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

Comparing Coro/Changes (file contents):
Revision 1.546 by root, Sun Feb 6 00:27:43 2011 UTC vs.
Revision 1.648 by root, Sun Oct 4 13:07:55 2015 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, andallow to unswap by repeat?
7TODO: swap_sv not undone in pool
8TODO: croak when async_pool tries to run canceled thread?
8 9
10 - fix memory leak when Coro::Handle uses Coro::EV internally
11 (testcase by Sten Sten).
12 - update code to libev 4 API (internally, Coro still used the
13 version 3 API calls).
14 - remove unused hv_sig.
15
166.47 Sat Jul 11 03:58:20 CEST 2015
17 - swap_svs - NVs can be in the SV head beginning with 5.20.
18
196.46 Tue Jun 30 14:36:30 CEST 2015
20 - restore portability to perls with windows fork emulation
21 (patch by Petr Písař).
22
236.45 Tue Jun 30 01:40:08 CEST 2015
24 - restore portability to perl 5.16 and below.
25
266.44 Tue Jun 30 00:41:54 CEST 2015
27 - avoid segfaulting (or worse) when tracing a canceled thread.
28 - the Event module silently broke it's public hook API, causing
29 failures on perls compiled with -Duselongdouble. Requiring
30 the latest version of Event because I don't know which release
31 changed this (it's not mentioned in the ChangeLog).
32 - allow xs level enterleave hooks via CoroAPI.
33
346.43 Thu Jun 4 15:38:14 CEST 2015
35 - use stability canary.
36 - port to stableperl-5.22.0-1.001.
37 - update libecb.
38
396.42 Wed Feb 11 20:29:52 CET 2015
40 - Coro::SemaphoreSet->try did not actually work (analyzed by
41 SATO Kentaro).
42 - upgrade libecb to be C11 compliant.
43
446.41 Sat Sep 6 22:08:46 CEST 2014
45 - restore portability to perl 5.8.x.
46 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
47 up some memory, but fixes "use feature" and similar modules
48 when used inside a Coro.
49 - allow zero as argument to Coro::Channel to mean the same thing
50 as no argument. this works with older versions as well,
51 but wasn't legal until now.
52 - slightly better c header file detection.
53
546.39 Mon Jun 2 00:00:08 CEST 2014
55 - work around more incompatible changes in 5.20.
56
576.38 Sun Jun 1 21:54:23 CEST 2014
58 - check that perl slots actually have enough space to hold
59 interpreter variables.
60 - untested port to perl 5.19 (64 bit tmps indices) (reported
61 by Andreas König).
62 - croak when cancel is called without a thread context.
63
646.37 Tue Mar 4 13:27:33 CET 2014
65 - *sigh*, removed leftover debugging code from debugging a
66 perl bug, of all things.
67
686.36 Tue Mar 4 07:11:59 CET 2014
69 - semaphores would not clear the destroy hook when interrupted
70 by ->throw, causing segfaults or worse.
71 - ->throw on a thread waiting for a semaphore did not acquire
72 the semaphore, but also didn't wake up other waiters,
73 possibly causing a deadlock.
74 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
75 to report a full stacktrace, which should help find out
76 where the actual call is.
77 - "atomically" destroy data in slf_destroy, because it is
78 the right thing to do, just in case.
79 - disable ecb.h libm dependency, to compile on ancient systems
80 or under adverse conditions.
81
826.33 Mon Nov 18 11:26:27 CET 2013
83 - do not crash when freeing padlists with holes (in 5.18).
84 - tentative SVt_BIND 5.19 port/fix.
85
866.32 Tue Nov 5 15:35:35 CET 2013
87 - use a new algorithm to derive padlists for perl 5.18. The old
88 one could lead to 0-pointer accesses inside perl (reported
89 by Darin McBride).
90
916.31 Thu May 9 07:39:48 CEST 2013
92 - Coro::AIO requests would crash if the thread was ready'd
93 while the request was ongoing.
94
956.29 Wed May 8 02:55:18 CEST 2013
96 - when an on_destroy handler destructs the coro currently being
97 destructed a perl scalar could be accessed after being freed,
98 likely causing a crash.
99
1006.28 Wed Mar 6 06:58:02 CET 2013
101 - clean remnants of existing __DIE__ and __WARN__ handlers so
102 they lose their magic and will not cause segfaults later
103 (testcase by Andrey Sagulin).
104 - improved Coro::State documentation a bit.
105 - Coro::Debug::command now flushes the output.
106 - add hack detection code for x32 abi, because the braindead slugs
107 who designed that made it look exactly like x86_64 without
108 providing proper compile time symbols to test for it. as a result,
109 this detection cannot work reliably.
110 - valgrind stack registering was broken.
111 - do not rely on Time::HiRes anymore in Coro::Debug.
112
1136.23 Fri Dec 7 23:36:37 CET 2012
114 - use experimental fiber implementation on native windows
115 perls.
116 - use sizeof (void *) as multiplication factor for stack sizes,
117 to accomodate the totally braindamaged microsoft 64 bit "os".
118 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
119 activeperl 5.16 crashes when PerlIO_define_layer is called due
120 to some bug in the perl dll, strawberry perl at least passes
121 the testsuite.
122 - implement Coro::Handle->peeraddr/host/port, for slightly
123 improved compatibility with LWP.
124 - implement 5.17 compatibility by almost blindly applying a
125 good-looking patch by Father Chrysostomos.
126 - move stack management functions into libcoro 3.
127 - libcoro version 3 "released".
128 - support magic values as timed_io_once args.
129 - recommend AnyEvent 7+ or EV 4+, also require EV
130 version 4 or newer for Coro::EV.
131
1326.10 Tue Oct 9 01:14:27 CEST 2012
133 - updated ecb.h, it had a typo that caused it to not compile on many
134 big endian systems (reported by many people).
135 - disable memory fences in ecb.h to improve portability.
136
1376.09 Sat Oct 6 23:25:02 CEST 2012
138 - Coro::EV I/O watchers were not interruptible by exceptions
139 (Coro::State::throw) (testcase by sten).
140 - ->throw now puts threads into the ready queue, as this seems to
141 be expected by existing code, and code that doesn't cope with spurious
142 wakeups needs fixing anyway.
143 - use fd -1 in mmap.
144 - cast I32 to int in error message printf.
145 - warn about broken so-called "hardened" kernels.
146
1476.08 Fri Apr 13 12:05:47 CEST 2012
148 - be more aggressive about exiting like perl does - formerly,
149 exiting from the non-main thread would not execute END blocks.
150
1516.07 Fri Nov 11 21:21:48 CET 2011
152 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
153 - update ecb.h.
154
1556.06 Mon Aug 8 23:59:48 CEST 2011
156 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
157 backend, which is an order of magnitude slower. unfortunately, cygwins
158 pthread implementation isn't very complete either, so allocate the stack
159 twice just to be sure.
160 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
161 in its ithreaded perl - best recompile cygwin and use the 'w'indows
162 backend for much better performance. also disable ithreads for
163 even better performance...).
164
1656.05 Thu Aug 4 21:36:36 CEST 2011
166 - blush, condvar values would not be propagated from send to recv anymore
167 (reported by Chip Salzenberg).
168 - use exponential increase for the readline buffer length in
169 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
170
1716.04 Wed Aug 3 17:15:45 CEST 2011
172 - use even more efficient and more compatible condvars for
173 compatibility to AnyEvent 6.x :)
174 - more inconsequential ecb.h updates.
175
1766.03 Wed Aug 3 11:41:30 CEST 2011
177 - change how Coro patches AnyEvent condvars for compatibility to
178 AnyEvent 6.x.
179 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
180 hasn't been included.
181
1826.02 Wed Jul 13 04:35:19 CEST 2011
183 - "improve portability to Gentoo" - gentoo manages to put perl variables
184 in memory areas that are farther than 2gb apart, which the jit couldn't
185 handle and barfed. now it's just a bit slower on gentoo and similar
186 systems.
187
1886.01 Sun Jul 3 12:31:14 CEST 2011
189 - workarounds are good, but the test for whether pthreads are used
190 was not good. this one should be better.
191 - check differently whether gcc generates cfi instructions itself.
192
1936.0 Wed Jun 29 19:43:35 CEST 2011
194 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
195 destroyed and cleaned up automatically (e.g. async { schedule }).
196 - implement a JIT compiler for part of the thread switch code,
197 which gives a 50% speed improvement on threaded perls, and
198 about 4% on non-threaded perls (so threaded perls now finally
199 reach about half the speed of non-threaded perls).
200 - slightly modernise Coro::Intro, add section about rouse functions.
201 - avoid DEFSV and ERRSV, giving another 10% improvement
202 in thread switching.
203 - Coro::State->is_destroyed is now called is_zombie.
204 - implement a Coro->safe_cancel method that might fail, but
205 cancels in a "safer" way if it succeeds.
206 - add preliminary support for DEBUGGING perls.
207 - get rid of two hash-accesses when initialising a new Coro - this
208 speeds up coro creation by almost a factor of two.
209 - croak when a coro that is being cancelled tries to block
210 (e.g. while executing a guard block), instead of crashing or
211 deadlocking.
212 - use a more robust and also faster method to identify Coro::State
213 objects - speeds up everything a bit.
214 - implement Coro->cancel in XS for a 20% speed improvement, and to
215 be able to implement mutual cancellation.
216 - speed up context switches by a percent or two by more efficiently
217 allocating context stack entries.
218 - implement Coro->join and Coro->on_destroy in XS for a speedup and
219 a reduction in memory use.
220 - cancelling a coro while it itself is cancelling another coro is
221 now supported and working, instead of triggering an assertion.
222 - be a bit more crash-resistant when calling (buggy) on_destroy
223 callbacks (best effort).
224 - move on_destroy into the slf_frame, to allow extension slf
225 functions to have destructors.
226 - get rid if coro refcounting - simply crash in other interpreter
227 threads by nulling the pointers on clone.
228 - simplify warn/die hook handling when loading Coro - the convoluted
229 logic seems to be no longer neccessary.
230 - use libecb instead of our own home-grown gcc hacks.
231 - document alternatives to Coro::LWP. Please use them :)
232 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
233 sigaltstack. Really. wine suffers from it, erlang suffers from it,
234 and it's known since at least 2006.
235
2365.372 Wed Feb 23 06:14:30 CET 2011
237 - apparently mingw doesn't provide a working gettimeofday, try to
238 work around that by relying on Time::HiRes (indirectly brought to
239 my attention by Max Maischein).
240 - fix some portability issues when Time::HiRes was used.
241
2425.371 Mon Feb 21 14:36:08 CET 2011
243 - backport to windows process emulation code again.
244
2455.37 Sat Feb 19 07:49:44 CET 2011
246 - add a big "Coro thread life cycle" section to "man Coro".
247 - try a tentative workaround against the breakage that 5.13 has
248 introduced without depreciation period. sigh.
249 - no longer use Time::HiRes if gettimeofday is available, which
250 saves quite a lot of memory.
251
2525.36 Sun Feb 13 05:33:41 CET 2011
9 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore, 253 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
10 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new" 254 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
11 method call. 255 method call.
12 - undocument Coro::Timer::sleep and obsolete whole module. 256 - undocument Coro::Timer::sleep and obsolete whole module.
13 - optimise Coro::Timer::timeout memory and cpu usage. 257 - optimise Coro::Timer::timeout memory and cpu usage.
14 - slightly updated Coro::Intro for recent changes. 258 - slightly updated Coro::Intro for recent changes.
259 - do not initialise PL_dirty anymore.
15 260
165.25 Thu Nov 11 01:08:39 CET 2010 2615.25 Thu Nov 11 01:08:39 CET 2010
17 - try a different approach on netbsd - netbsd 5 finally has marginally 262 - try a different approach on netbsd - netbsd 5 finally has marginally
18 working pthreads, but still broken ucontext/sigaltstack. 263 working pthreads, but still broken ucontext/sigaltstack.
19 - openbsd 4.8 finally got their act together, Coro works out of the box 264 - openbsd 4.8 finally got their act together, Coro works out of the box
… …
312 - ->throw is now supported on Coro::State objects. 557 - ->throw is now supported on Coro::State objects.
313 - clean up cctx creation code a bit. 558 - clean up cctx creation code a bit.
314 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 559 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
315 560
3164.9 Sat Nov 8 17:45:27 CET 2008 5614.9 Sat Nov 8 17:45:27 CET 2008
317 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 562 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
318 - (libcoro) no longer leak threads in the experimental pthread backend, 563 - (libcoro) no longer leak threads in the experimental pthread backend,
319 also speed it up considerably. 564 also speed it up considerably.
320 - (libcoro) do not rely on makecontext passing void *'s unscathed. 565 - (libcoro) do not rely on makecontext passing void *'s unscathed.
321 - fix compiletime dependencies on libcoro in the Makefile. 566 - fix compiletime dependencies on libcoro in the Makefile.
322 - cctx_count wasn't always updated properly. 567 - cctx_count wasn't always updated properly.
… …
384 - remove debugging code related to MgPV_nolen_const, also try to 629 - remove debugging code related to MgPV_nolen_const, also try to
385 make it compile with perl 5.8.6 (yes, apple apparently loves 630 make it compile with perl 5.8.6 (yes, apple apparently loves
386 outdated software). Reported by John S. 631 outdated software). Reported by John S.
387 632
3884.744 Tue Jul 8 22:06:35 CEST 2008 6334.744 Tue Jul 8 22:06:35 CEST 2008
389 - correctly provide default DIE/WARN handlers as documented. 634 - correctly provide default DIE/WARN handlers as documented.
390 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 635 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
391 implementations inside perl work fine for us. 636 implementations inside perl work fine for us.
392 637
3934.743 Mon Jun 16 00:21:57 CEST 2008 6384.743 Mon Jun 16 00:21:57 CEST 2008
394 - when using Coro::EV without running EV::loop it could 639 - when using Coro::EV without running EV::loop it could
… …
416 - sprinkle "no warnings" freely over everything, also suppress 661 - sprinkle "no warnings" freely over everything, also suppress
417 warnings for some other modules. 662 warnings for some other modules.
418 - fix typo in WSAEWOULDBLOCK. 663 - fix typo in WSAEWOULDBLOCK.
419 664
4204.72 Sun May 25 05:14:36 CEST 2008 6654.72 Sun May 25 05:14:36 CEST 2008
421 - tweak META.yaml a bit, unfortunately, there is no documented way 666 - tweak META.yaml a bit, unfortunately, there is no documented way
422 to have optional dependencies with CPAN. doh :( 667 to have optional dependencies with CPAN. doh :(
423 - avoid running some tests on windows because they would fail due to 668 - avoid running some tests on windows because they would fail due to
424 perl bug (broken fork, broken pipes...). 669 perl bug (broken fork, broken pipes...).
425 - work around perl on windows bugs where perl returns undocumented 670 - work around perl on windows bugs where perl returns undocumented
426 error codes for sysread, syswrite etc. by taking advantage 671 error codes for sysread, syswrite etc. by taking advantage
… …
430 - use unix domain sockets in testsuite to work around 675 - use unix domain sockets in testsuite to work around
431 common perl implementation bugs on widows (they are emulated by 676 common perl implementation bugs on widows (they are emulated by
432 tcp sockets on windows. ugh.) 677 tcp sockets on windows. ugh.)
433 678
4344.71 Sat May 24 20:01:27 CEST 2008 6794.71 Sat May 24 20:01:27 CEST 2008
435 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 680 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
436 - take advantage of async name resolution of AnyEvent::Util. 681 - take advantage of async name resolution of AnyEvent::Util.
437 - work around brutal inet_aton override in Coro::LWP. 682 - work around brutal inet_aton override in Coro::LWP.
438 - take advantage of the readyhook in Coro::EV, for smoother 683 - take advantage of the readyhook in Coro::EV, for smoother
439 scheduling. 684 scheduling.
440 685
4414.7 Sun May 11 00:32:19 CEST 2008 6864.7 Sun May 11 00:32:19 CEST 2008
442 - completely reworked the Coro manpage. 687 - completely reworked the Coro manpage.
443 - added Coro::AnyEvent, generic event loop integration. 688 - added Coro::AnyEvent, generic event loop integration.
444 - implement cancel, ready and kill commands in Coro::Debug. 689 - implement cancel, ready and kill commands in Coro::Debug.
445 - document find_coro in Coro::Debug. 690 - document find_coro in Coro::Debug.
446 - incompatible change: rename has_stack to has_cctx. 691 - incompatible change: rename has_stack to has_cctx.
447 - Coro::AIO and Coro::BDB no longer force event model detection, 692 - Coro::AIO and Coro::BDB no longer force event model detection,
448 use AnyEvent::AIO and AnyEvent::BDB. 693 use AnyEvent::AIO and AnyEvent::BDB.
… …
572 817
5734.11 Thu Oct 11 02:40:24 CEST 2007 8184.11 Thu Oct 11 02:40:24 CEST 2007
574 - port to threaded perls. 819 - port to threaded perls.
575 820
5764.1 Thu Oct 11 02:38:16 CEST 2007 8214.1 Thu Oct 11 02:38:16 CEST 2007
577 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 822 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
578 be local to each coro (see Coro::State). 823 be local to each coro (see Coro::State).
579 - incompatible change: for very deep reasons, cede and cede_notself 824 - incompatible change: for very deep reasons, cede and cede_notself
580 cannot return anything, so nothing will be returned. 825 cannot return anything, so nothing will be returned.
581 - possibly bring back 5.10 compatibility (untested). 826 - possibly bring back 5.10 compatibility (untested).
582 - work around stupid (and wrong) warning on 5.10 :(. 827 - work around stupid (and wrong) warning on 5.10 :(.
… …
654 but stateful semantics. 899 but stateful semantics.
655 - fixed a lot of typos in Coro.pm (patch submitted by David 900 - fixed a lot of typos in Coro.pm (patch submitted by David
656 Steinbrunner, which applied flawlessly). 901 Steinbrunner, which applied flawlessly).
657 902
6583.6 Sat Apr 14 17:13:31 CEST 2007 9033.6 Sat Apr 14 17:13:31 CEST 2007
659 - added some bugfixes to get eg/myhttpd working again. 904 - added some bugfixes to get eg/myhttpd working again.
660 - added Coro::Storable for often-cede'ing freeze/thaw. 905 - added Coro::Storable for often-cede'ing freeze/thaw.
661 - try to do a clean exit when a coroutine calls exit 906 - try to do a clean exit when a coroutine calls exit
662 (EXPERIMENTAL). 907 (EXPERIMENTAL).
663 - got rid of indirect call through _coro_init. 908 - got rid of indirect call through _coro_init.
664 - updated the partly antique examples in eg/ to 909 - updated the partly antique examples in eg/ to

Diff Legend

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