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

Comparing Coro/Changes (file contents):
Revision 1.557 by root, Sat Apr 30 05:17:42 2011 UTC vs.
Revision 1.670 by root, Fri Jul 14 23:20:07 2017 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?
6 8
96.513 Sat Jul 15 01:18:51 CEST 2017
10 - experimental 5.22 comaptibility changes (PadlistNAMES).
11 - removed attributes example and made sure the other examples
12 still work.
13 - fix some statevar types for newer perls, should be of no consequence.
14
156.512 Fri Jul 14 05:03:17 CEST 2017
16 - seems to pass test suite down till 5.8.9, so make this the
17 minimum requirement in Makefile.PL.
18 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
19 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
20 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
21 fortified longjmp variants in some libcs.
22 - re-try JIT allocation without PROT_EXEC, to hopefully improve
23 portability (still doesn't seem to work around broken selinux
24 mmap everywhere).
25 - speling fixes (Xavier Guimard).
26
276.511 Sun Jun 26 23:44:50 CEST 2016
28 - make it compile with threaded perls again.
29 - simplify and speed up __DIE__ and __WARN__ handling.
30
316.51 Sat Jun 25 20:57:20 CEST 2016
32 - 6.5 release didn't compile on most older perls due to
33 a logic error enabling experimental 5.24 code for all versions.
34 - since various broken or deficient patches float around
35 to work around the vtbl API breakage in perl, let's provide
36 our own workaround, which seems to be much less invasive
37 and more compatible than the approaches seen so far. YMMV.
38 - work around assertion failure in perl_destruct on 5.24
39 (analyzed by Dave Mitchell).
40
416.5 Wed Jun 22 22:23:50 CEST 2016
42 - swap_sv swapping order was not symmetrical, causing
43 wrong swaps when swap_sv was used multiple times on the
44 same sv in the same thread.
45 - swap_sv calls can now be undone by calling it again
46 with the same variables.
47 - swap_sv calls will now be undone in async_pool threads.
48 - split Coro::Semaphore::up/adjust into separate xs functions
49 for better error reporting, at slight codesize increase.
50 - (libcoro) arm assembly support, please test and report.
51 - adjust to PL_savestack changes in perl 5.24 (adapted from
52 the debian patch, which unfortunately gets it wrong).
53
546.49 Sat Oct 17 01:40:12 CEST 2015
55 - throwing an exception to a thread waiting in
56 Coro::Handle using Coro::EV did not stop the watchers,
57 causing the next call to fail (testcase by Martin Pritchard).
58 - bump minimum perl version to 5.10.
59
606.48 Sun Oct 4 19:03:51 CEST 2015
61 - fix memory leak when Coro::Handle uses Coro::EV internally
62 (testcase by Sten Sten).
63 - update code to libev 4 API (internally, Coro still used the
64 version 3 API calls).
65 - remove unused hv_sig.
66
676.47 Sat Jul 11 03:58:20 CEST 2015
68 - swap_svs - NVs can be in the SV head beginning with 5.20.
69
706.46 Tue Jun 30 14:36:30 CEST 2015
71 - restore portability to perls with windows fork emulation
72 (patch by Petr Písař).
73
746.45 Tue Jun 30 01:40:08 CEST 2015
75 - restore portability to perl 5.16 and below.
76
776.44 Tue Jun 30 00:41:54 CEST 2015
78 - avoid segfaulting (or worse) when tracing a canceled thread.
79 - the Event module silently broke it's public hook API, causing
80 failures on perls compiled with -Duselongdouble. Requiring
81 the latest version of Event because I don't know which release
82 changed this (it's not mentioned in the ChangeLog).
83 - allow xs level enterleave hooks via CoroAPI.
84
856.43 Thu Jun 4 15:38:14 CEST 2015
86 - use stability canary.
87 - port to stableperl-5.22.0-1.001.
88 - update libecb.
89
906.42 Wed Feb 11 20:29:52 CET 2015
91 - Coro::SemaphoreSet->try did not actually work (analyzed by
92 SATO Kentaro).
93 - upgrade libecb to be C11 compliant.
94
956.41 Sat Sep 6 22:08:46 CEST 2014
96 - restore portability to perl 5.8.x.
97 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
98 up some memory, but fixes "use feature" and similar modules
99 when used inside a Coro.
100 - allow zero as argument to Coro::Channel to mean the same thing
101 as no argument. this works with older versions as well,
102 but wasn't legal until now.
103 - slightly better c header file detection.
104
1056.39 Mon Jun 2 00:00:08 CEST 2014
106 - work around more incompatible changes in 5.20.
107
1086.38 Sun Jun 1 21:54:23 CEST 2014
109 - check that perl slots actually have enough space to hold
110 interpreter variables.
111 - untested port to perl 5.19 (64 bit tmps indices) (reported
112 by Andreas König).
113 - croak when cancel is called without a thread context.
114
1156.37 Tue Mar 4 13:27:33 CET 2014
116 - *sigh*, removed leftover debugging code from debugging a
117 perl bug, of all things.
118
1196.36 Tue Mar 4 07:11:59 CET 2014
120 - semaphores would not clear the destroy hook when interrupted
121 by ->throw, causing segfaults or worse.
122 - ->throw on a thread waiting for a semaphore did not acquire
123 the semaphore, but also didn't wake up other waiters,
124 possibly causing a deadlock.
125 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
126 to report a full stacktrace, which should help find out
127 where the actual call is.
128 - "atomically" destroy data in slf_destroy, because it is
129 the right thing to do, just in case.
130 - disable ecb.h libm dependency, to compile on ancient systems
131 or under adverse conditions.
132
1336.33 Mon Nov 18 11:26:27 CET 2013
134 - do not crash when freeing padlists with holes (in 5.18).
135 - tentative SVt_BIND 5.19 port/fix.
136
1376.32 Tue Nov 5 15:35:35 CET 2013
138 - use a new algorithm to derive padlists for perl 5.18. The old
139 one could lead to 0-pointer accesses inside perl (reported
140 by Darin McBride).
141
1426.31 Thu May 9 07:39:48 CEST 2013
143 - Coro::AIO requests would crash if the thread was ready'd
144 while the request was ongoing.
145
1466.29 Wed May 8 02:55:18 CEST 2013
147 - when an on_destroy handler destructs the coro currently being
148 destructed a perl scalar could be accessed after being freed,
149 likely causing a crash.
150
1516.28 Wed Mar 6 06:58:02 CET 2013
152 - clean remnants of existing __DIE__ and __WARN__ handlers so
153 they lose their magic and will not cause segfaults later
154 (testcase by Andrey Sagulin).
155 - improved Coro::State documentation a bit.
156 - Coro::Debug::command now flushes the output.
157 - add hack detection code for x32 abi, because the braindead slugs
158 who designed that made it look exactly like x86_64 without
159 providing proper compile time symbols to test for it. as a result,
160 this detection cannot work reliably.
161 - valgrind stack registering was broken.
162 - do not rely on Time::HiRes anymore in Coro::Debug.
163
1646.23 Fri Dec 7 23:36:37 CET 2012
165 - use experimental fiber implementation on native windows
166 perls.
167 - use sizeof (void *) as multiplication factor for stack sizes,
168 to accomodate the totally braindamaged microsoft 64 bit "os".
169 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
170 activeperl 5.16 crashes when PerlIO_define_layer is called due
171 to some bug in the perl dll, strawberry perl at least passes
172 the testsuite.
173 - implement Coro::Handle->peeraddr/host/port, for slightly
174 improved compatibility with LWP.
175 - implement 5.17 compatibility by almost blindly applying a
176 good-looking patch by Father Chrysostomos.
177 - move stack management functions into libcoro 3.
178 - libcoro version 3 "released".
179 - support magic values as timed_io_once args.
180 - recommend AnyEvent 7+ or EV 4+, also require EV
181 version 4 or newer for Coro::EV.
182
1836.10 Tue Oct 9 01:14:27 CEST 2012
184 - updated ecb.h, it had a typo that caused it to not compile on many
185 big endian systems (reported by many people).
186 - disable memory fences in ecb.h to improve portability.
187
1886.09 Sat Oct 6 23:25:02 CEST 2012
189 - Coro::EV I/O watchers were not interruptible by exceptions
190 (Coro::State::throw) (testcase by sten).
191 - ->throw now puts threads into the ready queue, as this seems to
192 be expected by existing code, and code that doesn't cope with spurious
193 wakeups needs fixing anyway.
194 - use fd -1 in mmap.
195 - cast I32 to int in error message printf.
196 - warn about broken so-called "hardened" kernels.
197
1986.08 Fri Apr 13 12:05:47 CEST 2012
199 - be more aggressive about exiting like perl does - formerly,
200 exiting from the non-main thread would not execute END blocks.
201
2026.07 Fri Nov 11 21:21:48 CET 2011
203 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
204 - update ecb.h.
205
2066.06 Mon Aug 8 23:59:48 CEST 2011
207 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
208 backend, which is an order of magnitude slower. unfortunately, cygwins
209 pthread implementation isn't very complete either, so allocate the stack
210 twice just to be sure.
211 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
212 in its ithreaded perl - best recompile cygwin and use the 'w'indows
213 backend for much better performance. also disable ithreads for
214 even better performance...).
215
2166.05 Thu Aug 4 21:36:36 CEST 2011
217 - blush, condvar values would not be propagated from send to recv anymore
218 (reported by Chip Salzenberg).
219 - use exponential increase for the readline buffer length in
220 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
221
2226.04 Wed Aug 3 17:15:45 CEST 2011
223 - use even more efficient and more compatible condvars for
224 compatibility to AnyEvent 6.x :)
225 - more inconsequential ecb.h updates.
226
2276.03 Wed Aug 3 11:41:30 CEST 2011
228 - change how Coro patches AnyEvent condvars for compatibility to
229 AnyEvent 6.x.
230 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
231 hasn't been included.
232
2336.02 Wed Jul 13 04:35:19 CEST 2011
234 - "improve portability to Gentoo" - gentoo manages to put perl variables
235 in memory areas that are farther than 2gb apart, which the jit couldn't
236 handle and barfed. now it's just a bit slower on gentoo and similar
237 systems.
238
2396.01 Sun Jul 3 12:31:14 CEST 2011
240 - workarounds are good, but the test for whether pthreads are used
241 was not good. this one should be better.
242 - check differently whether gcc generates cfi instructions itself.
243
2446.0 Wed Jun 29 19:43:35 CEST 2011
245 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
246 destroyed and cleaned up automatically (e.g. async { schedule }).
247 - implement a JIT compiler for part of the thread switch code,
248 which gives a 50% speed improvement on threaded perls, and
249 about 4% on non-threaded perls (so threaded perls now finally
250 reach about half the speed of non-threaded perls).
251 - slightly modernise Coro::Intro, add section about rouse functions.
252 - avoid DEFSV and ERRSV, giving another 10% improvement
253 in thread switching.
254 - Coro::State->is_destroyed is now called is_zombie.
255 - implement a Coro->safe_cancel method that might fail, but
256 cancels in a "safer" way if it succeeds.
257 - add preliminary support for DEBUGGING perls.
258 - get rid of two hash-accesses when initialising a new Coro - this
259 speeds up coro creation by almost a factor of two.
260 - croak when a coro that is being cancelled tries to block
261 (e.g. while executing a guard block), instead of crashing or
262 deadlocking.
263 - use a more robust and also faster method to identify Coro::State
264 objects - speeds up everything a bit.
7 - implement Coro->cancel in XS for a 20% speed improvement, and to 265 - implement Coro->cancel in XS for a 20% speed improvement, and to
8 be able to implement mutual cancellation. 266 be able to implement mutual cancellation.
267 - speed up context switches by a percent or two by more efficiently
268 allocating context stack entries.
269 - implement Coro->join and Coro->on_destroy in XS for a speedup and
270 a reduction in memory use.
9 - cancelling a coro while it itself is cancelling another coro is 271 - cancelling a coro while it itself is cancelling another coro is
10 now supported and working, instead of triggering an assertion. 272 now supported and working, instead of triggering an assertion.
11 - be a bit more crash-resistant when calling (buggy) on_destroy 273 - be a bit more crash-resistant when calling (buggy) on_destroy
12 callbacks (best effort). 274 callbacks (best effort).
275 - move on_destroy into the slf_frame, to allow extension slf
276 functions to have destructors.
277 - get rid if coro refcounting - simply crash in other interpreter
278 threads by nulling the pointers on clone.
279 - simplify warn/die hook handling when loading Coro - the convoluted
280 logic seems to be no longer neccessary.
281 - use libecb instead of our own home-grown gcc hacks.
282 - document alternatives to Coro::LWP. Please use them :)
283 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
284 sigaltstack. Really. wine suffers from it, erlang suffers from it,
285 and it's known since at least 2006.
13 286
145.372 Wed Feb 23 06:14:30 CET 2011 2875.372 Wed Feb 23 06:14:30 CET 2011
15 - apparently mingw doesn't provide a working gettimeofday, try to 288 - apparently mingw doesn't provide a working gettimeofday, try to
16 work around that by relying on Time::HiRes (indirectly brought to 289 work around that by relying on Time::HiRes (indirectly brought to
17 my attention by Max Maischein). 290 my attention by Max Maischein).
… …
335 - ->throw is now supported on Coro::State objects. 608 - ->throw is now supported on Coro::State objects.
336 - clean up cctx creation code a bit. 609 - clean up cctx creation code a bit.
337 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 610 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
338 611
3394.9 Sat Nov 8 17:45:27 CET 2008 6124.9 Sat Nov 8 17:45:27 CET 2008
340 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 613 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
341 - (libcoro) no longer leak threads in the experimental pthread backend, 614 - (libcoro) no longer leak threads in the experimental pthread backend,
342 also speed it up considerably. 615 also speed it up considerably.
343 - (libcoro) do not rely on makecontext passing void *'s unscathed. 616 - (libcoro) do not rely on makecontext passing void *'s unscathed.
344 - fix compiletime dependencies on libcoro in the Makefile. 617 - fix compiletime dependencies on libcoro in the Makefile.
345 - cctx_count wasn't always updated properly. 618 - cctx_count wasn't always updated properly.
… …
407 - remove debugging code related to MgPV_nolen_const, also try to 680 - remove debugging code related to MgPV_nolen_const, also try to
408 make it compile with perl 5.8.6 (yes, apple apparently loves 681 make it compile with perl 5.8.6 (yes, apple apparently loves
409 outdated software). Reported by John S. 682 outdated software). Reported by John S.
410 683
4114.744 Tue Jul 8 22:06:35 CEST 2008 6844.744 Tue Jul 8 22:06:35 CEST 2008
412 - correctly provide default DIE/WARN handlers as documented. 685 - correctly provide default DIE/WARN handlers as documented.
413 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 686 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
414 implementations inside perl work fine for us. 687 implementations inside perl work fine for us.
415 688
4164.743 Mon Jun 16 00:21:57 CEST 2008 6894.743 Mon Jun 16 00:21:57 CEST 2008
417 - when using Coro::EV without running EV::loop it could 690 - when using Coro::EV without running EV::loop it could
… …
439 - sprinkle "no warnings" freely over everything, also suppress 712 - sprinkle "no warnings" freely over everything, also suppress
440 warnings for some other modules. 713 warnings for some other modules.
441 - fix typo in WSAEWOULDBLOCK. 714 - fix typo in WSAEWOULDBLOCK.
442 715
4434.72 Sun May 25 05:14:36 CEST 2008 7164.72 Sun May 25 05:14:36 CEST 2008
444 - tweak META.yaml a bit, unfortunately, there is no documented way 717 - tweak META.yaml a bit, unfortunately, there is no documented way
445 to have optional dependencies with CPAN. doh :( 718 to have optional dependencies with CPAN. doh :(
446 - avoid running some tests on windows because they would fail due to 719 - avoid running some tests on windows because they would fail due to
447 perl bug (broken fork, broken pipes...). 720 perl bug (broken fork, broken pipes...).
448 - work around perl on windows bugs where perl returns undocumented 721 - work around perl on windows bugs where perl returns undocumented
449 error codes for sysread, syswrite etc. by taking advantage 722 error codes for sysread, syswrite etc. by taking advantage
… …
453 - use unix domain sockets in testsuite to work around 726 - use unix domain sockets in testsuite to work around
454 common perl implementation bugs on widows (they are emulated by 727 common perl implementation bugs on widows (they are emulated by
455 tcp sockets on windows. ugh.) 728 tcp sockets on windows. ugh.)
456 729
4574.71 Sat May 24 20:01:27 CEST 2008 7304.71 Sat May 24 20:01:27 CEST 2008
458 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 731 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
459 - take advantage of async name resolution of AnyEvent::Util. 732 - take advantage of async name resolution of AnyEvent::Util.
460 - work around brutal inet_aton override in Coro::LWP. 733 - work around brutal inet_aton override in Coro::LWP.
461 - take advantage of the readyhook in Coro::EV, for smoother 734 - take advantage of the readyhook in Coro::EV, for smoother
462 scheduling. 735 scheduling.
463 736
4644.7 Sun May 11 00:32:19 CEST 2008 7374.7 Sun May 11 00:32:19 CEST 2008
465 - completely reworked the Coro manpage. 738 - completely reworked the Coro manpage.
466 - added Coro::AnyEvent, generic event loop integration. 739 - added Coro::AnyEvent, generic event loop integration.
467 - implement cancel, ready and kill commands in Coro::Debug. 740 - implement cancel, ready and kill commands in Coro::Debug.
468 - document find_coro in Coro::Debug. 741 - document find_coro in Coro::Debug.
469 - incompatible change: rename has_stack to has_cctx. 742 - incompatible change: rename has_stack to has_cctx.
470 - Coro::AIO and Coro::BDB no longer force event model detection, 743 - Coro::AIO and Coro::BDB no longer force event model detection,
471 use AnyEvent::AIO and AnyEvent::BDB. 744 use AnyEvent::AIO and AnyEvent::BDB.
… …
595 868
5964.11 Thu Oct 11 02:40:24 CEST 2007 8694.11 Thu Oct 11 02:40:24 CEST 2007
597 - port to threaded perls. 870 - port to threaded perls.
598 871
5994.1 Thu Oct 11 02:38:16 CEST 2007 8724.1 Thu Oct 11 02:38:16 CEST 2007
600 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 873 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
601 be local to each coro (see Coro::State). 874 be local to each coro (see Coro::State).
602 - incompatible change: for very deep reasons, cede and cede_notself 875 - incompatible change: for very deep reasons, cede and cede_notself
603 cannot return anything, so nothing will be returned. 876 cannot return anything, so nothing will be returned.
604 - possibly bring back 5.10 compatibility (untested). 877 - possibly bring back 5.10 compatibility (untested).
605 - work around stupid (and wrong) warning on 5.10 :(. 878 - work around stupid (and wrong) warning on 5.10 :(.
… …
677 but stateful semantics. 950 but stateful semantics.
678 - fixed a lot of typos in Coro.pm (patch submitted by David 951 - fixed a lot of typos in Coro.pm (patch submitted by David
679 Steinbrunner, which applied flawlessly). 952 Steinbrunner, which applied flawlessly).
680 953
6813.6 Sat Apr 14 17:13:31 CEST 2007 9543.6 Sat Apr 14 17:13:31 CEST 2007
682 - added some bugfixes to get eg/myhttpd working again. 955 - added some bugfixes to get eg/myhttpd working again.
683 - added Coro::Storable for often-cede'ing freeze/thaw. 956 - added Coro::Storable for often-cede'ing freeze/thaw.
684 - try to do a clean exit when a coroutine calls exit 957 - try to do a clean exit when a coroutine calls exit
685 (EXPERIMENTAL). 958 (EXPERIMENTAL).
686 - got rid of indirect call through _coro_init. 959 - got rid of indirect call through _coro_init.
687 - updated the partly antique examples in eg/ to 960 - updated the partly antique examples in eg/ to

Diff Legend

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