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

Comparing Coro/Changes (file contents):
Revision 1.552 by root, Mon Feb 21 13:38:02 2011 UTC vs.
Revision 1.660 by root, Sat Jun 25 19:04:04 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
96.51 Sat Jun 25 20:57:20 CEST 2016
10 - 6.5 release didn't compile on most older perls due to
11 a logic error enabling experimental 5.24 code for all versions.
12 - since various broken or deficient patches float around
13 to work around the vtbl API breakage in perl, let's provide
14 our own workaround, which seems to be much less invasive
15 and more compatible than the approaches seen so far. YMMV.
16 - work around assertion failure in perl_destruct on 5.24
17 (analyzed by Dave Mitchell).
18
196.5 Wed Jun 22 22:23:50 CEST 2016
20 - swap_sv swapping order was not symmetrical, causing
21 wrong swaps when swap_sv was used multiple times on the
22 same sv in the same thread.
23 - swap_sv calls can now be undone by calling it again
24 with the same variables.
25 - swap_sv calls will now be undone in async_pool threads.
26 - split Coro::Semaphore::up/adjust into separate xs functions
27 for better error reporting, at slight codesize increase.
28 - (libcoro) arm assembly support, please test and report.
29 - adjust to PL_savestack changes in perl 5.24 (adapted from
30 the debian patch, which unfortunately gets it wrong).
31
326.49 Sat Oct 17 01:40:12 CEST 2015
33 - throwing an exception to a thread waiting in
34 Coro::Handle using Coro::EV did not stop the watchers,
35 causing the next call to fail (testcase by Martin Pritchard).
36 - bump minimum perl version to 5.10.
37
386.48 Sun Oct 4 19:03:51 CEST 2015
39 - fix memory leak when Coro::Handle uses Coro::EV internally
40 (testcase by Sten Sten).
41 - update code to libev 4 API (internally, Coro still used the
42 version 3 API calls).
43 - remove unused hv_sig.
44
456.47 Sat Jul 11 03:58:20 CEST 2015
46 - swap_svs - NVs can be in the SV head beginning with 5.20.
47
486.46 Tue Jun 30 14:36:30 CEST 2015
49 - restore portability to perls with windows fork emulation
50 (patch by Petr Písař).
51
526.45 Tue Jun 30 01:40:08 CEST 2015
53 - restore portability to perl 5.16 and below.
54
556.44 Tue Jun 30 00:41:54 CEST 2015
56 - avoid segfaulting (or worse) when tracing a canceled thread.
57 - the Event module silently broke it's public hook API, causing
58 failures on perls compiled with -Duselongdouble. Requiring
59 the latest version of Event because I don't know which release
60 changed this (it's not mentioned in the ChangeLog).
61 - allow xs level enterleave hooks via CoroAPI.
62
636.43 Thu Jun 4 15:38:14 CEST 2015
64 - use stability canary.
65 - port to stableperl-5.22.0-1.001.
66 - update libecb.
67
686.42 Wed Feb 11 20:29:52 CET 2015
69 - Coro::SemaphoreSet->try did not actually work (analyzed by
70 SATO Kentaro).
71 - upgrade libecb to be C11 compliant.
72
736.41 Sat Sep 6 22:08:46 CEST 2014
74 - restore portability to perl 5.8.x.
75 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
76 up some memory, but fixes "use feature" and similar modules
77 when used inside a Coro.
78 - allow zero as argument to Coro::Channel to mean the same thing
79 as no argument. this works with older versions as well,
80 but wasn't legal until now.
81 - slightly better c header file detection.
82
836.39 Mon Jun 2 00:00:08 CEST 2014
84 - work around more incompatible changes in 5.20.
85
866.38 Sun Jun 1 21:54:23 CEST 2014
87 - check that perl slots actually have enough space to hold
88 interpreter variables.
89 - untested port to perl 5.19 (64 bit tmps indices) (reported
90 by Andreas König).
91 - croak when cancel is called without a thread context.
92
936.37 Tue Mar 4 13:27:33 CET 2014
94 - *sigh*, removed leftover debugging code from debugging a
95 perl bug, of all things.
96
976.36 Tue Mar 4 07:11:59 CET 2014
98 - semaphores would not clear the destroy hook when interrupted
99 by ->throw, causing segfaults or worse.
100 - ->throw on a thread waiting for a semaphore did not acquire
101 the semaphore, but also didn't wake up other waiters,
102 possibly causing a deadlock.
103 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
104 to report a full stacktrace, which should help find out
105 where the actual call is.
106 - "atomically" destroy data in slf_destroy, because it is
107 the right thing to do, just in case.
108 - disable ecb.h libm dependency, to compile on ancient systems
109 or under adverse conditions.
110
1116.33 Mon Nov 18 11:26:27 CET 2013
112 - do not crash when freeing padlists with holes (in 5.18).
113 - tentative SVt_BIND 5.19 port/fix.
114
1156.32 Tue Nov 5 15:35:35 CET 2013
116 - use a new algorithm to derive padlists for perl 5.18. The old
117 one could lead to 0-pointer accesses inside perl (reported
118 by Darin McBride).
119
1206.31 Thu May 9 07:39:48 CEST 2013
121 - Coro::AIO requests would crash if the thread was ready'd
122 while the request was ongoing.
123
1246.29 Wed May 8 02:55:18 CEST 2013
125 - when an on_destroy handler destructs the coro currently being
126 destructed a perl scalar could be accessed after being freed,
127 likely causing a crash.
128
1296.28 Wed Mar 6 06:58:02 CET 2013
130 - clean remnants of existing __DIE__ and __WARN__ handlers so
131 they lose their magic and will not cause segfaults later
132 (testcase by Andrey Sagulin).
133 - improved Coro::State documentation a bit.
134 - Coro::Debug::command now flushes the output.
135 - add hack detection code for x32 abi, because the braindead slugs
136 who designed that made it look exactly like x86_64 without
137 providing proper compile time symbols to test for it. as a result,
138 this detection cannot work reliably.
139 - valgrind stack registering was broken.
140 - do not rely on Time::HiRes anymore in Coro::Debug.
141
1426.23 Fri Dec 7 23:36:37 CET 2012
143 - use experimental fiber implementation on native windows
144 perls.
145 - use sizeof (void *) as multiplication factor for stack sizes,
146 to accomodate the totally braindamaged microsoft 64 bit "os".
147 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
148 activeperl 5.16 crashes when PerlIO_define_layer is called due
149 to some bug in the perl dll, strawberry perl at least passes
150 the testsuite.
151 - implement Coro::Handle->peeraddr/host/port, for slightly
152 improved compatibility with LWP.
153 - implement 5.17 compatibility by almost blindly applying a
154 good-looking patch by Father Chrysostomos.
155 - move stack management functions into libcoro 3.
156 - libcoro version 3 "released".
157 - support magic values as timed_io_once args.
158 - recommend AnyEvent 7+ or EV 4+, also require EV
159 version 4 or newer for Coro::EV.
160
1616.10 Tue Oct 9 01:14:27 CEST 2012
162 - updated ecb.h, it had a typo that caused it to not compile on many
163 big endian systems (reported by many people).
164 - disable memory fences in ecb.h to improve portability.
165
1666.09 Sat Oct 6 23:25:02 CEST 2012
167 - Coro::EV I/O watchers were not interruptible by exceptions
168 (Coro::State::throw) (testcase by sten).
169 - ->throw now puts threads into the ready queue, as this seems to
170 be expected by existing code, and code that doesn't cope with spurious
171 wakeups needs fixing anyway.
172 - use fd -1 in mmap.
173 - cast I32 to int in error message printf.
174 - warn about broken so-called "hardened" kernels.
175
1766.08 Fri Apr 13 12:05:47 CEST 2012
177 - be more aggressive about exiting like perl does - formerly,
178 exiting from the non-main thread would not execute END blocks.
179
1806.07 Fri Nov 11 21:21:48 CET 2011
181 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
182 - update ecb.h.
183
1846.06 Mon Aug 8 23:59:48 CEST 2011
185 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
186 backend, which is an order of magnitude slower. unfortunately, cygwins
187 pthread implementation isn't very complete either, so allocate the stack
188 twice just to be sure.
189 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
190 in its ithreaded perl - best recompile cygwin and use the 'w'indows
191 backend for much better performance. also disable ithreads for
192 even better performance...).
193
1946.05 Thu Aug 4 21:36:36 CEST 2011
195 - blush, condvar values would not be propagated from send to recv anymore
196 (reported by Chip Salzenberg).
197 - use exponential increase for the readline buffer length in
198 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
199
2006.04 Wed Aug 3 17:15:45 CEST 2011
201 - use even more efficient and more compatible condvars for
202 compatibility to AnyEvent 6.x :)
203 - more inconsequential ecb.h updates.
204
2056.03 Wed Aug 3 11:41:30 CEST 2011
206 - change how Coro patches AnyEvent condvars for compatibility to
207 AnyEvent 6.x.
208 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
209 hasn't been included.
210
2116.02 Wed Jul 13 04:35:19 CEST 2011
212 - "improve portability to Gentoo" - gentoo manages to put perl variables
213 in memory areas that are farther than 2gb apart, which the jit couldn't
214 handle and barfed. now it's just a bit slower on gentoo and similar
215 systems.
216
2176.01 Sun Jul 3 12:31:14 CEST 2011
218 - workarounds are good, but the test for whether pthreads are used
219 was not good. this one should be better.
220 - check differently whether gcc generates cfi instructions itself.
221
2226.0 Wed Jun 29 19:43:35 CEST 2011
223 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
224 destroyed and cleaned up automatically (e.g. async { schedule }).
225 - implement a JIT compiler for part of the thread switch code,
226 which gives a 50% speed improvement on threaded perls, and
227 about 4% on non-threaded perls (so threaded perls now finally
228 reach about half the speed of non-threaded perls).
229 - slightly modernise Coro::Intro, add section about rouse functions.
230 - avoid DEFSV and ERRSV, giving another 10% improvement
231 in thread switching.
232 - Coro::State->is_destroyed is now called is_zombie.
233 - implement a Coro->safe_cancel method that might fail, but
234 cancels in a "safer" way if it succeeds.
235 - add preliminary support for DEBUGGING perls.
236 - get rid of two hash-accesses when initialising a new Coro - this
237 speeds up coro creation by almost a factor of two.
238 - croak when a coro that is being cancelled tries to block
239 (e.g. while executing a guard block), instead of crashing or
240 deadlocking.
241 - use a more robust and also faster method to identify Coro::State
242 objects - speeds up everything a bit.
243 - implement Coro->cancel in XS for a 20% speed improvement, and to
244 be able to implement mutual cancellation.
245 - speed up context switches by a percent or two by more efficiently
246 allocating context stack entries.
247 - implement Coro->join and Coro->on_destroy in XS for a speedup and
248 a reduction in memory use.
249 - cancelling a coro while it itself is cancelling another coro is
250 now supported and working, instead of triggering an assertion.
251 - be a bit more crash-resistant when calling (buggy) on_destroy
252 callbacks (best effort).
253 - move on_destroy into the slf_frame, to allow extension slf
254 functions to have destructors.
255 - get rid if coro refcounting - simply crash in other interpreter
256 threads by nulling the pointers on clone.
257 - simplify warn/die hook handling when loading Coro - the convoluted
258 logic seems to be no longer neccessary.
259 - use libecb instead of our own home-grown gcc hacks.
260 - document alternatives to Coro::LWP. Please use them :)
261 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
262 sigaltstack. Really. wine suffers from it, erlang suffers from it,
263 and it's known since at least 2006.
264
2655.372 Wed Feb 23 06:14:30 CET 2011
266 - apparently mingw doesn't provide a working gettimeofday, try to
267 work around that by relying on Time::HiRes (indirectly brought to
268 my attention by Max Maischein).
269 - fix some portability issues when Time::HiRes was used.
6 270
75.371 Mon Feb 21 14:36:08 CET 2011 2715.371 Mon Feb 21 14:36:08 CET 2011
8 - backport to windows process emulation code again. 272 - backport to windows process emulation code again.
9 273
105.37 Sat Feb 19 07:49:44 CET 2011 2745.37 Sat Feb 19 07:49:44 CET 2011
… …
322 - ->throw is now supported on Coro::State objects. 586 - ->throw is now supported on Coro::State objects.
323 - clean up cctx creation code a bit. 587 - clean up cctx creation code a bit.
324 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 588 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
325 589
3264.9 Sat Nov 8 17:45:27 CET 2008 5904.9 Sat Nov 8 17:45:27 CET 2008
327 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 591 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
328 - (libcoro) no longer leak threads in the experimental pthread backend, 592 - (libcoro) no longer leak threads in the experimental pthread backend,
329 also speed it up considerably. 593 also speed it up considerably.
330 - (libcoro) do not rely on makecontext passing void *'s unscathed. 594 - (libcoro) do not rely on makecontext passing void *'s unscathed.
331 - fix compiletime dependencies on libcoro in the Makefile. 595 - fix compiletime dependencies on libcoro in the Makefile.
332 - cctx_count wasn't always updated properly. 596 - cctx_count wasn't always updated properly.
… …
394 - remove debugging code related to MgPV_nolen_const, also try to 658 - remove debugging code related to MgPV_nolen_const, also try to
395 make it compile with perl 5.8.6 (yes, apple apparently loves 659 make it compile with perl 5.8.6 (yes, apple apparently loves
396 outdated software). Reported by John S. 660 outdated software). Reported by John S.
397 661
3984.744 Tue Jul 8 22:06:35 CEST 2008 6624.744 Tue Jul 8 22:06:35 CEST 2008
399 - correctly provide default DIE/WARN handlers as documented. 663 - correctly provide default DIE/WARN handlers as documented.
400 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 664 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
401 implementations inside perl work fine for us. 665 implementations inside perl work fine for us.
402 666
4034.743 Mon Jun 16 00:21:57 CEST 2008 6674.743 Mon Jun 16 00:21:57 CEST 2008
404 - when using Coro::EV without running EV::loop it could 668 - when using Coro::EV without running EV::loop it could
… …
426 - sprinkle "no warnings" freely over everything, also suppress 690 - sprinkle "no warnings" freely over everything, also suppress
427 warnings for some other modules. 691 warnings for some other modules.
428 - fix typo in WSAEWOULDBLOCK. 692 - fix typo in WSAEWOULDBLOCK.
429 693
4304.72 Sun May 25 05:14:36 CEST 2008 6944.72 Sun May 25 05:14:36 CEST 2008
431 - tweak META.yaml a bit, unfortunately, there is no documented way 695 - tweak META.yaml a bit, unfortunately, there is no documented way
432 to have optional dependencies with CPAN. doh :( 696 to have optional dependencies with CPAN. doh :(
433 - avoid running some tests on windows because they would fail due to 697 - avoid running some tests on windows because they would fail due to
434 perl bug (broken fork, broken pipes...). 698 perl bug (broken fork, broken pipes...).
435 - work around perl on windows bugs where perl returns undocumented 699 - work around perl on windows bugs where perl returns undocumented
436 error codes for sysread, syswrite etc. by taking advantage 700 error codes for sysread, syswrite etc. by taking advantage
… …
440 - use unix domain sockets in testsuite to work around 704 - use unix domain sockets in testsuite to work around
441 common perl implementation bugs on widows (they are emulated by 705 common perl implementation bugs on widows (they are emulated by
442 tcp sockets on windows. ugh.) 706 tcp sockets on windows. ugh.)
443 707
4444.71 Sat May 24 20:01:27 CEST 2008 7084.71 Sat May 24 20:01:27 CEST 2008
445 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 709 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
446 - take advantage of async name resolution of AnyEvent::Util. 710 - take advantage of async name resolution of AnyEvent::Util.
447 - work around brutal inet_aton override in Coro::LWP. 711 - work around brutal inet_aton override in Coro::LWP.
448 - take advantage of the readyhook in Coro::EV, for smoother 712 - take advantage of the readyhook in Coro::EV, for smoother
449 scheduling. 713 scheduling.
450 714
4514.7 Sun May 11 00:32:19 CEST 2008 7154.7 Sun May 11 00:32:19 CEST 2008
452 - completely reworked the Coro manpage. 716 - completely reworked the Coro manpage.
453 - added Coro::AnyEvent, generic event loop integration. 717 - added Coro::AnyEvent, generic event loop integration.
454 - implement cancel, ready and kill commands in Coro::Debug. 718 - implement cancel, ready and kill commands in Coro::Debug.
455 - document find_coro in Coro::Debug. 719 - document find_coro in Coro::Debug.
456 - incompatible change: rename has_stack to has_cctx. 720 - incompatible change: rename has_stack to has_cctx.
457 - Coro::AIO and Coro::BDB no longer force event model detection, 721 - Coro::AIO and Coro::BDB no longer force event model detection,
458 use AnyEvent::AIO and AnyEvent::BDB. 722 use AnyEvent::AIO and AnyEvent::BDB.
… …
582 846
5834.11 Thu Oct 11 02:40:24 CEST 2007 8474.11 Thu Oct 11 02:40:24 CEST 2007
584 - port to threaded perls. 848 - port to threaded perls.
585 849
5864.1 Thu Oct 11 02:38:16 CEST 2007 8504.1 Thu Oct 11 02:38:16 CEST 2007
587 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 851 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
588 be local to each coro (see Coro::State). 852 be local to each coro (see Coro::State).
589 - incompatible change: for very deep reasons, cede and cede_notself 853 - incompatible change: for very deep reasons, cede and cede_notself
590 cannot return anything, so nothing will be returned. 854 cannot return anything, so nothing will be returned.
591 - possibly bring back 5.10 compatibility (untested). 855 - possibly bring back 5.10 compatibility (untested).
592 - work around stupid (and wrong) warning on 5.10 :(. 856 - work around stupid (and wrong) warning on 5.10 :(.
… …
664 but stateful semantics. 928 but stateful semantics.
665 - fixed a lot of typos in Coro.pm (patch submitted by David 929 - fixed a lot of typos in Coro.pm (patch submitted by David
666 Steinbrunner, which applied flawlessly). 930 Steinbrunner, which applied flawlessly).
667 931
6683.6 Sat Apr 14 17:13:31 CEST 2007 9323.6 Sat Apr 14 17:13:31 CEST 2007
669 - added some bugfixes to get eg/myhttpd working again. 933 - added some bugfixes to get eg/myhttpd working again.
670 - added Coro::Storable for often-cede'ing freeze/thaw. 934 - added Coro::Storable for often-cede'ing freeze/thaw.
671 - try to do a clean exit when a coroutine calls exit 935 - try to do a clean exit when a coroutine calls exit
672 (EXPERIMENTAL). 936 (EXPERIMENTAL).
673 - got rid of indirect call through _coro_init. 937 - got rid of indirect call through _coro_init.
674 - updated the partly antique examples in eg/ to 938 - updated the partly antique examples in eg/ to

Diff Legend

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