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

Comparing Coro/Changes (file contents):
Revision 1.558 by root, Fri May 6 18:01:13 2011 UTC vs.
Revision 1.667 by root, Fri Jul 14 03:02:44 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
9 - seems to pass test suite down till 5.8.9, so make this the
10 minimum requirement in Makefile.PL.
11 - experimental 5.24 compatibility changes (SUB_ARGARRAY).
12 - experimental 5.26 compatibility changes (PL_comppad -> PADOFFSET).
13 - disable FORTIFY_SOURCE to avoid broken (but well-intentioned)
14 fortified longjmp variants in some libcs.
15 - re-try JIT allocation without PROT_EXEC, to hopefully improve
16 portability (still doesn't seem to work around broken selinux
17 mmap everywhere).
18 - speling fixes (Xavier Guimard).
19
206.511 Sun Jun 26 23:44:50 CEST 2016
21 - make it compile with threaded perls again.
22 - simplify and speed up __DIE__ and __WARN__ handling.
23
246.51 Sat Jun 25 20:57:20 CEST 2016
25 - 6.5 release didn't compile on most older perls due to
26 a logic error enabling experimental 5.24 code for all versions.
27 - since various broken or deficient patches float around
28 to work around the vtbl API breakage in perl, let's provide
29 our own workaround, which seems to be much less invasive
30 and more compatible than the approaches seen so far. YMMV.
31 - work around assertion failure in perl_destruct on 5.24
32 (analyzed by Dave Mitchell).
33
346.5 Wed Jun 22 22:23:50 CEST 2016
35 - swap_sv swapping order was not symmetrical, causing
36 wrong swaps when swap_sv was used multiple times on the
37 same sv in the same thread.
38 - swap_sv calls can now be undone by calling it again
39 with the same variables.
40 - swap_sv calls will now be undone in async_pool threads.
41 - split Coro::Semaphore::up/adjust into separate xs functions
42 for better error reporting, at slight codesize increase.
43 - (libcoro) arm assembly support, please test and report.
44 - adjust to PL_savestack changes in perl 5.24 (adapted from
45 the debian patch, which unfortunately gets it wrong).
46
476.49 Sat Oct 17 01:40:12 CEST 2015
48 - throwing an exception to a thread waiting in
49 Coro::Handle using Coro::EV did not stop the watchers,
50 causing the next call to fail (testcase by Martin Pritchard).
51 - bump minimum perl version to 5.10.
52
536.48 Sun Oct 4 19:03:51 CEST 2015
54 - fix memory leak when Coro::Handle uses Coro::EV internally
55 (testcase by Sten Sten).
56 - update code to libev 4 API (internally, Coro still used the
57 version 3 API calls).
58 - remove unused hv_sig.
59
606.47 Sat Jul 11 03:58:20 CEST 2015
61 - swap_svs - NVs can be in the SV head beginning with 5.20.
62
636.46 Tue Jun 30 14:36:30 CEST 2015
64 - restore portability to perls with windows fork emulation
65 (patch by Petr Písař).
66
676.45 Tue Jun 30 01:40:08 CEST 2015
68 - restore portability to perl 5.16 and below.
69
706.44 Tue Jun 30 00:41:54 CEST 2015
71 - avoid segfaulting (or worse) when tracing a canceled thread.
72 - the Event module silently broke it's public hook API, causing
73 failures on perls compiled with -Duselongdouble. Requiring
74 the latest version of Event because I don't know which release
75 changed this (it's not mentioned in the ChangeLog).
76 - allow xs level enterleave hooks via CoroAPI.
77
786.43 Thu Jun 4 15:38:14 CEST 2015
79 - use stability canary.
80 - port to stableperl-5.22.0-1.001.
81 - update libecb.
82
836.42 Wed Feb 11 20:29:52 CET 2015
84 - Coro::SemaphoreSet->try did not actually work (analyzed by
85 SATO Kentaro).
86 - upgrade libecb to be C11 compliant.
87
886.41 Sat Sep 6 22:08:46 CEST 2014
89 - restore portability to perl 5.8.x.
90 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
91 up some memory, but fixes "use feature" and similar modules
92 when used inside a Coro.
93 - allow zero as argument to Coro::Channel to mean the same thing
94 as no argument. this works with older versions as well,
95 but wasn't legal until now.
96 - slightly better c header file detection.
97
986.39 Mon Jun 2 00:00:08 CEST 2014
99 - work around more incompatible changes in 5.20.
100
1016.38 Sun Jun 1 21:54:23 CEST 2014
102 - check that perl slots actually have enough space to hold
103 interpreter variables.
104 - untested port to perl 5.19 (64 bit tmps indices) (reported
105 by Andreas König).
106 - croak when cancel is called without a thread context.
107
1086.37 Tue Mar 4 13:27:33 CET 2014
109 - *sigh*, removed leftover debugging code from debugging a
110 perl bug, of all things.
111
1126.36 Tue Mar 4 07:11:59 CET 2014
113 - semaphores would not clear the destroy hook when interrupted
114 by ->throw, causing segfaults or worse.
115 - ->throw on a thread waiting for a semaphore did not acquire
116 the semaphore, but also didn't wake up other waiters,
117 possibly causing a deadlock.
118 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
119 to report a full stacktrace, which should help find out
120 where the actual call is.
121 - "atomically" destroy data in slf_destroy, because it is
122 the right thing to do, just in case.
123 - disable ecb.h libm dependency, to compile on ancient systems
124 or under adverse conditions.
125
1266.33 Mon Nov 18 11:26:27 CET 2013
127 - do not crash when freeing padlists with holes (in 5.18).
128 - tentative SVt_BIND 5.19 port/fix.
129
1306.32 Tue Nov 5 15:35:35 CET 2013
131 - use a new algorithm to derive padlists for perl 5.18. The old
132 one could lead to 0-pointer accesses inside perl (reported
133 by Darin McBride).
134
1356.31 Thu May 9 07:39:48 CEST 2013
136 - Coro::AIO requests would crash if the thread was ready'd
137 while the request was ongoing.
138
1396.29 Wed May 8 02:55:18 CEST 2013
140 - when an on_destroy handler destructs the coro currently being
141 destructed a perl scalar could be accessed after being freed,
142 likely causing a crash.
143
1446.28 Wed Mar 6 06:58:02 CET 2013
145 - clean remnants of existing __DIE__ and __WARN__ handlers so
146 they lose their magic and will not cause segfaults later
147 (testcase by Andrey Sagulin).
148 - improved Coro::State documentation a bit.
149 - Coro::Debug::command now flushes the output.
150 - add hack detection code for x32 abi, because the braindead slugs
151 who designed that made it look exactly like x86_64 without
152 providing proper compile time symbols to test for it. as a result,
153 this detection cannot work reliably.
154 - valgrind stack registering was broken.
155 - do not rely on Time::HiRes anymore in Coro::Debug.
156
1576.23 Fri Dec 7 23:36:37 CET 2012
158 - use experimental fiber implementation on native windows
159 perls.
160 - use sizeof (void *) as multiplication factor for stack sizes,
161 to accomodate the totally braindamaged microsoft 64 bit "os".
162 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
163 activeperl 5.16 crashes when PerlIO_define_layer is called due
164 to some bug in the perl dll, strawberry perl at least passes
165 the testsuite.
166 - implement Coro::Handle->peeraddr/host/port, for slightly
167 improved compatibility with LWP.
168 - implement 5.17 compatibility by almost blindly applying a
169 good-looking patch by Father Chrysostomos.
170 - move stack management functions into libcoro 3.
171 - libcoro version 3 "released".
172 - support magic values as timed_io_once args.
173 - recommend AnyEvent 7+ or EV 4+, also require EV
174 version 4 or newer for Coro::EV.
175
1766.10 Tue Oct 9 01:14:27 CEST 2012
177 - updated ecb.h, it had a typo that caused it to not compile on many
178 big endian systems (reported by many people).
179 - disable memory fences in ecb.h to improve portability.
180
1816.09 Sat Oct 6 23:25:02 CEST 2012
182 - Coro::EV I/O watchers were not interruptible by exceptions
183 (Coro::State::throw) (testcase by sten).
184 - ->throw now puts threads into the ready queue, as this seems to
185 be expected by existing code, and code that doesn't cope with spurious
186 wakeups needs fixing anyway.
187 - use fd -1 in mmap.
188 - cast I32 to int in error message printf.
189 - warn about broken so-called "hardened" kernels.
190
1916.08 Fri Apr 13 12:05:47 CEST 2012
192 - be more aggressive about exiting like perl does - formerly,
193 exiting from the non-main thread would not execute END blocks.
194
1956.07 Fri Nov 11 21:21:48 CET 2011
196 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
197 - update ecb.h.
198
1996.06 Mon Aug 8 23:59:48 CEST 2011
200 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
201 backend, which is an order of magnitude slower. unfortunately, cygwins
202 pthread implementation isn't very complete either, so allocate the stack
203 twice just to be sure.
204 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
205 in its ithreaded perl - best recompile cygwin and use the 'w'indows
206 backend for much better performance. also disable ithreads for
207 even better performance...).
208
2096.05 Thu Aug 4 21:36:36 CEST 2011
210 - blush, condvar values would not be propagated from send to recv anymore
211 (reported by Chip Salzenberg).
212 - use exponential increase for the readline buffer length in
213 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
214
2156.04 Wed Aug 3 17:15:45 CEST 2011
216 - use even more efficient and more compatible condvars for
217 compatibility to AnyEvent 6.x :)
218 - more inconsequential ecb.h updates.
219
2206.03 Wed Aug 3 11:41:30 CEST 2011
221 - change how Coro patches AnyEvent condvars for compatibility to
222 AnyEvent 6.x.
223 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
224 hasn't been included.
225
2266.02 Wed Jul 13 04:35:19 CEST 2011
227 - "improve portability to Gentoo" - gentoo manages to put perl variables
228 in memory areas that are farther than 2gb apart, which the jit couldn't
229 handle and barfed. now it's just a bit slower on gentoo and similar
230 systems.
231
2326.01 Sun Jul 3 12:31:14 CEST 2011
233 - workarounds are good, but the test for whether pthreads are used
234 was not good. this one should be better.
235 - check differently whether gcc generates cfi instructions itself.
236
2376.0 Wed Jun 29 19:43:35 CEST 2011
238 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
239 destroyed and cleaned up automatically (e.g. async { schedule }).
240 - implement a JIT compiler for part of the thread switch code,
241 which gives a 50% speed improvement on threaded perls, and
242 about 4% on non-threaded perls (so threaded perls now finally
243 reach about half the speed of non-threaded perls).
244 - slightly modernise Coro::Intro, add section about rouse functions.
245 - avoid DEFSV and ERRSV, giving another 10% improvement
246 in thread switching.
247 - Coro::State->is_destroyed is now called is_zombie.
248 - implement a Coro->safe_cancel method that might fail, but
249 cancels in a "safer" way if it succeeds.
250 - add preliminary support for DEBUGGING perls.
251 - get rid of two hash-accesses when initialising a new Coro - this
252 speeds up coro creation by almost a factor of two.
7 - croak when a coro being cancelled tries to block (e.g. while executing 253 - croak when a coro that is being cancelled tries to block
8 a guard block). 254 (e.g. while executing a guard block), instead of crashing or
255 deadlocking.
256 - use a more robust and also faster method to identify Coro::State
257 objects - speeds up everything a bit.
9 - implement Coro->cancel in XS for a 20% speed improvement, and to 258 - implement Coro->cancel in XS for a 20% speed improvement, and to
10 be able to implement mutual cancellation. 259 be able to implement mutual cancellation.
260 - speed up context switches by a percent or two by more efficiently
261 allocating context stack entries.
262 - implement Coro->join and Coro->on_destroy in XS for a speedup and
263 a reduction in memory use.
11 - cancelling a coro while it itself is cancelling another coro is 264 - cancelling a coro while it itself is cancelling another coro is
12 now supported and working, instead of triggering an assertion. 265 now supported and working, instead of triggering an assertion.
13 - be a bit more crash-resistant when calling (buggy) on_destroy 266 - be a bit more crash-resistant when calling (buggy) on_destroy
14 callbacks (best effort). 267 callbacks (best effort).
268 - move on_destroy into the slf_frame, to allow extension slf
269 functions to have destructors.
270 - get rid if coro refcounting - simply crash in other interpreter
271 threads by nulling the pointers on clone.
272 - simplify warn/die hook handling when loading Coro - the convoluted
273 logic seems to be no longer neccessary.
274 - use libecb instead of our own home-grown gcc hacks.
275 - document alternatives to Coro::LWP. Please use them :)
276 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
277 sigaltstack. Really. wine suffers from it, erlang suffers from it,
278 and it's known since at least 2006.
15 279
165.372 Wed Feb 23 06:14:30 CET 2011 2805.372 Wed Feb 23 06:14:30 CET 2011
17 - apparently mingw doesn't provide a working gettimeofday, try to 281 - apparently mingw doesn't provide a working gettimeofday, try to
18 work around that by relying on Time::HiRes (indirectly brought to 282 work around that by relying on Time::HiRes (indirectly brought to
19 my attention by Max Maischein). 283 my attention by Max Maischein).
… …
337 - ->throw is now supported on Coro::State objects. 601 - ->throw is now supported on Coro::State objects.
338 - clean up cctx creation code a bit. 602 - clean up cctx creation code a bit.
339 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 603 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
340 604
3414.9 Sat Nov 8 17:45:27 CET 2008 6054.9 Sat Nov 8 17:45:27 CET 2008
342 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 606 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
343 - (libcoro) no longer leak threads in the experimental pthread backend, 607 - (libcoro) no longer leak threads in the experimental pthread backend,
344 also speed it up considerably. 608 also speed it up considerably.
345 - (libcoro) do not rely on makecontext passing void *'s unscathed. 609 - (libcoro) do not rely on makecontext passing void *'s unscathed.
346 - fix compiletime dependencies on libcoro in the Makefile. 610 - fix compiletime dependencies on libcoro in the Makefile.
347 - cctx_count wasn't always updated properly. 611 - cctx_count wasn't always updated properly.
… …
409 - remove debugging code related to MgPV_nolen_const, also try to 673 - remove debugging code related to MgPV_nolen_const, also try to
410 make it compile with perl 5.8.6 (yes, apple apparently loves 674 make it compile with perl 5.8.6 (yes, apple apparently loves
411 outdated software). Reported by John S. 675 outdated software). Reported by John S.
412 676
4134.744 Tue Jul 8 22:06:35 CEST 2008 6774.744 Tue Jul 8 22:06:35 CEST 2008
414 - correctly provide default DIE/WARN handlers as documented. 678 - correctly provide default DIE/WARN handlers as documented.
415 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 679 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
416 implementations inside perl work fine for us. 680 implementations inside perl work fine for us.
417 681
4184.743 Mon Jun 16 00:21:57 CEST 2008 6824.743 Mon Jun 16 00:21:57 CEST 2008
419 - when using Coro::EV without running EV::loop it could 683 - when using Coro::EV without running EV::loop it could
… …
441 - sprinkle "no warnings" freely over everything, also suppress 705 - sprinkle "no warnings" freely over everything, also suppress
442 warnings for some other modules. 706 warnings for some other modules.
443 - fix typo in WSAEWOULDBLOCK. 707 - fix typo in WSAEWOULDBLOCK.
444 708
4454.72 Sun May 25 05:14:36 CEST 2008 7094.72 Sun May 25 05:14:36 CEST 2008
446 - tweak META.yaml a bit, unfortunately, there is no documented way 710 - tweak META.yaml a bit, unfortunately, there is no documented way
447 to have optional dependencies with CPAN. doh :( 711 to have optional dependencies with CPAN. doh :(
448 - avoid running some tests on windows because they would fail due to 712 - avoid running some tests on windows because they would fail due to
449 perl bug (broken fork, broken pipes...). 713 perl bug (broken fork, broken pipes...).
450 - work around perl on windows bugs where perl returns undocumented 714 - work around perl on windows bugs where perl returns undocumented
451 error codes for sysread, syswrite etc. by taking advantage 715 error codes for sysread, syswrite etc. by taking advantage
… …
455 - use unix domain sockets in testsuite to work around 719 - use unix domain sockets in testsuite to work around
456 common perl implementation bugs on widows (they are emulated by 720 common perl implementation bugs on widows (they are emulated by
457 tcp sockets on windows. ugh.) 721 tcp sockets on windows. ugh.)
458 722
4594.71 Sat May 24 20:01:27 CEST 2008 7234.71 Sat May 24 20:01:27 CEST 2008
460 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 724 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
461 - take advantage of async name resolution of AnyEvent::Util. 725 - take advantage of async name resolution of AnyEvent::Util.
462 - work around brutal inet_aton override in Coro::LWP. 726 - work around brutal inet_aton override in Coro::LWP.
463 - take advantage of the readyhook in Coro::EV, for smoother 727 - take advantage of the readyhook in Coro::EV, for smoother
464 scheduling. 728 scheduling.
465 729
4664.7 Sun May 11 00:32:19 CEST 2008 7304.7 Sun May 11 00:32:19 CEST 2008
467 - completely reworked the Coro manpage. 731 - completely reworked the Coro manpage.
468 - added Coro::AnyEvent, generic event loop integration. 732 - added Coro::AnyEvent, generic event loop integration.
469 - implement cancel, ready and kill commands in Coro::Debug. 733 - implement cancel, ready and kill commands in Coro::Debug.
470 - document find_coro in Coro::Debug. 734 - document find_coro in Coro::Debug.
471 - incompatible change: rename has_stack to has_cctx. 735 - incompatible change: rename has_stack to has_cctx.
472 - Coro::AIO and Coro::BDB no longer force event model detection, 736 - Coro::AIO and Coro::BDB no longer force event model detection,
473 use AnyEvent::AIO and AnyEvent::BDB. 737 use AnyEvent::AIO and AnyEvent::BDB.
… …
597 861
5984.11 Thu Oct 11 02:40:24 CEST 2007 8624.11 Thu Oct 11 02:40:24 CEST 2007
599 - port to threaded perls. 863 - port to threaded perls.
600 864
6014.1 Thu Oct 11 02:38:16 CEST 2007 8654.1 Thu Oct 11 02:38:16 CEST 2007
602 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 866 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
603 be local to each coro (see Coro::State). 867 be local to each coro (see Coro::State).
604 - incompatible change: for very deep reasons, cede and cede_notself 868 - incompatible change: for very deep reasons, cede and cede_notself
605 cannot return anything, so nothing will be returned. 869 cannot return anything, so nothing will be returned.
606 - possibly bring back 5.10 compatibility (untested). 870 - possibly bring back 5.10 compatibility (untested).
607 - work around stupid (and wrong) warning on 5.10 :(. 871 - work around stupid (and wrong) warning on 5.10 :(.
… …
679 but stateful semantics. 943 but stateful semantics.
680 - fixed a lot of typos in Coro.pm (patch submitted by David 944 - fixed a lot of typos in Coro.pm (patch submitted by David
681 Steinbrunner, which applied flawlessly). 945 Steinbrunner, which applied flawlessly).
682 946
6833.6 Sat Apr 14 17:13:31 CEST 2007 9473.6 Sat Apr 14 17:13:31 CEST 2007
684 - added some bugfixes to get eg/myhttpd working again. 948 - added some bugfixes to get eg/myhttpd working again.
685 - added Coro::Storable for often-cede'ing freeze/thaw. 949 - added Coro::Storable for often-cede'ing freeze/thaw.
686 - try to do a clean exit when a coroutine calls exit 950 - try to do a clean exit when a coroutine calls exit
687 (EXPERIMENTAL). 951 (EXPERIMENTAL).
688 - got rid of indirect call through _coro_init. 952 - got rid of indirect call through _coro_init.
689 - updated the partly antique examples in eg/ to 953 - updated the partly antique examples in eg/ to

Diff Legend

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