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

Comparing Coro/Changes (file contents):
Revision 1.542 by root, Sat Oct 23 09:28:48 2010 UTC vs.
Revision 1.665 by root, Fri Jul 14 02:51:11 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: 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?
7TODO: croak when async_pool tries to run canceled thread?
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 - re-try JIT allocation without PROT_EXEC, to hopefully improve
14 portability.
15
166.511 Sun Jun 26 23:44:50 CEST 2016
17 - make it compile with threaded perls again.
18 - simplify and speed up __DIE__ and __WARN__ handling.
19
206.51 Sat Jun 25 20:57:20 CEST 2016
21 - 6.5 release didn't compile on most older perls due to
22 a logic error enabling experimental 5.24 code for all versions.
23 - since various broken or deficient patches float around
24 to work around the vtbl API breakage in perl, let's provide
25 our own workaround, which seems to be much less invasive
26 and more compatible than the approaches seen so far. YMMV.
27 - work around assertion failure in perl_destruct on 5.24
28 (analyzed by Dave Mitchell).
29
306.5 Wed Jun 22 22:23:50 CEST 2016
31 - swap_sv swapping order was not symmetrical, causing
32 wrong swaps when swap_sv was used multiple times on the
33 same sv in the same thread.
34 - swap_sv calls can now be undone by calling it again
35 with the same variables.
36 - swap_sv calls will now be undone in async_pool threads.
37 - split Coro::Semaphore::up/adjust into separate xs functions
38 for better error reporting, at slight codesize increase.
39 - (libcoro) arm assembly support, please test and report.
40 - adjust to PL_savestack changes in perl 5.24 (adapted from
41 the debian patch, which unfortunately gets it wrong).
42
436.49 Sat Oct 17 01:40:12 CEST 2015
44 - throwing an exception to a thread waiting in
45 Coro::Handle using Coro::EV did not stop the watchers,
46 causing the next call to fail (testcase by Martin Pritchard).
47 - bump minimum perl version to 5.10.
48
496.48 Sun Oct 4 19:03:51 CEST 2015
50 - fix memory leak when Coro::Handle uses Coro::EV internally
51 (testcase by Sten Sten).
52 - update code to libev 4 API (internally, Coro still used the
53 version 3 API calls).
54 - remove unused hv_sig.
55
566.47 Sat Jul 11 03:58:20 CEST 2015
57 - swap_svs - NVs can be in the SV head beginning with 5.20.
58
596.46 Tue Jun 30 14:36:30 CEST 2015
60 - restore portability to perls with windows fork emulation
61 (patch by Petr Písař).
62
636.45 Tue Jun 30 01:40:08 CEST 2015
64 - restore portability to perl 5.16 and below.
65
666.44 Tue Jun 30 00:41:54 CEST 2015
67 - avoid segfaulting (or worse) when tracing a canceled thread.
68 - the Event module silently broke it's public hook API, causing
69 failures on perls compiled with -Duselongdouble. Requiring
70 the latest version of Event because I don't know which release
71 changed this (it's not mentioned in the ChangeLog).
72 - allow xs level enterleave hooks via CoroAPI.
73
746.43 Thu Jun 4 15:38:14 CEST 2015
75 - use stability canary.
76 - port to stableperl-5.22.0-1.001.
77 - update libecb.
78
796.42 Wed Feb 11 20:29:52 CET 2015
80 - Coro::SemaphoreSet->try did not actually work (analyzed by
81 SATO Kentaro).
82 - upgrade libecb to be C11 compliant.
83
846.41 Sat Sep 6 22:08:46 CEST 2014
85 - restore portability to perl 5.8.x.
86 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
87 up some memory, but fixes "use feature" and similar modules
88 when used inside a Coro.
89 - allow zero as argument to Coro::Channel to mean the same thing
90 as no argument. this works with older versions as well,
91 but wasn't legal until now.
92 - slightly better c header file detection.
93
946.39 Mon Jun 2 00:00:08 CEST 2014
95 - work around more incompatible changes in 5.20.
96
976.38 Sun Jun 1 21:54:23 CEST 2014
98 - check that perl slots actually have enough space to hold
99 interpreter variables.
100 - untested port to perl 5.19 (64 bit tmps indices) (reported
101 by Andreas König).
102 - croak when cancel is called without a thread context.
103
1046.37 Tue Mar 4 13:27:33 CET 2014
105 - *sigh*, removed leftover debugging code from debugging a
106 perl bug, of all things.
107
1086.36 Tue Mar 4 07:11:59 CET 2014
109 - semaphores would not clear the destroy hook when interrupted
110 by ->throw, causing segfaults or worse.
111 - ->throw on a thread waiting for a semaphore did not acquire
112 the semaphore, but also didn't wake up other waiters,
113 possibly causing a deadlock.
114 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
115 to report a full stacktrace, which should help find out
116 where the actual call is.
117 - "atomically" destroy data in slf_destroy, because it is
118 the right thing to do, just in case.
119 - disable ecb.h libm dependency, to compile on ancient systems
120 or under adverse conditions.
121
1226.33 Mon Nov 18 11:26:27 CET 2013
123 - do not crash when freeing padlists with holes (in 5.18).
124 - tentative SVt_BIND 5.19 port/fix.
125
1266.32 Tue Nov 5 15:35:35 CET 2013
127 - use a new algorithm to derive padlists for perl 5.18. The old
128 one could lead to 0-pointer accesses inside perl (reported
129 by Darin McBride).
130
1316.31 Thu May 9 07:39:48 CEST 2013
132 - Coro::AIO requests would crash if the thread was ready'd
133 while the request was ongoing.
134
1356.29 Wed May 8 02:55:18 CEST 2013
136 - when an on_destroy handler destructs the coro currently being
137 destructed a perl scalar could be accessed after being freed,
138 likely causing a crash.
139
1406.28 Wed Mar 6 06:58:02 CET 2013
141 - clean remnants of existing __DIE__ and __WARN__ handlers so
142 they lose their magic and will not cause segfaults later
143 (testcase by Andrey Sagulin).
144 - improved Coro::State documentation a bit.
145 - Coro::Debug::command now flushes the output.
146 - add hack detection code for x32 abi, because the braindead slugs
147 who designed that made it look exactly like x86_64 without
148 providing proper compile time symbols to test for it. as a result,
149 this detection cannot work reliably.
150 - valgrind stack registering was broken.
151 - do not rely on Time::HiRes anymore in Coro::Debug.
152
1536.23 Fri Dec 7 23:36:37 CET 2012
154 - use experimental fiber implementation on native windows
155 perls.
156 - use sizeof (void *) as multiplication factor for stack sizes,
157 to accomodate the totally braindamaged microsoft 64 bit "os".
158 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
159 activeperl 5.16 crashes when PerlIO_define_layer is called due
160 to some bug in the perl dll, strawberry perl at least passes
161 the testsuite.
162 - implement Coro::Handle->peeraddr/host/port, for slightly
163 improved compatibility with LWP.
164 - implement 5.17 compatibility by almost blindly applying a
165 good-looking patch by Father Chrysostomos.
166 - move stack management functions into libcoro 3.
167 - libcoro version 3 "released".
168 - support magic values as timed_io_once args.
169 - recommend AnyEvent 7+ or EV 4+, also require EV
170 version 4 or newer for Coro::EV.
171
1726.10 Tue Oct 9 01:14:27 CEST 2012
173 - updated ecb.h, it had a typo that caused it to not compile on many
174 big endian systems (reported by many people).
175 - disable memory fences in ecb.h to improve portability.
176
1776.09 Sat Oct 6 23:25:02 CEST 2012
178 - Coro::EV I/O watchers were not interruptible by exceptions
179 (Coro::State::throw) (testcase by sten).
180 - ->throw now puts threads into the ready queue, as this seems to
181 be expected by existing code, and code that doesn't cope with spurious
182 wakeups needs fixing anyway.
183 - use fd -1 in mmap.
184 - cast I32 to int in error message printf.
185 - warn about broken so-called "hardened" kernels.
186
1876.08 Fri Apr 13 12:05:47 CEST 2012
188 - be more aggressive about exiting like perl does - formerly,
189 exiting from the non-main thread would not execute END blocks.
190
1916.07 Fri Nov 11 21:21:48 CET 2011
192 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
193 - update ecb.h.
194
1956.06 Mon Aug 8 23:59:48 CEST 2011
196 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
197 backend, which is an order of magnitude slower. unfortunately, cygwins
198 pthread implementation isn't very complete either, so allocate the stack
199 twice just to be sure.
200 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
201 in its ithreaded perl - best recompile cygwin and use the 'w'indows
202 backend for much better performance. also disable ithreads for
203 even better performance...).
204
2056.05 Thu Aug 4 21:36:36 CEST 2011
206 - blush, condvar values would not be propagated from send to recv anymore
207 (reported by Chip Salzenberg).
208 - use exponential increase for the readline buffer length in
209 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
210
2116.04 Wed Aug 3 17:15:45 CEST 2011
212 - use even more efficient and more compatible condvars for
213 compatibility to AnyEvent 6.x :)
214 - more inconsequential ecb.h updates.
215
2166.03 Wed Aug 3 11:41:30 CEST 2011
217 - change how Coro patches AnyEvent condvars for compatibility to
218 AnyEvent 6.x.
219 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
220 hasn't been included.
221
2226.02 Wed Jul 13 04:35:19 CEST 2011
223 - "improve portability to Gentoo" - gentoo manages to put perl variables
224 in memory areas that are farther than 2gb apart, which the jit couldn't
225 handle and barfed. now it's just a bit slower on gentoo and similar
226 systems.
227
2286.01 Sun Jul 3 12:31:14 CEST 2011
229 - workarounds are good, but the test for whether pthreads are used
230 was not good. this one should be better.
231 - check differently whether gcc generates cfi instructions itself.
232
2336.0 Wed Jun 29 19:43:35 CEST 2011
234 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
235 destroyed and cleaned up automatically (e.g. async { schedule }).
236 - implement a JIT compiler for part of the thread switch code,
237 which gives a 50% speed improvement on threaded perls, and
238 about 4% on non-threaded perls (so threaded perls now finally
239 reach about half the speed of non-threaded perls).
240 - slightly modernise Coro::Intro, add section about rouse functions.
241 - avoid DEFSV and ERRSV, giving another 10% improvement
242 in thread switching.
243 - Coro::State->is_destroyed is now called is_zombie.
244 - implement a Coro->safe_cancel method that might fail, but
245 cancels in a "safer" way if it succeeds.
246 - add preliminary support for DEBUGGING perls.
247 - get rid of two hash-accesses when initialising a new Coro - this
248 speeds up coro creation by almost a factor of two.
249 - croak when a coro that is being cancelled tries to block
250 (e.g. while executing a guard block), instead of crashing or
251 deadlocking.
252 - use a more robust and also faster method to identify Coro::State
253 objects - speeds up everything a bit.
254 - implement Coro->cancel in XS for a 20% speed improvement, and to
255 be able to implement mutual cancellation.
256 - speed up context switches by a percent or two by more efficiently
257 allocating context stack entries.
258 - implement Coro->join and Coro->on_destroy in XS for a speedup and
259 a reduction in memory use.
260 - cancelling a coro while it itself is cancelling another coro is
261 now supported and working, instead of triggering an assertion.
262 - be a bit more crash-resistant when calling (buggy) on_destroy
263 callbacks (best effort).
264 - move on_destroy into the slf_frame, to allow extension slf
265 functions to have destructors.
266 - get rid if coro refcounting - simply crash in other interpreter
267 threads by nulling the pointers on clone.
268 - simplify warn/die hook handling when loading Coro - the convoluted
269 logic seems to be no longer neccessary.
270 - use libecb instead of our own home-grown gcc hacks.
271 - document alternatives to Coro::LWP. Please use them :)
272 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
273 sigaltstack. Really. wine suffers from it, erlang suffers from it,
274 and it's known since at least 2006.
275
2765.372 Wed Feb 23 06:14:30 CET 2011
277 - apparently mingw doesn't provide a working gettimeofday, try to
278 work around that by relying on Time::HiRes (indirectly brought to
279 my attention by Max Maischein).
280 - fix some portability issues when Time::HiRes was used.
281
2825.371 Mon Feb 21 14:36:08 CET 2011
283 - backport to windows process emulation code again.
284
2855.37 Sat Feb 19 07:49:44 CET 2011
286 - add a big "Coro thread life cycle" section to "man Coro".
287 - try a tentative workaround against the breakage that 5.13 has
288 introduced without depreciation period. sigh.
289 - no longer use Time::HiRes if gettimeofday is available, which
290 saves quite a lot of memory.
291
2925.36 Sun Feb 13 05:33:41 CET 2011
293 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
294 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
295 method call.
296 - undocument Coro::Timer::sleep and obsolete whole module.
297 - optimise Coro::Timer::timeout memory and cpu usage.
298 - slightly updated Coro::Intro for recent changes.
299 - do not initialise PL_dirty anymore.
300
3015.25 Thu Nov 11 01:08:39 CET 2010
302 - try a different approach on netbsd - netbsd 5 finally has marginally
303 working pthreads, but still broken ucontext/sigaltstack.
304 - openbsd 4.8 finally got their act together, Coro works out of the box
305 with asm, setjmp and pthreads (no change, just informational).
8 306
95.24 Sat Oct 23 11:27:12 CEST 2010 3075.24 Sat Oct 23 11:27:12 CEST 2010
10 - port to the EV 4.0 API. 308 - port to the EV 4.0 API.
11 - work around bugs in mingw32, making strawberry perl work 309 - work around bugs in mingw32, making strawberry perl work
12 out of the box. 310 out of the box.
… …
299 - ->throw is now supported on Coro::State objects. 597 - ->throw is now supported on Coro::State objects.
300 - clean up cctx creation code a bit. 598 - clean up cctx creation code a bit.
301 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 599 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
302 600
3034.9 Sat Nov 8 17:45:27 CET 2008 6014.9 Sat Nov 8 17:45:27 CET 2008
304 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 602 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
305 - (libcoro) no longer leak threads in the experimental pthread backend, 603 - (libcoro) no longer leak threads in the experimental pthread backend,
306 also speed it up considerably. 604 also speed it up considerably.
307 - (libcoro) do not rely on makecontext passing void *'s unscathed. 605 - (libcoro) do not rely on makecontext passing void *'s unscathed.
308 - fix compiletime dependencies on libcoro in the Makefile. 606 - fix compiletime dependencies on libcoro in the Makefile.
309 - cctx_count wasn't always updated properly. 607 - cctx_count wasn't always updated properly.
… …
371 - remove debugging code related to MgPV_nolen_const, also try to 669 - remove debugging code related to MgPV_nolen_const, also try to
372 make it compile with perl 5.8.6 (yes, apple apparently loves 670 make it compile with perl 5.8.6 (yes, apple apparently loves
373 outdated software). Reported by John S. 671 outdated software). Reported by John S.
374 672
3754.744 Tue Jul 8 22:06:35 CEST 2008 6734.744 Tue Jul 8 22:06:35 CEST 2008
376 - correctly provide default DIE/WARN handlers as documented. 674 - correctly provide default DIE/WARN handlers as documented.
377 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 675 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
378 implementations inside perl work fine for us. 676 implementations inside perl work fine for us.
379 677
3804.743 Mon Jun 16 00:21:57 CEST 2008 6784.743 Mon Jun 16 00:21:57 CEST 2008
381 - when using Coro::EV without running EV::loop it could 679 - when using Coro::EV without running EV::loop it could
… …
403 - sprinkle "no warnings" freely over everything, also suppress 701 - sprinkle "no warnings" freely over everything, also suppress
404 warnings for some other modules. 702 warnings for some other modules.
405 - fix typo in WSAEWOULDBLOCK. 703 - fix typo in WSAEWOULDBLOCK.
406 704
4074.72 Sun May 25 05:14:36 CEST 2008 7054.72 Sun May 25 05:14:36 CEST 2008
408 - tweak META.yaml a bit, unfortunately, there is no documented way 706 - tweak META.yaml a bit, unfortunately, there is no documented way
409 to have optional dependencies with CPAN. doh :( 707 to have optional dependencies with CPAN. doh :(
410 - avoid running some tests on windows because they would fail due to 708 - avoid running some tests on windows because they would fail due to
411 perl bug (broken fork, broken pipes...). 709 perl bug (broken fork, broken pipes...).
412 - work around perl on windows bugs where perl returns undocumented 710 - work around perl on windows bugs where perl returns undocumented
413 error codes for sysread, syswrite etc. by taking advantage 711 error codes for sysread, syswrite etc. by taking advantage
… …
417 - use unix domain sockets in testsuite to work around 715 - use unix domain sockets in testsuite to work around
418 common perl implementation bugs on widows (they are emulated by 716 common perl implementation bugs on widows (they are emulated by
419 tcp sockets on windows. ugh.) 717 tcp sockets on windows. ugh.)
420 718
4214.71 Sat May 24 20:01:27 CEST 2008 7194.71 Sat May 24 20:01:27 CEST 2008
422 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 720 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
423 - take advantage of async name resolution of AnyEvent::Util. 721 - take advantage of async name resolution of AnyEvent::Util.
424 - work around brutal inet_aton override in Coro::LWP. 722 - work around brutal inet_aton override in Coro::LWP.
425 - take advantage of the readyhook in Coro::EV, for smoother 723 - take advantage of the readyhook in Coro::EV, for smoother
426 scheduling. 724 scheduling.
427 725
4284.7 Sun May 11 00:32:19 CEST 2008 7264.7 Sun May 11 00:32:19 CEST 2008
429 - completely reworked the Coro manpage. 727 - completely reworked the Coro manpage.
430 - added Coro::AnyEvent, generic event loop integration. 728 - added Coro::AnyEvent, generic event loop integration.
431 - implement cancel, ready and kill commands in Coro::Debug. 729 - implement cancel, ready and kill commands in Coro::Debug.
432 - document find_coro in Coro::Debug. 730 - document find_coro in Coro::Debug.
433 - incompatible change: rename has_stack to has_cctx. 731 - incompatible change: rename has_stack to has_cctx.
434 - Coro::AIO and Coro::BDB no longer force event model detection, 732 - Coro::AIO and Coro::BDB no longer force event model detection,
435 use AnyEvent::AIO and AnyEvent::BDB. 733 use AnyEvent::AIO and AnyEvent::BDB.
… …
559 857
5604.11 Thu Oct 11 02:40:24 CEST 2007 8584.11 Thu Oct 11 02:40:24 CEST 2007
561 - port to threaded perls. 859 - port to threaded perls.
562 860
5634.1 Thu Oct 11 02:38:16 CEST 2007 8614.1 Thu Oct 11 02:38:16 CEST 2007
564 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 862 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
565 be local to each coro (see Coro::State). 863 be local to each coro (see Coro::State).
566 - incompatible change: for very deep reasons, cede and cede_notself 864 - incompatible change: for very deep reasons, cede and cede_notself
567 cannot return anything, so nothing will be returned. 865 cannot return anything, so nothing will be returned.
568 - possibly bring back 5.10 compatibility (untested). 866 - possibly bring back 5.10 compatibility (untested).
569 - work around stupid (and wrong) warning on 5.10 :(. 867 - work around stupid (and wrong) warning on 5.10 :(.
… …
641 but stateful semantics. 939 but stateful semantics.
642 - fixed a lot of typos in Coro.pm (patch submitted by David 940 - fixed a lot of typos in Coro.pm (patch submitted by David
643 Steinbrunner, which applied flawlessly). 941 Steinbrunner, which applied flawlessly).
644 942
6453.6 Sat Apr 14 17:13:31 CEST 2007 9433.6 Sat Apr 14 17:13:31 CEST 2007
646 - added some bugfixes to get eg/myhttpd working again. 944 - added some bugfixes to get eg/myhttpd working again.
647 - added Coro::Storable for often-cede'ing freeze/thaw. 945 - added Coro::Storable for often-cede'ing freeze/thaw.
648 - try to do a clean exit when a coroutine calls exit 946 - try to do a clean exit when a coroutine calls exit
649 (EXPERIMENTAL). 947 (EXPERIMENTAL).
650 - got rid of indirect call through _coro_init. 948 - got rid of indirect call through _coro_init.
651 - updated the partly antique examples in eg/ to 949 - updated the partly antique examples in eg/ to

Diff Legend

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