking advantage 640 error codes for sysread, syswrite etc. by taking advantage
… …
462 - use unix domain sockets in testsuite to work around 644 - use unix domain sockets in testsuite to work around
463 common perl implementation bugs on widows (they are emulated by 645 common perl implementation bugs on widows (they are emulated by
464 tcp sockets on windows. ugh.) 646 tcp sockets on windows. ugh.)
465 647
4664.71 Sat May 24 20:01:27 CEST 2008 6484.71 Sat May 24 20:01:27 CEST 2008
467 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 649 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
468 - take advantage of async name resolution of AnyEvent::Util. 650 - take advantage of async name resolution of AnyEvent::Util.
469 - work around brutal inet_aton override in Coro::LWP. 651 - work around brutal inet_aton override in Coro::LWP.
470 - take advantage of the readyhook in Coro::EV, for smoother 652 - take advantage of the readyhook in Coro::EV, for smoother
471 scheduling. 653 scheduling.
472 654
4734.7 Sun May 11 00:32:19 CEST 2008 6554.7 Sun May 11 00:32:19 CEST 2008
474 - completely reworked the Coro manpage. 656 - completely reworked the Coro manpage.
475 - added Coro::AnyEvent, generic event loop integration. 657 - added Coro::AnyEvent, generic event loop integration.
476 - implement cancel, ready and kill commands in Coro::Debug. 658 - implement cancel, ready and kill commands in Coro::Debug.
477 - document find_coro in Coro::Debug. 659 - document find_coro in Coro::Debug.
478 - incompatible change: rename has_stack to has_cctx. 660 - incompatible change: rename has_stack to has_cctx.
479 - Coro::AIO and Coro::BDB no longer force event model detection, 661 - Coro::AIO and Coro::BDB no longer force event model detection,
480 use AnyEvent::AIO and AnyEvent::BDB. 662 use AnyEvent::AIO and AnyEvent::BDB.
… …
604 786
6054.11 Thu Oct 11 02:40:24 CEST 2007 7874.11 Thu Oct 11 02:40:24 CEST 2007
606 - port to threaded perls. 788 - port to threaded perls.
607 789
6084.1 Thu Oct 11 02:38:16 CEST 2007 7904.1 Thu Oct 11 02:38:16 CEST 2007
609 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 791 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
610 be local to each coro (see Coro::State). 792 be local to each coro (see Coro::State).
611 - incompatible change: for very deep reasons, cede and cede_notself 793 - incompatible change: for very deep reasons, cede and cede_notself
612 cannot return anything, so nothing will be returned. 794 cannot return anything, so nothing will be returned.
613 - possibly bring back 5.10 compatibility (untested). 795 - possibly bring back 5.10 compatibility (untested).
614 - work around stupid (and wrong) warning on 5.10 :(. 796 - work around stupid (and wrong) warning on 5.10 :(.
… …
686 but stateful semantics. 868 but stateful semantics.
687 - fixed a lot of typos in Coro.pm (patch submitted by David 869 - fixed a lot of typos in Coro.pm (patch submitted by David
688 Steinbrunner, which applied flawlessly). 870 Steinbrunner, which applied flawlessly).
689 871
6903.6 Sat Apr 14 17:13:31 CEST 2007 8723.6 Sat Apr 14 17:13:31 CEST 2007
691 - added some bugfixes to get eg/myhttpd working again. 873 - added some bugfixes to get eg/myhttpd working again.
692 - added Coro::Storable for often-cede'ing freeze/thaw. 874 - added Coro::Storable for often-cede'ing freeze/thaw.
693 - try to do a clean exit when a coroutine calls exit 875 - try to do a clean exit when a coroutine calls exit
694 (EXPERIMENTAL). 876 (EXPERIMENTAL).
695 - got rid of indirect call through _coro_init. 877 - got rid of indirect call through _coro_init.
696 - updated the partly antique examples in eg/ to 878 - updated the partly antique examples in eg/ to

Diff Legend

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