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

Comparing Coro/Changes (file contents):
Revision 1.547 by root, Sun Feb 13 04:39:14 2011 UTC vs.
Revision 1.654 by root, Sun Mar 6 06:26:21 2016 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?
9
10 - split Coro::Semaphore::up/adjust into separate xs functions
11 for better error reporting, at slight codesize increase.
12
136.49 Sat Oct 17 01:40:12 CEST 2015
14 - throwing an exception to a thread waiting in
15 Coro::Handle using Coro::EV did not stop the watchers,
16 causing the next call to fail (testcase by Martin Pritchard).
17 - bump minimum perl version to 5.10.
18
196.48 Sun Oct 4 19:03:51 CEST 2015
20 - fix memory leak when Coro::Handle uses Coro::EV internally
21 (testcase by Sten Sten).
22 - update code to libev 4 API (internally, Coro still used the
23 version 3 API calls).
24 - remove unused hv_sig.
25
266.47 Sat Jul 11 03:58:20 CEST 2015
27 - swap_svs - NVs can be in the SV head beginning with 5.20.
28
296.46 Tue Jun 30 14:36:30 CEST 2015
30 - restore portability to perls with windows fork emulation
31 (patch by Petr Písař).
32
336.45 Tue Jun 30 01:40:08 CEST 2015
34 - restore portability to perl 5.16 and below.
35
366.44 Tue Jun 30 00:41:54 CEST 2015
37 - avoid segfaulting (or worse) when tracing a canceled thread.
38 - the Event module silently broke it's public hook API, causing
39 failures on perls compiled with -Duselongdouble. Requiring
40 the latest version of Event because I don't know which release
41 changed this (it's not mentioned in the ChangeLog).
42 - allow xs level enterleave hooks via CoroAPI.
43
446.43 Thu Jun 4 15:38:14 CEST 2015
45 - use stability canary.
46 - port to stableperl-5.22.0-1.001.
47 - update libecb.
48
496.42 Wed Feb 11 20:29:52 CET 2015
50 - Coro::SemaphoreSet->try did not actually work (analyzed by
51 SATO Kentaro).
52 - upgrade libecb to be C11 compliant.
53
546.41 Sat Sep 6 22:08:46 CEST 2014
55 - restore portability to perl 5.8.x.
56 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
57 up some memory, but fixes "use feature" and similar modules
58 when used inside a Coro.
59 - allow zero as argument to Coro::Channel to mean the same thing
60 as no argument. this works with older versions as well,
61 but wasn't legal until now.
62 - slightly better c header file detection.
63
646.39 Mon Jun 2 00:00:08 CEST 2014
65 - work around more incompatible changes in 5.20.
66
676.38 Sun Jun 1 21:54:23 CEST 2014
68 - check that perl slots actually have enough space to hold
69 interpreter variables.
70 - untested port to perl 5.19 (64 bit tmps indices) (reported
71 by Andreas König).
72 - croak when cancel is called without a thread context.
73
746.37 Tue Mar 4 13:27:33 CET 2014
75 - *sigh*, removed leftover debugging code from debugging a
76 perl bug, of all things.
77
786.36 Tue Mar 4 07:11:59 CET 2014
79 - semaphores would not clear the destroy hook when interrupted
80 by ->throw, causing segfaults or worse.
81 - ->throw on a thread waiting for a semaphore did not acquire
82 the semaphore, but also didn't wake up other waiters,
83 possibly causing a deadlock.
84 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
85 to report a full stacktrace, which should help find out
86 where the actual call is.
87 - "atomically" destroy data in slf_destroy, because it is
88 the right thing to do, just in case.
89 - disable ecb.h libm dependency, to compile on ancient systems
90 or under adverse conditions.
91
926.33 Mon Nov 18 11:26:27 CET 2013
93 - do not crash when freeing padlists with holes (in 5.18).
94 - tentative SVt_BIND 5.19 port/fix.
95
966.32 Tue Nov 5 15:35:35 CET 2013
97 - use a new algorithm to derive padlists for perl 5.18. The old
98 one could lead to 0-pointer accesses inside perl (reported
99 by Darin McBride).
100
1016.31 Thu May 9 07:39:48 CEST 2013
102 - Coro::AIO requests would crash if the thread was ready'd
103 while the request was ongoing.
104
1056.29 Wed May 8 02:55:18 CEST 2013
106 - when an on_destroy handler destructs the coro currently being
107 destructed a perl scalar could be accessed after being freed,
108 likely causing a crash.
109
1106.28 Wed Mar 6 06:58:02 CET 2013
111 - clean remnants of existing __DIE__ and __WARN__ handlers so
112 they lose their magic and will not cause segfaults later
113 (testcase by Andrey Sagulin).
114 - improved Coro::State documentation a bit.
115 - Coro::Debug::command now flushes the output.
116 - add hack detection code for x32 abi, because the braindead slugs
117 who designed that made it look exactly like x86_64 without
118 providing proper compile time symbols to test for it. as a result,
119 this detection cannot work reliably.
120 - valgrind stack registering was broken.
121 - do not rely on Time::HiRes anymore in Coro::Debug.
122
1236.23 Fri Dec 7 23:36:37 CET 2012
124 - use experimental fiber implementation on native windows
125 perls.
126 - use sizeof (void *) as multiplication factor for stack sizes,
127 to accomodate the totally braindamaged microsoft 64 bit "os".
128 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
129 activeperl 5.16 crashes when PerlIO_define_layer is called due
130 to some bug in the perl dll, strawberry perl at least passes
131 the testsuite.
132 - implement Coro::Handle->peeraddr/host/port, for slightly
133 improved compatibility with LWP.
134 - implement 5.17 compatibility by almost blindly applying a
135 good-looking patch by Father Chrysostomos.
136 - move stack management functions into libcoro 3.
137 - libcoro version 3 "released".
138 - support magic values as timed_io_once args.
139 - recommend AnyEvent 7+ or EV 4+, also require EV
140 version 4 or newer for Coro::EV.
141
1426.10 Tue Oct 9 01:14:27 CEST 2012
143 - updated ecb.h, it had a typo that caused it to not compile on many
144 big endian systems (reported by many people).
145 - disable memory fences in ecb.h to improve portability.
146
1476.09 Sat Oct 6 23:25:02 CEST 2012
148 - Coro::EV I/O watchers were not interruptible by exceptions
149 (Coro::State::throw) (testcase by sten).
150 - ->throw now puts threads into the ready queue, as this seems to
151 be expected by existing code, and code that doesn't cope with spurious
152 wakeups needs fixing anyway.
153 - use fd -1 in mmap.
154 - cast I32 to int in error message printf.
155 - warn about broken so-called "hardened" kernels.
156
1576.08 Fri Apr 13 12:05:47 CEST 2012
158 - be more aggressive about exiting like perl does - formerly,
159 exiting from the non-main thread would not execute END blocks.
160
1616.07 Fri Nov 11 21:21:48 CET 2011
162 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
163 - update ecb.h.
164
1656.06 Mon Aug 8 23:59:48 CEST 2011
166 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
167 backend, which is an order of magnitude slower. unfortunately, cygwins
168 pthread implementation isn't very complete either, so allocate the stack
169 twice just to be sure.
170 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
171 in its ithreaded perl - best recompile cygwin and use the 'w'indows
172 backend for much better performance. also disable ithreads for
173 even better performance...).
174
1756.05 Thu Aug 4 21:36:36 CEST 2011
176 - blush, condvar values would not be propagated from send to recv anymore
177 (reported by Chip Salzenberg).
178 - use exponential increase for the readline buffer length in
179 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
180
1816.04 Wed Aug 3 17:15:45 CEST 2011
182 - use even more efficient and more compatible condvars for
183 compatibility to AnyEvent 6.x :)
184 - more inconsequential ecb.h updates.
185
1866.03 Wed Aug 3 11:41:30 CEST 2011
187 - change how Coro patches AnyEvent condvars for compatibility to
188 AnyEvent 6.x.
189 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
190 hasn't been included.
191
1926.02 Wed Jul 13 04:35:19 CEST 2011
193 - "improve portability to Gentoo" - gentoo manages to put perl variables
194 in memory areas that are farther than 2gb apart, which the jit couldn't
195 handle and barfed. now it's just a bit slower on gentoo and similar
196 systems.
197
1986.01 Sun Jul 3 12:31:14 CEST 2011
199 - workarounds are good, but the test for whether pthreads are used
200 was not good. this one should be better.
201 - check differently whether gcc generates cfi instructions itself.
202
2036.0 Wed Jun 29 19:43:35 CEST 2011
204 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
205 destroyed and cleaned up automatically (e.g. async { schedule }).
206 - implement a JIT compiler for part of the thread switch code,
207 which gives a 50% speed improvement on threaded perls, and
208 about 4% on non-threaded perls (so threaded perls now finally
209 reach about half the speed of non-threaded perls).
210 - slightly modernise Coro::Intro, add section about rouse functions.
211 - avoid DEFSV and ERRSV, giving another 10% improvement
212 in thread switching.
213 - Coro::State->is_destroyed is now called is_zombie.
214 - implement a Coro->safe_cancel method that might fail, but
215 cancels in a "safer" way if it succeeds.
216 - add preliminary support for DEBUGGING perls.
217 - get rid of two hash-accesses when initialising a new Coro - this
218 speeds up coro creation by almost a factor of two.
219 - croak when a coro that is being cancelled tries to block
220 (e.g. while executing a guard block), instead of crashing or
221 deadlocking.
222 - use a more robust and also faster method to identify Coro::State
223 objects - speeds up everything a bit.
224 - implement Coro->cancel in XS for a 20% speed improvement, and to
225 be able to implement mutual cancellation.
226 - speed up context switches by a percent or two by more efficiently
227 allocating context stack entries.
228 - implement Coro->join and Coro->on_destroy in XS for a speedup and
229 a reduction in memory use.
230 - cancelling a coro while it itself is cancelling another coro is
231 now supported and working, instead of triggering an assertion.
232 - be a bit more crash-resistant when calling (buggy) on_destroy
233 callbacks (best effort).
234 - move on_destroy into the slf_frame, to allow extension slf
235 functions to have destructors.
236 - get rid if coro refcounting - simply crash in other interpreter
237 threads by nulling the pointers on clone.
238 - simplify warn/die hook handling when loading Coro - the convoluted
239 logic seems to be no longer neccessary.
240 - use libecb instead of our own home-grown gcc hacks.
241 - document alternatives to Coro::LWP. Please use them :)
242 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
243 sigaltstack. Really. wine suffers from it, erlang suffers from it,
244 and it's known since at least 2006.
245
2465.372 Wed Feb 23 06:14:30 CET 2011
247 - apparently mingw doesn't provide a working gettimeofday, try to
248 work around that by relying on Time::HiRes (indirectly brought to
249 my attention by Max Maischein).
250 - fix some portability issues when Time::HiRes was used.
251
2525.371 Mon Feb 21 14:36:08 CET 2011
253 - backport to windows process emulation code again.
254
2555.37 Sat Feb 19 07:49:44 CET 2011
256 - add a big "Coro thread life cycle" section to "man Coro".
257 - try a tentative workaround against the breakage that 5.13 has
258 introduced without depreciation period. sigh.
259 - no longer use Time::HiRes if gettimeofday is available, which
260 saves quite a lot of memory.
8 261
95.36 Sun Feb 13 05:33:41 CET 2011 2625.36 Sun Feb 13 05:33:41 CET 2011
10 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore, 263 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
11 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new" 264 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
12 method call. 265 method call.
… …
314 - ->throw is now supported on Coro::State objects. 567 - ->throw is now supported on Coro::State objects.
315 - clean up cctx creation code a bit. 568 - clean up cctx creation code a bit.
316 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 569 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
317 570
3184.9 Sat Nov 8 17:45:27 CET 2008 5714.9 Sat Nov 8 17:45:27 CET 2008
319 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 572 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
320 - (libcoro) no longer leak threads in the experimental pthread backend, 573 - (libcoro) no longer leak threads in the experimental pthread backend,
321 also speed it up considerably. 574 also speed it up considerably.
322 - (libcoro) do not rely on makecontext passing void *'s unscathed. 575 - (libcoro) do not rely on makecontext passing void *'s unscathed.
323 - fix compiletime dependencies on libcoro in the Makefile. 576 - fix compiletime dependencies on libcoro in the Makefile.
324 - cctx_count wasn't always updated properly. 577 - cctx_count wasn't always updated properly.
… …
386 - remove debugging code related to MgPV_nolen_const, also try to 639 - remove debugging code related to MgPV_nolen_const, also try to
387 make it compile with perl 5.8.6 (yes, apple apparently loves 640 make it compile with perl 5.8.6 (yes, apple apparently loves
388 outdated software). Reported by John S. 641 outdated software). Reported by John S.
389 642
3904.744 Tue Jul 8 22:06:35 CEST 2008 6434.744 Tue Jul 8 22:06:35 CEST 2008
391 - correctly provide default DIE/WARN handlers as documented. 644 - correctly provide default DIE/WARN handlers as documented.
392 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 645 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
393 implementations inside perl work fine for us. 646 implementations inside perl work fine for us.
394 647
3954.743 Mon Jun 16 00:21:57 CEST 2008 6484.743 Mon Jun 16 00:21:57 CEST 2008
396 - when using Coro::EV without running EV::loop it could 649 - when using Coro::EV without running EV::loop it could
… …
418 - sprinkle "no warnings" freely over everything, also suppress 671 - sprinkle "no warnings" freely over everything, also suppress
419 warnings for some other modules. 672 warnings for some other modules.
420 - fix typo in WSAEWOULDBLOCK. 673 - fix typo in WSAEWOULDBLOCK.
421 674
4224.72 Sun May 25 05:14:36 CEST 2008 6754.72 Sun May 25 05:14:36 CEST 2008
423 - tweak META.yaml a bit, unfortunately, there is no documented way 676 - tweak META.yaml a bit, unfortunately, there is no documented way
424 to have optional dependencies with CPAN. doh :( 677 to have optional dependencies with CPAN. doh :(
425 - avoid running some tests on windows because they would fail due to 678 - avoid running some tests on windows because they would fail due to
426 perl bug (broken fork, broken pipes...). 679 perl bug (broken fork, broken pipes...).
427 - work around perl on windows bugs where perl returns undocumented 680 - work around perl on windows bugs where perl returns undocumented
428 error codes for sysread, syswrite etc. by taking advantage 681 error codes for sysread, syswrite etc. by taking advantage
… …
432 - use unix domain sockets in testsuite to work around 685 - use unix domain sockets in testsuite to work around
433 common perl implementation bugs on widows (they are emulated by 686 common perl implementation bugs on widows (they are emulated by
434 tcp sockets on windows. ugh.) 687 tcp sockets on windows. ugh.)
435 688
4364.71 Sat May 24 20:01:27 CEST 2008 6894.71 Sat May 24 20:01:27 CEST 2008
437 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 690 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
438 - take advantage of async name resolution of AnyEvent::Util. 691 - take advantage of async name resolution of AnyEvent::Util.
439 - work around brutal inet_aton override in Coro::LWP. 692 - work around brutal inet_aton override in Coro::LWP.
440 - take advantage of the readyhook in Coro::EV, for smoother 693 - take advantage of the readyhook in Coro::EV, for smoother
441 scheduling. 694 scheduling.
442 695
4434.7 Sun May 11 00:32:19 CEST 2008 6964.7 Sun May 11 00:32:19 CEST 2008
444 - completely reworked the Coro manpage. 697 - completely reworked the Coro manpage.
445 - added Coro::AnyEvent, generic event loop integration. 698 - added Coro::AnyEvent, generic event loop integration.
446 - implement cancel, ready and kill commands in Coro::Debug. 699 - implement cancel, ready and kill commands in Coro::Debug.
447 - document find_coro in Coro::Debug. 700 - document find_coro in Coro::Debug.
448 - incompatible change: rename has_stack to has_cctx. 701 - incompatible change: rename has_stack to has_cctx.
449 - Coro::AIO and Coro::BDB no longer force event model detection, 702 - Coro::AIO and Coro::BDB no longer force event model detection,
450 use AnyEvent::AIO and AnyEvent::BDB. 703 use AnyEvent::AIO and AnyEvent::BDB.
… …
574 827
5754.11 Thu Oct 11 02:40:24 CEST 2007 8284.11 Thu Oct 11 02:40:24 CEST 2007
576 - port to threaded perls. 829 - port to threaded perls.
577 830
5784.1 Thu Oct 11 02:38:16 CEST 2007 8314.1 Thu Oct 11 02:38:16 CEST 2007
579 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 832 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
580 be local to each coro (see Coro::State). 833 be local to each coro (see Coro::State).
581 - incompatible change: for very deep reasons, cede and cede_notself 834 - incompatible change: for very deep reasons, cede and cede_notself
582 cannot return anything, so nothing will be returned. 835 cannot return anything, so nothing will be returned.
583 - possibly bring back 5.10 compatibility (untested). 836 - possibly bring back 5.10 compatibility (untested).
584 - work around stupid (and wrong) warning on 5.10 :(. 837 - work around stupid (and wrong) warning on 5.10 :(.
… …
656 but stateful semantics. 909 but stateful semantics.
657 - fixed a lot of typos in Coro.pm (patch submitted by David 910 - fixed a lot of typos in Coro.pm (patch submitted by David
658 Steinbrunner, which applied flawlessly). 911 Steinbrunner, which applied flawlessly).
659 912
6603.6 Sat Apr 14 17:13:31 CEST 2007 9133.6 Sat Apr 14 17:13:31 CEST 2007
661 - added some bugfixes to get eg/myhttpd working again. 914 - added some bugfixes to get eg/myhttpd working again.
662 - added Coro::Storable for often-cede'ing freeze/thaw. 915 - added Coro::Storable for often-cede'ing freeze/thaw.
663 - try to do a clean exit when a coroutine calls exit 916 - try to do a clean exit when a coroutine calls exit
664 (EXPERIMENTAL). 917 (EXPERIMENTAL).
665 - got rid of indirect call through _coro_init. 918 - got rid of indirect call through _coro_init.
666 - updated the partly antique examples in eg/ to 919 - updated the partly antique examples in eg/ to

Diff Legend

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