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

Comparing Coro/Changes (file contents):
Revision 1.549 by root, Sat Feb 19 04:34:58 2011 UTC vs.
Revision 1.633 by root, Mon Jun 2 18:32:31 2014 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: unready_all
5TODO: myhttpd header parsing
6TODO: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
7TODO: 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 runc anceled thread?
8 9
10 - restore portability to perl 5.8.x.
11
126.39 Mon Jun 2 00:00:08 CEST 2014
13 - work around more incompatible changes in 5.20.
14
156.38 Sun Jun 1 21:54:23 CEST 2014
16 - check that perl slots actually have enough space to hold
17 interpreter variables.
18 - untested port to perl 5.19 (64 bit tmps indices) (reported
19 by Andreas König).
20 - croak when cancel is called without a thread context.
21
226.37 Tue Mar 4 13:27:33 CET 2014
23 - *sigh*, removed leftover debugging code from debugging a
24 perl bug, of all things.
25
266.36 Tue Mar 4 07:11:59 CET 2014
27 - semaphores would not clear the destroy hook when interrupted
28 by ->throw, causing segfaults or worse.
29 - ->throw on a thread waiting for a semaphore did not acquire
30 the semaphore, but also didn't wake up other waiters,
31 possibly causing a deadlock.
32 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
33 to report a full stacktrace, which should help find out
34 where the actual call is.
35 - "atomically" destroy data in slf_destroy, because it is
36 the right thing to do, just in case.
37 - disable ecb.h libm dependency, to compile on ancient systems
38 or under adverse conditions.
39
406.33 Mon Nov 18 11:26:27 CET 2013
41 - do not crash when freeing padlists with holes (in 5.18).
42 - tentative SVt_BIND 5.19 port/fix.
43
446.32 Tue Nov 5 15:35:35 CET 2013
45 - use a new algorithm to derive padlists for perl 5.18. The old
46 one could lead to 0-pointer accesses inside perl (reported
47 by Darin McBride).
48
496.31 Thu May 9 07:39:48 CEST 2013
50 - Coro::AIO requests would crash if the thread was ready'd
51 while the request was ongoing.
52
536.29 Wed May 8 02:55:18 CEST 2013
54 - when an on_destroy handler destructs the coro currently being
55 destructed a perl scalar could be accessed after being freed,
56 likely causing a crash.
57
586.28 Wed Mar 6 06:58:02 CET 2013
59 - clean remnants of existing __DIE__ and __WARN__ handlers so
60 they lose their magic and will not cause segfaults later
61 (testcase by Andrey Sagulin).
62 - improved Coro::State documentation a bit.
63 - Coro::Debug::command now flushes the output.
64 - add hack detection code for x32 abi, because the braindead slugs
65 who designed that made it look exactly like x86_64 without
66 providing proper compile time symbols to test for it. as a result,
67 this detection cannot work reliably.
68 - valgrind stack registering was broken.
69 - do not rely on Time::HiRes anymore in Coro::Debug.
70
716.23 Fri Dec 7 23:36:37 CET 2012
72 - use experimental fiber implementation on native windows
73 perls.
74 - use sizeof (void *) as multiplication factor for stack sizes,
75 to accomodate the totally braindamaged microsoft 64 bit "os".
76 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
77 activeperl 5.16 crashes when PerlIO_define_layer is called due
78 to some bug in the perl dll, strawberry perl at least passes
79 the testsuite.
80 - implement Coro::Handle->peeraddr/host/port, for slightly
81 improved compatibility with LWP.
82 - implement 5.17 compatibility by almost blindly applying a
83 good-looking patch by Father Chrysostomos.
84 - move stack management functions into libcoro 3.
85 - libcoro version 3 "released".
86 - support magic values as timed_io_once args.
87 - recommend AnyEvent 7+ or EV 4+, also require EV
88 version 4 or newer for Coro::EV.
89
906.10 Tue Oct 9 01:14:27 CEST 2012
91 - updated ecb.h, it had a typo that caused it to not compile on many
92 big endian systems (reported by many people).
93 - disable memory fences in ecb.h to improve portability.
94
956.09 Sat Oct 6 23:25:02 CEST 2012
96 - Coro::EV I/O watchers were not interruptible by exceptions
97 (Coro::State::throw) (testcase by sten).
98 - ->throw now puts threads into the ready queue, as this seems to
99 be expected by existing code, and code that doesn't cope with spurious
100 wakeups needs fixing anyway.
101 - use fd -1 in mmap.
102 - cast I32 to int in error message printf.
103 - warn about broken so-called "hardened" kernels.
104
1056.08 Fri Apr 13 12:05:47 CEST 2012
106 - be more aggressive about exiting like perl does - formerly,
107 exiting from the non-main thread would not execute END blocks.
108
1096.07 Fri Nov 11 21:21:48 CET 2011
110 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
111 - update ecb.h.
112
1136.06 Mon Aug 8 23:59:48 CEST 2011
114 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
115 backend, which is an order of magnitude slower. unfortunately, cygwins
116 pthread implementation isn't very complete either, so allocate the stack
117 twice just to be sure.
118 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
119 in its ithreaded perl - best recompile cygwin and use the 'w'indows
120 backend for much better performance. also disable ithreads for
121 even better performance...).
122
1236.05 Thu Aug 4 21:36:36 CEST 2011
124 - blush, condvar values would not be propagated from send to recv anymore
125 (reported by Chip Salzenberg).
126 - use exponential increase for the readline buffer length in
127 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
128
1296.04 Wed Aug 3 17:15:45 CEST 2011
130 - use even more efficient and more compatible condvars for
131 compatibility to AnyEvent 6.x :)
132 - more inconsequential ecb.h updates.
133
1346.03 Wed Aug 3 11:41:30 CEST 2011
135 - change how Coro patches AnyEvent condvars for compatibility to
136 AnyEvent 6.x.
137 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
138 hasn't been included.
139
1406.02 Wed Jul 13 04:35:19 CEST 2011
141 - "improve portability to Gentoo" - gentoo manages to put perl variables
142 in memory areas that are farther than 2gb apart, which the jit couldn't
143 handle and barfed. now it's just a bit slower on gentoo and similar
144 systems.
145
1466.01 Sun Jul 3 12:31:14 CEST 2011
147 - workarounds are good, but the test for whether pthreads are used
148 was not good. this one should be better.
149 - check differently whether gcc generates cfi instructions itself.
150
1516.0 Wed Jun 29 19:43:35 CEST 2011
152 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
153 destroyed and cleaned up automatically (e.g. async { schedule }).
154 - implement a JIT compiler for part of the thread switch code,
155 which gives a 50% speed improvement on threaded perls, and
156 about 4% on non-threaded perls (so threaded perls now finally
157 reach about half the speed of non-threaded perls).
158 - slightly modernise Coro::Intro, add section about rouse functions.
159 - avoid DEFSV and ERRSV, giving another 10% improvement
160 in thread switching.
161 - Coro::State->is_destroyed is now called is_zombie.
162 - implement a Coro->safe_cancel method that might fail, but
163 cancels in a "safer" way if it succeeds.
164 - add preliminary support for DEBUGGING perls.
165 - get rid of two hash-accesses when initialising a new Coro - this
166 speeds up coro creation by almost a factor of two.
167 - croak when a coro that is being cancelled tries to block
168 (e.g. while executing a guard block), instead of crashing or
169 deadlocking.
170 - use a more robust and also faster method to identify Coro::State
171 objects - speeds up everything a bit.
172 - implement Coro->cancel in XS for a 20% speed improvement, and to
173 be able to implement mutual cancellation.
174 - speed up context switches by a percent or two by more efficiently
175 allocating context stack entries.
176 - implement Coro->join and Coro->on_destroy in XS for a speedup and
177 a reduction in memory use.
178 - cancelling a coro while it itself is cancelling another coro is
179 now supported and working, instead of triggering an assertion.
180 - be a bit more crash-resistant when calling (buggy) on_destroy
181 callbacks (best effort).
182 - move on_destroy into the slf_frame, to allow extension slf
183 functions to have destructors.
184 - get rid if coro refcounting - simply crash in other interpreter
185 threads by nulling the pointers on clone.
186 - simplify warn/die hook handling when loading Coro - the convoluted
187 logic seems to be no longer neccessary.
188 - use libecb instead of our own home-grown gcc hacks.
189 - document alternatives to Coro::LWP. Please use them :)
190 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
191 sigaltstack. Really. wine suffers from it, erlang suffers from it,
192 and it's known since at least 2006.
193
1945.372 Wed Feb 23 06:14:30 CET 2011
195 - apparently mingw doesn't provide a working gettimeofday, try to
196 work around that by relying on Time::HiRes (indirectly brought to
197 my attention by Max Maischein).
198 - fix some portability issues when Time::HiRes was used.
199
2005.371 Mon Feb 21 14:36:08 CET 2011
201 - backport to windows process emulation code again.
202
2035.37 Sat Feb 19 07:49:44 CET 2011
9 - add a big "Coro thread life cycle" section to "man Coro". 204 - add a big "Coro thread life cycle" section to "man Coro".
10 - try a tentative workaround against the breakage that 5.13 has 205 - try a tentative workaround against the breakage that 5.13 has
11 introduced without depreciation period. sigh. 206 introduced without depreciation period. sigh.
207 - no longer use Time::HiRes if gettimeofday is available, which
208 saves quite a lot of memory.
12 209
135.36 Sun Feb 13 05:33:41 CET 2011 2105.36 Sun Feb 13 05:33:41 CET 2011
14 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore, 211 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
15 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new" 212 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
16 method call. 213 method call.
… …
318 - ->throw is now supported on Coro::State objects. 515 - ->throw is now supported on Coro::State objects.
319 - clean up cctx creation code a bit. 516 - clean up cctx creation code a bit.
320 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 517 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
321 518
3224.9 Sat Nov 8 17:45:27 CET 2008 5194.9 Sat Nov 8 17:45:27 CET 2008
323 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 520 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
324 - (libcoro) no longer leak threads in the experimental pthread backend, 521 - (libcoro) no longer leak threads in the experimental pthread backend,
325 also speed it up considerably. 522 also speed it up considerably.
326 - (libcoro) do not rely on makecontext passing void *'s unscathed. 523 - (libcoro) do not rely on makecontext passing void *'s unscathed.
327 - fix compiletime dependencies on libcoro in the Makefile. 524 - fix compiletime dependencies on libcoro in the Makefile.
328 - cctx_count wasn't always updated properly. 525 - cctx_count wasn't always updated properly.
… …
390 - remove debugging code related to MgPV_nolen_const, also try to 587 - remove debugging code related to MgPV_nolen_const, also try to
391 make it compile with perl 5.8.6 (yes, apple apparently loves 588 make it compile with perl 5.8.6 (yes, apple apparently loves
392 outdated software). Reported by John S. 589 outdated software). Reported by John S.
393 590
3944.744 Tue Jul 8 22:06:35 CEST 2008 5914.744 Tue Jul 8 22:06:35 CEST 2008
395 - correctly provide default DIE/WARN handlers as documented. 592 - correctly provide default DIE/WARN handlers as documented.
396 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 593 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
397 implementations inside perl work fine for us. 594 implementations inside perl work fine for us.
398 595
3994.743 Mon Jun 16 00:21:57 CEST 2008 5964.743 Mon Jun 16 00:21:57 CEST 2008
400 - when using Coro::EV without running EV::loop it could 597 - when using Coro::EV without running EV::loop it could
… …
422 - sprinkle "no warnings" freely over everything, also suppress 619 - sprinkle "no warnings" freely over everything, also suppress
423 warnings for some other modules. 620 warnings for some other modules.
424 - fix typo in WSAEWOULDBLOCK. 621 - fix typo in WSAEWOULDBLOCK.
425 622
4264.72 Sun May 25 05:14:36 CEST 2008 6234.72 Sun May 25 05:14:36 CEST 2008
427 - tweak META.yaml a bit, unfortunately, there is no documented way 624 - tweak META.yaml a bit, unfortunately, there is no documented way
428 to have optional dependencies with CPAN. doh :( 625 to have optional dependencies with CPAN. doh :(
429 - avoid running some tests on windows because they would fail due to 626 - avoid running some tests on windows because they would fail due to
430 perl bug (broken fork, broken pipes...). 627 perl bug (broken fork, broken pipes...).
431 - work around perl on windows bugs where perl returns undocumented 628 - work around perl on windows bugs where perl returns undocumented
432 error codes for sysread, syswrite etc. by taking advantage 629 error codes for sysread, syswrite etc. by taking advantage
… …
436 - use unix domain sockets in testsuite to work around 633 - use unix domain sockets in testsuite to work around
437 common perl implementation bugs on widows (they are emulated by 634 common perl implementation bugs on widows (they are emulated by
438 tcp sockets on windows. ugh.) 635 tcp sockets on windows. ugh.)
439 636
4404.71 Sat May 24 20:01:27 CEST 2008 6374.71 Sat May 24 20:01:27 CEST 2008
441 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 638 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
442 - take advantage of async name resolution of AnyEvent::Util. 639 - take advantage of async name resolution of AnyEvent::Util.
443 - work around brutal inet_aton override in Coro::LWP. 640 - work around brutal inet_aton override in Coro::LWP.
444 - take advantage of the readyhook in Coro::EV, for smoother 641 - take advantage of the readyhook in Coro::EV, for smoother
445 scheduling. 642 scheduling.
446 643
4474.7 Sun May 11 00:32:19 CEST 2008 6444.7 Sun May 11 00:32:19 CEST 2008
448 - completely reworked the Coro manpage. 645 - completely reworked the Coro manpage.
449 - added Coro::AnyEvent, generic event loop integration. 646 - added Coro::AnyEvent, generic event loop integration.
450 - implement cancel, ready and kill commands in Coro::Debug. 647 - implement cancel, ready and kill commands in Coro::Debug.
451 - document find_coro in Coro::Debug. 648 - document find_coro in Coro::Debug.
452 - incompatible change: rename has_stack to has_cctx. 649 - incompatible change: rename has_stack to has_cctx.
453 - Coro::AIO and Coro::BDB no longer force event model detection, 650 - Coro::AIO and Coro::BDB no longer force event model detection,
454 use AnyEvent::AIO and AnyEvent::BDB. 651 use AnyEvent::AIO and AnyEvent::BDB.
… …
578 775
5794.11 Thu Oct 11 02:40:24 CEST 2007 7764.11 Thu Oct 11 02:40:24 CEST 2007
580 - port to threaded perls. 777 - port to threaded perls.
581 778
5824.1 Thu Oct 11 02:38:16 CEST 2007 7794.1 Thu Oct 11 02:38:16 CEST 2007
583 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 780 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
584 be local to each coro (see Coro::State). 781 be local to each coro (see Coro::State).
585 - incompatible change: for very deep reasons, cede and cede_notself 782 - incompatible change: for very deep reasons, cede and cede_notself
586 cannot return anything, so nothing will be returned. 783 cannot return anything, so nothing will be returned.
587 - possibly bring back 5.10 compatibility (untested). 784 - possibly bring back 5.10 compatibility (untested).
588 - work around stupid (and wrong) warning on 5.10 :(. 785 - work around stupid (and wrong) warning on 5.10 :(.
… …
660 but stateful semantics. 857 but stateful semantics.
661 - fixed a lot of typos in Coro.pm (patch submitted by David 858 - fixed a lot of typos in Coro.pm (patch submitted by David
662 Steinbrunner, which applied flawlessly). 859 Steinbrunner, which applied flawlessly).
663 860
6643.6 Sat Apr 14 17:13:31 CEST 2007 8613.6 Sat Apr 14 17:13:31 CEST 2007
665 - added some bugfixes to get eg/myhttpd working again. 862 - added some bugfixes to get eg/myhttpd working again.
666 - added Coro::Storable for often-cede'ing freeze/thaw. 863 - added Coro::Storable for often-cede'ing freeze/thaw.
667 - try to do a clean exit when a coroutine calls exit 864 - try to do a clean exit when a coroutine calls exit
668 (EXPERIMENTAL). 865 (EXPERIMENTAL).
669 - got rid of indirect call through _coro_init. 866 - got rid of indirect call through _coro_init.
670 - updated the partly antique examples in eg/ to 867 - updated the partly antique examples in eg/ to

Diff Legend

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