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

Comparing Coro/Changes (file contents):
Revision 1.540 by root, Tue Jul 27 14:42:34 2010 UTC vs.
Revision 1.664 by root, Fri Jul 14 02:43:50 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 8
9 - experimental 5.24 compatibility changes.
10 - re-try JIT allocation without PROT_EXEC, to hopefully improve
11 portability.
12
136.511 Sun Jun 26 23:44:50 CEST 2016
14 - make it compile with threaded perls again.
15 - simplify and speed up __DIE__ and __WARN__ handling.
16
176.51 Sat Jun 25 20:57:20 CEST 2016
18 - 6.5 release didn't compile on most older perls due to
19 a logic error enabling experimental 5.24 code for all versions.
20 - since various broken or deficient patches float around
21 to work around the vtbl API breakage in perl, let's provide
22 our own workaround, which seems to be much less invasive
23 and more compatible than the approaches seen so far. YMMV.
24 - work around assertion failure in perl_destruct on 5.24
25 (analyzed by Dave Mitchell).
26
276.5 Wed Jun 22 22:23:50 CEST 2016
28 - swap_sv swapping order was not symmetrical, causing
29 wrong swaps when swap_sv was used multiple times on the
30 same sv in the same thread.
31 - swap_sv calls can now be undone by calling it again
32 with the same variables.
33 - swap_sv calls will now be undone in async_pool threads.
34 - split Coro::Semaphore::up/adjust into separate xs functions
35 for better error reporting, at slight codesize increase.
36 - (libcoro) arm assembly support, please test and report.
37 - adjust to PL_savestack changes in perl 5.24 (adapted from
38 the debian patch, which unfortunately gets it wrong).
39
406.49 Sat Oct 17 01:40:12 CEST 2015
41 - throwing an exception to a thread waiting in
42 Coro::Handle using Coro::EV did not stop the watchers,
43 causing the next call to fail (testcase by Martin Pritchard).
44 - bump minimum perl version to 5.10.
45
466.48 Sun Oct 4 19:03:51 CEST 2015
47 - fix memory leak when Coro::Handle uses Coro::EV internally
48 (testcase by Sten Sten).
49 - update code to libev 4 API (internally, Coro still used the
50 version 3 API calls).
51 - remove unused hv_sig.
52
536.47 Sat Jul 11 03:58:20 CEST 2015
54 - swap_svs - NVs can be in the SV head beginning with 5.20.
55
566.46 Tue Jun 30 14:36:30 CEST 2015
57 - restore portability to perls with windows fork emulation
58 (patch by Petr Písař).
59
606.45 Tue Jun 30 01:40:08 CEST 2015
61 - restore portability to perl 5.16 and below.
62
636.44 Tue Jun 30 00:41:54 CEST 2015
64 - avoid segfaulting (or worse) when tracing a canceled thread.
65 - the Event module silently broke it's public hook API, causing
66 failures on perls compiled with -Duselongdouble. Requiring
67 the latest version of Event because I don't know which release
68 changed this (it's not mentioned in the ChangeLog).
69 - allow xs level enterleave hooks via CoroAPI.
70
716.43 Thu Jun 4 15:38:14 CEST 2015
72 - use stability canary.
73 - port to stableperl-5.22.0-1.001.
74 - update libecb.
75
766.42 Wed Feb 11 20:29:52 CET 2015
77 - Coro::SemaphoreSet->try did not actually work (analyzed by
78 SATO Kentaro).
79 - upgrade libecb to be C11 compliant.
80
816.41 Sat Sep 6 22:08:46 CEST 2014
82 - restore portability to perl 5.8.x.
83 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
84 up some memory, but fixes "use feature" and similar modules
85 when used inside a Coro.
86 - allow zero as argument to Coro::Channel to mean the same thing
87 as no argument. this works with older versions as well,
88 but wasn't legal until now.
89 - slightly better c header file detection.
90
916.39 Mon Jun 2 00:00:08 CEST 2014
92 - work around more incompatible changes in 5.20.
93
946.38 Sun Jun 1 21:54:23 CEST 2014
95 - check that perl slots actually have enough space to hold
96 interpreter variables.
97 - untested port to perl 5.19 (64 bit tmps indices) (reported
98 by Andreas König).
99 - croak when cancel is called without a thread context.
100
1016.37 Tue Mar 4 13:27:33 CET 2014
102 - *sigh*, removed leftover debugging code from debugging a
103 perl bug, of all things.
104
1056.36 Tue Mar 4 07:11:59 CET 2014
106 - semaphores would not clear the destroy hook when interrupted
107 by ->throw, causing segfaults or worse.
108 - ->throw on a thread waiting for a semaphore did not acquire
109 the semaphore, but also didn't wake up other waiters,
110 possibly causing a deadlock.
111 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
112 to report a full stacktrace, which should help find out
113 where the actual call is.
114 - "atomically" destroy data in slf_destroy, because it is
115 the right thing to do, just in case.
116 - disable ecb.h libm dependency, to compile on ancient systems
117 or under adverse conditions.
118
1196.33 Mon Nov 18 11:26:27 CET 2013
120 - do not crash when freeing padlists with holes (in 5.18).
121 - tentative SVt_BIND 5.19 port/fix.
122
1236.32 Tue Nov 5 15:35:35 CET 2013
124 - use a new algorithm to derive padlists for perl 5.18. The old
125 one could lead to 0-pointer accesses inside perl (reported
126 by Darin McBride).
127
1286.31 Thu May 9 07:39:48 CEST 2013
129 - Coro::AIO requests would crash if the thread was ready'd
130 while the request was ongoing.
131
1326.29 Wed May 8 02:55:18 CEST 2013
133 - when an on_destroy handler destructs the coro currently being
134 destructed a perl scalar could be accessed after being freed,
135 likely causing a crash.
136
1376.28 Wed Mar 6 06:58:02 CET 2013
138 - clean remnants of existing __DIE__ and __WARN__ handlers so
139 they lose their magic and will not cause segfaults later
140 (testcase by Andrey Sagulin).
141 - improved Coro::State documentation a bit.
142 - Coro::Debug::command now flushes the output.
143 - add hack detection code for x32 abi, because the braindead slugs
144 who designed that made it look exactly like x86_64 without
145 providing proper compile time symbols to test for it. as a result,
146 this detection cannot work reliably.
147 - valgrind stack registering was broken.
148 - do not rely on Time::HiRes anymore in Coro::Debug.
149
1506.23 Fri Dec 7 23:36:37 CET 2012
151 - use experimental fiber implementation on native windows
152 perls.
153 - use sizeof (void *) as multiplication factor for stack sizes,
154 to accomodate the totally braindamaged microsoft 64 bit "os".
155 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
156 activeperl 5.16 crashes when PerlIO_define_layer is called due
157 to some bug in the perl dll, strawberry perl at least passes
158 the testsuite.
159 - implement Coro::Handle->peeraddr/host/port, for slightly
160 improved compatibility with LWP.
161 - implement 5.17 compatibility by almost blindly applying a
162 good-looking patch by Father Chrysostomos.
163 - move stack management functions into libcoro 3.
164 - libcoro version 3 "released".
165 - support magic values as timed_io_once args.
166 - recommend AnyEvent 7+ or EV 4+, also require EV
167 version 4 or newer for Coro::EV.
168
1696.10 Tue Oct 9 01:14:27 CEST 2012
170 - updated ecb.h, it had a typo that caused it to not compile on many
171 big endian systems (reported by many people).
172 - disable memory fences in ecb.h to improve portability.
173
1746.09 Sat Oct 6 23:25:02 CEST 2012
175 - Coro::EV I/O watchers were not interruptible by exceptions
176 (Coro::State::throw) (testcase by sten).
177 - ->throw now puts threads into the ready queue, as this seems to
178 be expected by existing code, and code that doesn't cope with spurious
179 wakeups needs fixing anyway.
180 - use fd -1 in mmap.
181 - cast I32 to int in error message printf.
182 - warn about broken so-called "hardened" kernels.
183
1846.08 Fri Apr 13 12:05:47 CEST 2012
185 - be more aggressive about exiting like perl does - formerly,
186 exiting from the non-main thread would not execute END blocks.
187
1886.07 Fri Nov 11 21:21:48 CET 2011
189 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
190 - update ecb.h.
191
1926.06 Mon Aug 8 23:59:48 CEST 2011
193 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
194 backend, which is an order of magnitude slower. unfortunately, cygwins
195 pthread implementation isn't very complete either, so allocate the stack
196 twice just to be sure.
197 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
198 in its ithreaded perl - best recompile cygwin and use the 'w'indows
199 backend for much better performance. also disable ithreads for
200 even better performance...).
201
2026.05 Thu Aug 4 21:36:36 CEST 2011
203 - blush, condvar values would not be propagated from send to recv anymore
204 (reported by Chip Salzenberg).
205 - use exponential increase for the readline buffer length in
206 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
207
2086.04 Wed Aug 3 17:15:45 CEST 2011
209 - use even more efficient and more compatible condvars for
210 compatibility to AnyEvent 6.x :)
211 - more inconsequential ecb.h updates.
212
2136.03 Wed Aug 3 11:41:30 CEST 2011
214 - change how Coro patches AnyEvent condvars for compatibility to
215 AnyEvent 6.x.
216 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
217 hasn't been included.
218
2196.02 Wed Jul 13 04:35:19 CEST 2011
220 - "improve portability to Gentoo" - gentoo manages to put perl variables
221 in memory areas that are farther than 2gb apart, which the jit couldn't
222 handle and barfed. now it's just a bit slower on gentoo and similar
223 systems.
224
2256.01 Sun Jul 3 12:31:14 CEST 2011
226 - workarounds are good, but the test for whether pthreads are used
227 was not good. this one should be better.
228 - check differently whether gcc generates cfi instructions itself.
229
2306.0 Wed Jun 29 19:43:35 CEST 2011
231 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
232 destroyed and cleaned up automatically (e.g. async { schedule }).
233 - implement a JIT compiler for part of the thread switch code,
234 which gives a 50% speed improvement on threaded perls, and
235 about 4% on non-threaded perls (so threaded perls now finally
236 reach about half the speed of non-threaded perls).
237 - slightly modernise Coro::Intro, add section about rouse functions.
238 - avoid DEFSV and ERRSV, giving another 10% improvement
239 in thread switching.
240 - Coro::State->is_destroyed is now called is_zombie.
241 - implement a Coro->safe_cancel method that might fail, but
242 cancels in a "safer" way if it succeeds.
243 - add preliminary support for DEBUGGING perls.
244 - get rid of two hash-accesses when initialising a new Coro - this
245 speeds up coro creation by almost a factor of two.
246 - croak when a coro that is being cancelled tries to block
247 (e.g. while executing a guard block), instead of crashing or
248 deadlocking.
249 - use a more robust and also faster method to identify Coro::State
250 objects - speeds up everything a bit.
251 - implement Coro->cancel in XS for a 20% speed improvement, and to
252 be able to implement mutual cancellation.
253 - speed up context switches by a percent or two by more efficiently
254 allocating context stack entries.
255 - implement Coro->join and Coro->on_destroy in XS for a speedup and
256 a reduction in memory use.
257 - cancelling a coro while it itself is cancelling another coro is
258 now supported and working, instead of triggering an assertion.
259 - be a bit more crash-resistant when calling (buggy) on_destroy
260 callbacks (best effort).
261 - move on_destroy into the slf_frame, to allow extension slf
262 functions to have destructors.
263 - get rid if coro refcounting - simply crash in other interpreter
264 threads by nulling the pointers on clone.
265 - simplify warn/die hook handling when loading Coro - the convoluted
266 logic seems to be no longer neccessary.
267 - use libecb instead of our own home-grown gcc hacks.
268 - document alternatives to Coro::LWP. Please use them :)
269 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
270 sigaltstack. Really. wine suffers from it, erlang suffers from it,
271 and it's known since at least 2006.
272
2735.372 Wed Feb 23 06:14:30 CET 2011
274 - apparently mingw doesn't provide a working gettimeofday, try to
275 work around that by relying on Time::HiRes (indirectly brought to
276 my attention by Max Maischein).
277 - fix some portability issues when Time::HiRes was used.
278
2795.371 Mon Feb 21 14:36:08 CET 2011
280 - backport to windows process emulation code again.
281
2825.37 Sat Feb 19 07:49:44 CET 2011
283 - add a big "Coro thread life cycle" section to "man Coro".
284 - try a tentative workaround against the breakage that 5.13 has
285 introduced without depreciation period. sigh.
286 - no longer use Time::HiRes if gettimeofday is available, which
287 saves quite a lot of memory.
288
2895.36 Sun Feb 13 05:33:41 CET 2011
290 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
291 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
292 method call.
293 - undocument Coro::Timer::sleep and obsolete whole module.
294 - optimise Coro::Timer::timeout memory and cpu usage.
295 - slightly updated Coro::Intro for recent changes.
296 - do not initialise PL_dirty anymore.
297
2985.25 Thu Nov 11 01:08:39 CET 2010
299 - try a different approach on netbsd - netbsd 5 finally has marginally
300 working pthreads, but still broken ucontext/sigaltstack.
301 - openbsd 4.8 finally got their act together, Coro works out of the box
302 with asm, setjmp and pthreads (no change, just informational).
303
3045.24 Sat Oct 23 11:27:12 CEST 2010
305 - port to the EV 4.0 API.
306 - work around bugs in mingw32, making strawberry perl work
307 out of the box.
308 - correctly modify Coro::AIO function prototypes
309 so that they reflect the "no optional parameters" rule.
9 - "ported" libcoro to C++. 310 - "ported" libcoro to C++.
10 311
115.23 Mon May 17 18:50:42 CEST 2010 3125.23 Mon May 17 18:50:42 CEST 2010
12 - be more resistant to ordering changes when initialising 313 - be more resistant to ordering changes when initialising
13 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias 314 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
… …
293 - ->throw is now supported on Coro::State objects. 594 - ->throw is now supported on Coro::State objects.
294 - clean up cctx creation code a bit. 595 - clean up cctx creation code a bit.
295 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 596 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
296 597
2974.9 Sat Nov 8 17:45:27 CET 2008 5984.9 Sat Nov 8 17:45:27 CET 2008
298 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 599 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
299 - (libcoro) no longer leak threads in the experimental pthread backend, 600 - (libcoro) no longer leak threads in the experimental pthread backend,
300 also speed it up considerably. 601 also speed it up considerably.
301 - (libcoro) do not rely on makecontext passing void *'s unscathed. 602 - (libcoro) do not rely on makecontext passing void *'s unscathed.
302 - fix compiletime dependencies on libcoro in the Makefile. 603 - fix compiletime dependencies on libcoro in the Makefile.
303 - cctx_count wasn't always updated properly. 604 - cctx_count wasn't always updated properly.
… …
365 - remove debugging code related to MgPV_nolen_const, also try to 666 - remove debugging code related to MgPV_nolen_const, also try to
366 make it compile with perl 5.8.6 (yes, apple apparently loves 667 make it compile with perl 5.8.6 (yes, apple apparently loves
367 outdated software). Reported by John S. 668 outdated software). Reported by John S.
368 669
3694.744 Tue Jul 8 22:06:35 CEST 2008 6704.744 Tue Jul 8 22:06:35 CEST 2008
370 - correctly provide default DIE/WARN handlers as documented. 671 - correctly provide default DIE/WARN handlers as documented.
371 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 672 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
372 implementations inside perl work fine for us. 673 implementations inside perl work fine for us.
373 674
3744.743 Mon Jun 16 00:21:57 CEST 2008 6754.743 Mon Jun 16 00:21:57 CEST 2008
375 - when using Coro::EV without running EV::loop it could 676 - when using Coro::EV without running EV::loop it could
… …
397 - sprinkle "no warnings" freely over everything, also suppress 698 - sprinkle "no warnings" freely over everything, also suppress
398 warnings for some other modules. 699 warnings for some other modules.
399 - fix typo in WSAEWOULDBLOCK. 700 - fix typo in WSAEWOULDBLOCK.
400 701
4014.72 Sun May 25 05:14:36 CEST 2008 7024.72 Sun May 25 05:14:36 CEST 2008
402 - tweak META.yaml a bit, unfortunately, there is no documented way 703 - tweak META.yaml a bit, unfortunately, there is no documented way
403 to have optional dependencies with CPAN. doh :( 704 to have optional dependencies with CPAN. doh :(
404 - avoid running some tests on windows because they would fail due to 705 - avoid running some tests on windows because they would fail due to
405 perl bug (broken fork, broken pipes...). 706 perl bug (broken fork, broken pipes...).
406 - work around perl on windows bugs where perl returns undocumented 707 - work around perl on windows bugs where perl returns undocumented
407 error codes for sysread, syswrite etc. by taking advantage 708 error codes for sysread, syswrite etc. by taking advantage
… …
411 - use unix domain sockets in testsuite to work around 712 - use unix domain sockets in testsuite to work around
412 common perl implementation bugs on widows (they are emulated by 713 common perl implementation bugs on widows (they are emulated by
413 tcp sockets on windows. ugh.) 714 tcp sockets on windows. ugh.)
414 715
4154.71 Sat May 24 20:01:27 CEST 2008 7164.71 Sat May 24 20:01:27 CEST 2008
416 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 717 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
417 - take advantage of async name resolution of AnyEvent::Util. 718 - take advantage of async name resolution of AnyEvent::Util.
418 - work around brutal inet_aton override in Coro::LWP. 719 - work around brutal inet_aton override in Coro::LWP.
419 - take advantage of the readyhook in Coro::EV, for smoother 720 - take advantage of the readyhook in Coro::EV, for smoother
420 scheduling. 721 scheduling.
421 722
4224.7 Sun May 11 00:32:19 CEST 2008 7234.7 Sun May 11 00:32:19 CEST 2008
423 - completely reworked the Coro manpage. 724 - completely reworked the Coro manpage.
424 - added Coro::AnyEvent, generic event loop integration. 725 - added Coro::AnyEvent, generic event loop integration.
425 - implement cancel, ready and kill commands in Coro::Debug. 726 - implement cancel, ready and kill commands in Coro::Debug.
426 - document find_coro in Coro::Debug. 727 - document find_coro in Coro::Debug.
427 - incompatible change: rename has_stack to has_cctx. 728 - incompatible change: rename has_stack to has_cctx.
428 - Coro::AIO and Coro::BDB no longer force event model detection, 729 - Coro::AIO and Coro::BDB no longer force event model detection,
429 use AnyEvent::AIO and AnyEvent::BDB. 730 use AnyEvent::AIO and AnyEvent::BDB.
… …
553 854
5544.11 Thu Oct 11 02:40:24 CEST 2007 8554.11 Thu Oct 11 02:40:24 CEST 2007
555 - port to threaded perls. 856 - port to threaded perls.
556 857
5574.1 Thu Oct 11 02:38:16 CEST 2007 8584.1 Thu Oct 11 02:38:16 CEST 2007
558 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 859 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
559 be local to each coro (see Coro::State). 860 be local to each coro (see Coro::State).
560 - incompatible change: for very deep reasons, cede and cede_notself 861 - incompatible change: for very deep reasons, cede and cede_notself
561 cannot return anything, so nothing will be returned. 862 cannot return anything, so nothing will be returned.
562 - possibly bring back 5.10 compatibility (untested). 863 - possibly bring back 5.10 compatibility (untested).
563 - work around stupid (and wrong) warning on 5.10 :(. 864 - work around stupid (and wrong) warning on 5.10 :(.
… …
635 but stateful semantics. 936 but stateful semantics.
636 - fixed a lot of typos in Coro.pm (patch submitted by David 937 - fixed a lot of typos in Coro.pm (patch submitted by David
637 Steinbrunner, which applied flawlessly). 938 Steinbrunner, which applied flawlessly).
638 939
6393.6 Sat Apr 14 17:13:31 CEST 2007 9403.6 Sat Apr 14 17:13:31 CEST 2007
640 - added some bugfixes to get eg/myhttpd working again. 941 - added some bugfixes to get eg/myhttpd working again.
641 - added Coro::Storable for often-cede'ing freeze/thaw. 942 - added Coro::Storable for often-cede'ing freeze/thaw.
642 - try to do a clean exit when a coroutine calls exit 943 - try to do a clean exit when a coroutine calls exit
643 (EXPERIMENTAL). 944 (EXPERIMENTAL).
644 - got rid of indirect call through _coro_init. 945 - got rid of indirect call through _coro_init.
645 - updated the partly antique examples in eg/ to 946 - updated the partly antique examples in eg/ to

Diff Legend

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