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

Comparing Coro/Changes (file contents):
Revision 1.539 by root, Mon May 17 16:51:44 2010 UTC vs.
Revision 1.632 by root, Sun Jun 1 22:00:45 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?
9
106.39 Mon Jun 2 00:00:08 CEST 2014
11 - work around more incompatible changes in 5.20.
12
136.38 Sun Jun 1 21:54:23 CEST 2014
14 - check that perl slots actually have enough space to hold
15 interpreter variables.
16 - untested port to perl 5.19 (64 bit tmps indices) (reported
17 by Andreas König).
18 - croak when cancel is called without a thread context.
19
206.37 Tue Mar 4 13:27:33 CET 2014
21 - *sigh*, removed leftover debugging code from debugging a
22 perl bug, of all things.
23
246.36 Tue Mar 4 07:11:59 CET 2014
25 - semaphores would not clear the destroy hook when interrupted
26 by ->throw, causing segfaults or worse.
27 - ->throw on a thread waiting for a semaphore did not acquire
28 the semaphore, but also didn't wake up other waiters,
29 possibly causing a deadlock.
30 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
31 to report a full stacktrace, which should help find out
32 where the actual call is.
33 - "atomically" destroy data in slf_destroy, because it is
34 the right thing to do, just in case.
35 - disable ecb.h libm dependency, to compile on ancient systems
36 or under adverse conditions.
37
386.33 Mon Nov 18 11:26:27 CET 2013
39 - do not crash when freeing padlists with holes (in 5.18).
40 - tentative SVt_BIND 5.19 port/fix.
41
426.32 Tue Nov 5 15:35:35 CET 2013
43 - use a new algorithm to derive padlists for perl 5.18. The old
44 one could lead to 0-pointer accesses inside perl (reported
45 by Darin McBride).
46
476.31 Thu May 9 07:39:48 CEST 2013
48 - Coro::AIO requests would crash if the thread was ready'd
49 while the request was ongoing.
50
516.29 Wed May 8 02:55:18 CEST 2013
52 - when an on_destroy handler destructs the coro currently being
53 destructed a perl scalar could be accessed after being freed,
54 likely causing a crash.
55
566.28 Wed Mar 6 06:58:02 CET 2013
57 - clean remnants of existing __DIE__ and __WARN__ handlers so
58 they lose their magic and will not cause segfaults later
59 (testcase by Andrey Sagulin).
60 - improved Coro::State documentation a bit.
61 - Coro::Debug::command now flushes the output.
62 - add hack detection code for x32 abi, because the braindead slugs
63 who designed that made it look exactly like x86_64 without
64 providing proper compile time symbols to test for it. as a result,
65 this detection cannot work reliably.
66 - valgrind stack registering was broken.
67 - do not rely on Time::HiRes anymore in Coro::Debug.
68
696.23 Fri Dec 7 23:36:37 CET 2012
70 - use experimental fiber implementation on native windows
71 perls.
72 - use sizeof (void *) as multiplication factor for stack sizes,
73 to accomodate the totally braindamaged microsoft 64 bit "os".
74 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
75 activeperl 5.16 crashes when PerlIO_define_layer is called due
76 to some bug in the perl dll, strawberry perl at least passes
77 the testsuite.
78 - implement Coro::Handle->peeraddr/host/port, for slightly
79 improved compatibility with LWP.
80 - implement 5.17 compatibility by almost blindly applying a
81 good-looking patch by Father Chrysostomos.
82 - move stack management functions into libcoro 3.
83 - libcoro version 3 "released".
84 - support magic values as timed_io_once args.
85 - recommend AnyEvent 7+ or EV 4+, also require EV
86 version 4 or newer for Coro::EV.
87
886.10 Tue Oct 9 01:14:27 CEST 2012
89 - updated ecb.h, it had a typo that caused it to not compile on many
90 big endian systems (reported by many people).
91 - disable memory fences in ecb.h to improve portability.
92
936.09 Sat Oct 6 23:25:02 CEST 2012
94 - Coro::EV I/O watchers were not interruptible by exceptions
95 (Coro::State::throw) (testcase by sten).
96 - ->throw now puts threads into the ready queue, as this seems to
97 be expected by existing code, and code that doesn't cope with spurious
98 wakeups needs fixing anyway.
99 - use fd -1 in mmap.
100 - cast I32 to int in error message printf.
101 - warn about broken so-called "hardened" kernels.
102
1036.08 Fri Apr 13 12:05:47 CEST 2012
104 - be more aggressive about exiting like perl does - formerly,
105 exiting from the non-main thread would not execute END blocks.
106
1076.07 Fri Nov 11 21:21:48 CET 2011
108 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
109 - update ecb.h.
110
1116.06 Mon Aug 8 23:59:48 CEST 2011
112 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
113 backend, which is an order of magnitude slower. unfortunately, cygwins
114 pthread implementation isn't very complete either, so allocate the stack
115 twice just to be sure.
116 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
117 in its ithreaded perl - best recompile cygwin and use the 'w'indows
118 backend for much better performance. also disable ithreads for
119 even better performance...).
120
1216.05 Thu Aug 4 21:36:36 CEST 2011
122 - blush, condvar values would not be propagated from send to recv anymore
123 (reported by Chip Salzenberg).
124 - use exponential increase for the readline buffer length in
125 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
126
1276.04 Wed Aug 3 17:15:45 CEST 2011
128 - use even more efficient and more compatible condvars for
129 compatibility to AnyEvent 6.x :)
130 - more inconsequential ecb.h updates.
131
1326.03 Wed Aug 3 11:41:30 CEST 2011
133 - change how Coro patches AnyEvent condvars for compatibility to
134 AnyEvent 6.x.
135 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
136 hasn't been included.
137
1386.02 Wed Jul 13 04:35:19 CEST 2011
139 - "improve portability to Gentoo" - gentoo manages to put perl variables
140 in memory areas that are farther than 2gb apart, which the jit couldn't
141 handle and barfed. now it's just a bit slower on gentoo and similar
142 systems.
143
1446.01 Sun Jul 3 12:31:14 CEST 2011
145 - workarounds are good, but the test for whether pthreads are used
146 was not good. this one should be better.
147 - check differently whether gcc generates cfi instructions itself.
148
1496.0 Wed Jun 29 19:43:35 CEST 2011
150 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
151 destroyed and cleaned up automatically (e.g. async { schedule }).
152 - implement a JIT compiler for part of the thread switch code,
153 which gives a 50% speed improvement on threaded perls, and
154 about 4% on non-threaded perls (so threaded perls now finally
155 reach about half the speed of non-threaded perls).
156 - slightly modernise Coro::Intro, add section about rouse functions.
157 - avoid DEFSV and ERRSV, giving another 10% improvement
158 in thread switching.
159 - Coro::State->is_destroyed is now called is_zombie.
160 - implement a Coro->safe_cancel method that might fail, but
161 cancels in a "safer" way if it succeeds.
162 - add preliminary support for DEBUGGING perls.
163 - get rid of two hash-accesses when initialising a new Coro - this
164 speeds up coro creation by almost a factor of two.
165 - croak when a coro that is being cancelled tries to block
166 (e.g. while executing a guard block), instead of crashing or
167 deadlocking.
168 - use a more robust and also faster method to identify Coro::State
169 objects - speeds up everything a bit.
170 - implement Coro->cancel in XS for a 20% speed improvement, and to
171 be able to implement mutual cancellation.
172 - speed up context switches by a percent or two by more efficiently
173 allocating context stack entries.
174 - implement Coro->join and Coro->on_destroy in XS for a speedup and
175 a reduction in memory use.
176 - cancelling a coro while it itself is cancelling another coro is
177 now supported and working, instead of triggering an assertion.
178 - be a bit more crash-resistant when calling (buggy) on_destroy
179 callbacks (best effort).
180 - move on_destroy into the slf_frame, to allow extension slf
181 functions to have destructors.
182 - get rid if coro refcounting - simply crash in other interpreter
183 threads by nulling the pointers on clone.
184 - simplify warn/die hook handling when loading Coro - the convoluted
185 logic seems to be no longer neccessary.
186 - use libecb instead of our own home-grown gcc hacks.
187 - document alternatives to Coro::LWP. Please use them :)
188 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
189 sigaltstack. Really. wine suffers from it, erlang suffers from it,
190 and it's known since at least 2006.
191
1925.372 Wed Feb 23 06:14:30 CET 2011
193 - apparently mingw doesn't provide a working gettimeofday, try to
194 work around that by relying on Time::HiRes (indirectly brought to
195 my attention by Max Maischein).
196 - fix some portability issues when Time::HiRes was used.
197
1985.371 Mon Feb 21 14:36:08 CET 2011
199 - backport to windows process emulation code again.
200
2015.37 Sat Feb 19 07:49:44 CET 2011
202 - add a big "Coro thread life cycle" section to "man Coro".
203 - try a tentative workaround against the breakage that 5.13 has
204 introduced without depreciation period. sigh.
205 - no longer use Time::HiRes if gettimeofday is available, which
206 saves quite a lot of memory.
207
2085.36 Sun Feb 13 05:33:41 CET 2011
209 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
210 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
211 method call.
212 - undocument Coro::Timer::sleep and obsolete whole module.
213 - optimise Coro::Timer::timeout memory and cpu usage.
214 - slightly updated Coro::Intro for recent changes.
215 - do not initialise PL_dirty anymore.
216
2175.25 Thu Nov 11 01:08:39 CET 2010
218 - try a different approach on netbsd - netbsd 5 finally has marginally
219 working pthreads, but still broken ucontext/sigaltstack.
220 - openbsd 4.8 finally got their act together, Coro works out of the box
221 with asm, setjmp and pthreads (no change, just informational).
222
2235.24 Sat Oct 23 11:27:12 CEST 2010
224 - port to the EV 4.0 API.
225 - work around bugs in mingw32, making strawberry perl work
226 out of the box.
227 - correctly modify Coro::AIO function prototypes
228 so that they reflect the "no optional parameters" rule.
229 - "ported" libcoro to C++.
8 230
95.23 Mon May 17 18:50:42 CEST 2010 2315.23 Mon May 17 18:50:42 CEST 2010
10 - be more resistant to ordering changes when initialising 232 - be more resistant to ordering changes when initialising
11 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias 233 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
12 Waldorf). 234 Waldorf).
… …
291 - ->throw is now supported on Coro::State objects. 513 - ->throw is now supported on Coro::State objects.
292 - clean up cctx creation code a bit. 514 - clean up cctx creation code a bit.
293 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 515 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
294 516
2954.9 Sat Nov 8 17:45:27 CET 2008 5174.9 Sat Nov 8 17:45:27 CET 2008
296 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 518 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
297 - (libcoro) no longer leak threads in the experimental pthread backend, 519 - (libcoro) no longer leak threads in the experimental pthread backend,
298 also speed it up considerably. 520 also speed it up considerably.
299 - (libcoro) do not rely on makecontext passing void *'s unscathed. 521 - (libcoro) do not rely on makecontext passing void *'s unscathed.
300 - fix compiletime dependencies on libcoro in the Makefile. 522 - fix compiletime dependencies on libcoro in the Makefile.
301 - cctx_count wasn't always updated properly. 523 - cctx_count wasn't always updated properly.
… …
363 - remove debugging code related to MgPV_nolen_const, also try to 585 - remove debugging code related to MgPV_nolen_const, also try to
364 make it compile with perl 5.8.6 (yes, apple apparently loves 586 make it compile with perl 5.8.6 (yes, apple apparently loves
365 outdated software). Reported by John S. 587 outdated software). Reported by John S.
366 588
3674.744 Tue Jul 8 22:06:35 CEST 2008 5894.744 Tue Jul 8 22:06:35 CEST 2008
368 - correctly provide default DIE/WARN handlers as documented. 590 - correctly provide default DIE/WARN handlers as documented.
369 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 591 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
370 implementations inside perl work fine for us. 592 implementations inside perl work fine for us.
371 593
3724.743 Mon Jun 16 00:21:57 CEST 2008 5944.743 Mon Jun 16 00:21:57 CEST 2008
373 - when using Coro::EV without running EV::loop it could 595 - when using Coro::EV without running EV::loop it could
… …
395 - sprinkle "no warnings" freely over everything, also suppress 617 - sprinkle "no warnings" freely over everything, also suppress
396 warnings for some other modules. 618 warnings for some other modules.
397 - fix typo in WSAEWOULDBLOCK. 619 - fix typo in WSAEWOULDBLOCK.
398 620
3994.72 Sun May 25 05:14:36 CEST 2008 6214.72 Sun May 25 05:14:36 CEST 2008
400 - tweak META.yaml a bit, unfortunately, there is no documented way 622 - tweak META.yaml a bit, unfortunately, there is no documented way
401 to have optional dependencies with CPAN. doh :( 623 to have optional dependencies with CPAN. doh :(
402 - avoid running some tests on windows because they would fail due to 624 - avoid running some tests on windows because they would fail due to
403 perl bug (broken fork, broken pipes...). 625 perl bug (broken fork, broken pipes...).
404 - work around perl on windows bugs where perl returns undocumented 626 - work around perl on windows bugs where perl returns undocumented
405 error codes for sysread, syswrite etc. by taking advantage 627 error codes for sysread, syswrite etc. by taking advantage
… …
409 - use unix domain sockets in testsuite to work around 631 - use unix domain sockets in testsuite to work around
410 common perl implementation bugs on widows (they are emulated by 632 common perl implementation bugs on widows (they are emulated by
411 tcp sockets on windows. ugh.) 633 tcp sockets on windows. ugh.)
412 634
4134.71 Sat May 24 20:01:27 CEST 2008 6354.71 Sat May 24 20:01:27 CEST 2008
414 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 636 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
415 - take advantage of async name resolution of AnyEvent::Util. 637 - take advantage of async name resolution of AnyEvent::Util.
416 - work around brutal inet_aton override in Coro::LWP. 638 - work around brutal inet_aton override in Coro::LWP.
417 - take advantage of the readyhook in Coro::EV, for smoother 639 - take advantage of the readyhook in Coro::EV, for smoother
418 scheduling. 640 scheduling.
419 641
4204.7 Sun May 11 00:32:19 CEST 2008 6424.7 Sun May 11 00:32:19 CEST 2008
421 - completely reworked the Coro manpage. 643 - completely reworked the Coro manpage.
422 - added Coro::AnyEvent, generic event loop integration. 644 - added Coro::AnyEvent, generic event loop integration.
423 - implement cancel, ready and kill commands in Coro::Debug. 645 - implement cancel, ready and kill commands in Coro::Debug.
424 - document find_coro in Coro::Debug. 646 - document find_coro in Coro::Debug.
425 - incompatible change: rename has_stack to has_cctx. 647 - incompatible change: rename has_stack to has_cctx.
426 - Coro::AIO and Coro::BDB no longer force event model detection, 648 - Coro::AIO and Coro::BDB no longer force event model detection,
427 use AnyEvent::AIO and AnyEvent::BDB. 649 use AnyEvent::AIO and AnyEvent::BDB.
… …
551 773
5524.11 Thu Oct 11 02:40:24 CEST 2007 7744.11 Thu Oct 11 02:40:24 CEST 2007
553 - port to threaded perls. 775 - port to threaded perls.
554 776
5554.1 Thu Oct 11 02:38:16 CEST 2007 7774.1 Thu Oct 11 02:38:16 CEST 2007
556 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 778 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
557 be local to each coro (see Coro::State). 779 be local to each coro (see Coro::State).
558 - incompatible change: for very deep reasons, cede and cede_notself 780 - incompatible change: for very deep reasons, cede and cede_notself
559 cannot return anything, so nothing will be returned. 781 cannot return anything, so nothing will be returned.
560 - possibly bring back 5.10 compatibility (untested). 782 - possibly bring back 5.10 compatibility (untested).
561 - work around stupid (and wrong) warning on 5.10 :(. 783 - work around stupid (and wrong) warning on 5.10 :(.
… …
633 but stateful semantics. 855 but stateful semantics.
634 - fixed a lot of typos in Coro.pm (patch submitted by David 856 - fixed a lot of typos in Coro.pm (patch submitted by David
635 Steinbrunner, which applied flawlessly). 857 Steinbrunner, which applied flawlessly).
636 858
6373.6 Sat Apr 14 17:13:31 CEST 2007 8593.6 Sat Apr 14 17:13:31 CEST 2007
638 - added some bugfixes to get eg/myhttpd working again. 860 - added some bugfixes to get eg/myhttpd working again.
639 - added Coro::Storable for often-cede'ing freeze/thaw. 861 - added Coro::Storable for often-cede'ing freeze/thaw.
640 - try to do a clean exit when a coroutine calls exit 862 - try to do a clean exit when a coroutine calls exit
641 (EXPERIMENTAL). 863 (EXPERIMENTAL).
642 - got rid of indirect call through _coro_init. 864 - got rid of indirect call through _coro_init.
643 - updated the partly antique examples in eg/ to 865 - updated the partly antique examples in eg/ to

Diff Legend

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