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

Comparing Coro/Changes (file contents):
Revision 1.536 by root, Wed Apr 14 01:57:34 2010 UTC vs.
Revision 1.644 by root, Mon Jun 29 23:49:34 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
106.45 Tue Jun 30 01:40:08 CEST 2015
11 - restore portability to perl 5.16 and below.
12
136.44 Tue Jun 30 00:41:54 CEST 2015
14 - avoid segfaulting (or worse) when tracing a canceled thread.
15 - the Event module silently broke it's public hook API, causing
16 failures on perls compiled with -Duselongdouble. Requiring
17 the latest version of Event because I don't know which release
18 changed this (it's not mentioned in the ChangeLog).
19 - allow xs level enterleave hooks via CoroAPI.
20
216.43 Thu Jun 4 15:38:14 CEST 2015
22 - use stability canary.
23 - port to stableperl-5.22.0-1.001.
24 - update libecb.
25
266.42 Wed Feb 11 20:29:52 CET 2015
27 - Coro::SemaphoreSet->try did not actually work (analyzed by
28 SATO Kentaro).
29 - upgrade libecb to be C11 compliant.
30
316.41 Sat Sep 6 22:08:46 CEST 2014
32 - restore portability to perl 5.8.x.
33 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
34 up some memory, but fixes "use feature" and similar modules
35 when used inside a Coro.
36 - allow zero as argument to Coro::Channel to mean the same thing
37 as no argument. this works with older versions as well,
38 but wasn't legal until now.
39 - slightly better c header file detection.
40
416.39 Mon Jun 2 00:00:08 CEST 2014
42 - work around more incompatible changes in 5.20.
43
446.38 Sun Jun 1 21:54:23 CEST 2014
45 - check that perl slots actually have enough space to hold
46 interpreter variables.
47 - untested port to perl 5.19 (64 bit tmps indices) (reported
48 by Andreas König).
49 - croak when cancel is called without a thread context.
50
516.37 Tue Mar 4 13:27:33 CET 2014
52 - *sigh*, removed leftover debugging code from debugging a
53 perl bug, of all things.
54
556.36 Tue Mar 4 07:11:59 CET 2014
56 - semaphores would not clear the destroy hook when interrupted
57 by ->throw, causing segfaults or worse.
58 - ->throw on a thread waiting for a semaphore did not acquire
59 the semaphore, but also didn't wake up other waiters,
60 possibly causing a deadlock.
61 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
62 to report a full stacktrace, which should help find out
63 where the actual call is.
64 - "atomically" destroy data in slf_destroy, because it is
65 the right thing to do, just in case.
66 - disable ecb.h libm dependency, to compile on ancient systems
67 or under adverse conditions.
68
696.33 Mon Nov 18 11:26:27 CET 2013
70 - do not crash when freeing padlists with holes (in 5.18).
71 - tentative SVt_BIND 5.19 port/fix.
72
736.32 Tue Nov 5 15:35:35 CET 2013
74 - use a new algorithm to derive padlists for perl 5.18. The old
75 one could lead to 0-pointer accesses inside perl (reported
76 by Darin McBride).
77
786.31 Thu May 9 07:39:48 CEST 2013
79 - Coro::AIO requests would crash if the thread was ready'd
80 while the request was ongoing.
81
826.29 Wed May 8 02:55:18 CEST 2013
83 - when an on_destroy handler destructs the coro currently being
84 destructed a perl scalar could be accessed after being freed,
85 likely causing a crash.
86
876.28 Wed Mar 6 06:58:02 CET 2013
88 - clean remnants of existing __DIE__ and __WARN__ handlers so
89 they lose their magic and will not cause segfaults later
90 (testcase by Andrey Sagulin).
91 - improved Coro::State documentation a bit.
92 - Coro::Debug::command now flushes the output.
93 - add hack detection code for x32 abi, because the braindead slugs
94 who designed that made it look exactly like x86_64 without
95 providing proper compile time symbols to test for it. as a result,
96 this detection cannot work reliably.
97 - valgrind stack registering was broken.
98 - do not rely on Time::HiRes anymore in Coro::Debug.
99
1006.23 Fri Dec 7 23:36:37 CET 2012
101 - use experimental fiber implementation on native windows
102 perls.
103 - use sizeof (void *) as multiplication factor for stack sizes,
104 to accomodate the totally braindamaged microsoft 64 bit "os".
105 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
106 activeperl 5.16 crashes when PerlIO_define_layer is called due
107 to some bug in the perl dll, strawberry perl at least passes
108 the testsuite.
109 - implement Coro::Handle->peeraddr/host/port, for slightly
110 improved compatibility with LWP.
111 - implement 5.17 compatibility by almost blindly applying a
112 good-looking patch by Father Chrysostomos.
113 - move stack management functions into libcoro 3.
114 - libcoro version 3 "released".
115 - support magic values as timed_io_once args.
116 - recommend AnyEvent 7+ or EV 4+, also require EV
117 version 4 or newer for Coro::EV.
118
1196.10 Tue Oct 9 01:14:27 CEST 2012
120 - updated ecb.h, it had a typo that caused it to not compile on many
121 big endian systems (reported by many people).
122 - disable memory fences in ecb.h to improve portability.
123
1246.09 Sat Oct 6 23:25:02 CEST 2012
125 - Coro::EV I/O watchers were not interruptible by exceptions
126 (Coro::State::throw) (testcase by sten).
127 - ->throw now puts threads into the ready queue, as this seems to
128 be expected by existing code, and code that doesn't cope with spurious
129 wakeups needs fixing anyway.
130 - use fd -1 in mmap.
131 - cast I32 to int in error message printf.
132 - warn about broken so-called "hardened" kernels.
133
1346.08 Fri Apr 13 12:05:47 CEST 2012
135 - be more aggressive about exiting like perl does - formerly,
136 exiting from the non-main thread would not execute END blocks.
137
1386.07 Fri Nov 11 21:21:48 CET 2011
139 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
140 - update ecb.h.
141
1426.06 Mon Aug 8 23:59:48 CEST 2011
143 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
144 backend, which is an order of magnitude slower. unfortunately, cygwins
145 pthread implementation isn't very complete either, so allocate the stack
146 twice just to be sure.
147 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
148 in its ithreaded perl - best recompile cygwin and use the 'w'indows
149 backend for much better performance. also disable ithreads for
150 even better performance...).
151
1526.05 Thu Aug 4 21:36:36 CEST 2011
153 - blush, condvar values would not be propagated from send to recv anymore
154 (reported by Chip Salzenberg).
155 - use exponential increase for the readline buffer length in
156 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
157
1586.04 Wed Aug 3 17:15:45 CEST 2011
159 - use even more efficient and more compatible condvars for
160 compatibility to AnyEvent 6.x :)
161 - more inconsequential ecb.h updates.
162
1636.03 Wed Aug 3 11:41:30 CEST 2011
164 - change how Coro patches AnyEvent condvars for compatibility to
165 AnyEvent 6.x.
166 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
167 hasn't been included.
168
1696.02 Wed Jul 13 04:35:19 CEST 2011
170 - "improve portability to Gentoo" - gentoo manages to put perl variables
171 in memory areas that are farther than 2gb apart, which the jit couldn't
172 handle and barfed. now it's just a bit slower on gentoo and similar
173 systems.
174
1756.01 Sun Jul 3 12:31:14 CEST 2011
176 - workarounds are good, but the test for whether pthreads are used
177 was not good. this one should be better.
178 - check differently whether gcc generates cfi instructions itself.
179
1806.0 Wed Jun 29 19:43:35 CEST 2011
181 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
182 destroyed and cleaned up automatically (e.g. async { schedule }).
183 - implement a JIT compiler for part of the thread switch code,
184 which gives a 50% speed improvement on threaded perls, and
185 about 4% on non-threaded perls (so threaded perls now finally
186 reach about half the speed of non-threaded perls).
187 - slightly modernise Coro::Intro, add section about rouse functions.
188 - avoid DEFSV and ERRSV, giving another 10% improvement
189 in thread switching.
190 - Coro::State->is_destroyed is now called is_zombie.
191 - implement a Coro->safe_cancel method that might fail, but
192 cancels in a "safer" way if it succeeds.
193 - add preliminary support for DEBUGGING perls.
194 - get rid of two hash-accesses when initialising a new Coro - this
195 speeds up coro creation by almost a factor of two.
196 - croak when a coro that is being cancelled tries to block
197 (e.g. while executing a guard block), instead of crashing or
198 deadlocking.
199 - use a more robust and also faster method to identify Coro::State
200 objects - speeds up everything a bit.
201 - implement Coro->cancel in XS for a 20% speed improvement, and to
202 be able to implement mutual cancellation.
203 - speed up context switches by a percent or two by more efficiently
204 allocating context stack entries.
205 - implement Coro->join and Coro->on_destroy in XS for a speedup and
206 a reduction in memory use.
207 - cancelling a coro while it itself is cancelling another coro is
208 now supported and working, instead of triggering an assertion.
209 - be a bit more crash-resistant when calling (buggy) on_destroy
210 callbacks (best effort).
211 - move on_destroy into the slf_frame, to allow extension slf
212 functions to have destructors.
213 - get rid if coro refcounting - simply crash in other interpreter
214 threads by nulling the pointers on clone.
215 - simplify warn/die hook handling when loading Coro - the convoluted
216 logic seems to be no longer neccessary.
217 - use libecb instead of our own home-grown gcc hacks.
218 - document alternatives to Coro::LWP. Please use them :)
219 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
220 sigaltstack. Really. wine suffers from it, erlang suffers from it,
221 and it's known since at least 2006.
222
2235.372 Wed Feb 23 06:14:30 CET 2011
224 - apparently mingw doesn't provide a working gettimeofday, try to
225 work around that by relying on Time::HiRes (indirectly brought to
226 my attention by Max Maischein).
227 - fix some portability issues when Time::HiRes was used.
228
2295.371 Mon Feb 21 14:36:08 CET 2011
230 - backport to windows process emulation code again.
231
2325.37 Sat Feb 19 07:49:44 CET 2011
233 - add a big "Coro thread life cycle" section to "man Coro".
234 - try a tentative workaround against the breakage that 5.13 has
235 introduced without depreciation period. sigh.
236 - no longer use Time::HiRes if gettimeofday is available, which
237 saves quite a lot of memory.
238
2395.36 Sun Feb 13 05:33:41 CET 2011
240 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
241 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
242 method call.
243 - undocument Coro::Timer::sleep and obsolete whole module.
244 - optimise Coro::Timer::timeout memory and cpu usage.
245 - slightly updated Coro::Intro for recent changes.
246 - do not initialise PL_dirty anymore.
247
2485.25 Thu Nov 11 01:08:39 CET 2010
249 - try a different approach on netbsd - netbsd 5 finally has marginally
250 working pthreads, but still broken ucontext/sigaltstack.
251 - openbsd 4.8 finally got their act together, Coro works out of the box
252 with asm, setjmp and pthreads (no change, just informational).
253
2545.24 Sat Oct 23 11:27:12 CEST 2010
255 - port to the EV 4.0 API.
256 - work around bugs in mingw32, making strawberry perl work
257 out of the box.
258 - correctly modify Coro::AIO function prototypes
259 so that they reflect the "no optional parameters" rule.
260 - "ported" libcoro to C++.
261
2625.23 Mon May 17 18:50:42 CEST 2010
263 - be more resistant to ordering changes when initialising
264 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
265 Waldorf).
9 - document that perl 5.12 deliberately removed support for cloning. 266 - document that perl 5.12 deliberately removed support for cloning.
10 267
115.22 Wed Apr 14 03:55:35 CEST 2010 2685.22 Wed Apr 14 03:55:35 CEST 2010
12 - correctly return udnef on errors in Coro::Handle::read/write 269 - correctly return udnef on errors in Coro::Handle::read/write
13 (testcase by Marc Wims). 270 (testcase by Marc Mims).
14 - convert Coro::Util into a "perl compatibility wrapper" - the functions 271 - convert Coro::Util into a "perl compatibility wrapper" - the functions
15 are less useful now, but are drop-in replacements for existing 272 are less useful now, but are drop-in replacements for existing
16 functions, listing better alternatives in the documentation. This also 273 functions, listing better alternatives in the documentation. This also
17 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton 274 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
18 with Coro::Util::inet_aton. 275 with Coro::Util::inet_aton.
… …
287 - ->throw is now supported on Coro::State objects. 544 - ->throw is now supported on Coro::State objects.
288 - clean up cctx creation code a bit. 545 - clean up cctx creation code a bit.
289 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 546 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
290 547
2914.9 Sat Nov 8 17:45:27 CET 2008 5484.9 Sat Nov 8 17:45:27 CET 2008
292 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 549 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
293 - (libcoro) no longer leak threads in the experimental pthread backend, 550 - (libcoro) no longer leak threads in the experimental pthread backend,
294 also speed it up considerably. 551 also speed it up considerably.
295 - (libcoro) do not rely on makecontext passing void *'s unscathed. 552 - (libcoro) do not rely on makecontext passing void *'s unscathed.
296 - fix compiletime dependencies on libcoro in the Makefile. 553 - fix compiletime dependencies on libcoro in the Makefile.
297 - cctx_count wasn't always updated properly. 554 - cctx_count wasn't always updated properly.
… …
359 - remove debugging code related to MgPV_nolen_const, also try to 616 - remove debugging code related to MgPV_nolen_const, also try to
360 make it compile with perl 5.8.6 (yes, apple apparently loves 617 make it compile with perl 5.8.6 (yes, apple apparently loves
361 outdated software). Reported by John S. 618 outdated software). Reported by John S.
362 619
3634.744 Tue Jul 8 22:06:35 CEST 2008 6204.744 Tue Jul 8 22:06:35 CEST 2008
364 - correctly provide default DIE/WARN handlers as documented. 621 - correctly provide default DIE/WARN handlers as documented.
365 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 622 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
366 implementations inside perl work fine for us. 623 implementations inside perl work fine for us.
367 624
3684.743 Mon Jun 16 00:21:57 CEST 2008 6254.743 Mon Jun 16 00:21:57 CEST 2008
369 - when using Coro::EV without running EV::loop it could 626 - when using Coro::EV without running EV::loop it could
… …
391 - sprinkle "no warnings" freely over everything, also suppress 648 - sprinkle "no warnings" freely over everything, also suppress
392 warnings for some other modules. 649 warnings for some other modules.
393 - fix typo in WSAEWOULDBLOCK. 650 - fix typo in WSAEWOULDBLOCK.
394 651
3954.72 Sun May 25 05:14:36 CEST 2008 6524.72 Sun May 25 05:14:36 CEST 2008
396 - tweak META.yaml a bit, unfortunately, there is no documented way 653 - tweak META.yaml a bit, unfortunately, there is no documented way
397 to have optional dependencies with CPAN. doh :( 654 to have optional dependencies with CPAN. doh :(
398 - avoid running some tests on windows because they would fail due to 655 - avoid running some tests on windows because they would fail due to
399 perl bug (broken fork, broken pipes...). 656 perl bug (broken fork, broken pipes...).
400 - work around perl on windows bugs where perl returns undocumented 657 - work around perl on windows bugs where perl returns undocumented
401 error codes for sysread, syswrite etc. by taking advantage 658 error codes for sysread, syswrite etc. by taking advantage
… …
405 - use unix domain sockets in testsuite to work around 662 - use unix domain sockets in testsuite to work around
406 common perl implementation bugs on widows (they are emulated by 663 common perl implementation bugs on widows (they are emulated by
407 tcp sockets on windows. ugh.) 664 tcp sockets on windows. ugh.)
408 665
4094.71 Sat May 24 20:01:27 CEST 2008 6664.71 Sat May 24 20:01:27 CEST 2008
410 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 667 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
411 - take advantage of async name resolution of AnyEvent::Util. 668 - take advantage of async name resolution of AnyEvent::Util.
412 - work around brutal inet_aton override in Coro::LWP. 669 - work around brutal inet_aton override in Coro::LWP.
413 - take advantage of the readyhook in Coro::EV, for smoother 670 - take advantage of the readyhook in Coro::EV, for smoother
414 scheduling. 671 scheduling.
415 672
4164.7 Sun May 11 00:32:19 CEST 2008 6734.7 Sun May 11 00:32:19 CEST 2008
417 - completely reworked the Coro manpage. 674 - completely reworked the Coro manpage.
418 - added Coro::AnyEvent, generic event loop integration. 675 - added Coro::AnyEvent, generic event loop integration.
419 - implement cancel, ready and kill commands in Coro::Debug. 676 - implement cancel, ready and kill commands in Coro::Debug.
420 - document find_coro in Coro::Debug. 677 - document find_coro in Coro::Debug.
421 - incompatible change: rename has_stack to has_cctx. 678 - incompatible change: rename has_stack to has_cctx.
422 - Coro::AIO and Coro::BDB no longer force event model detection, 679 - Coro::AIO and Coro::BDB no longer force event model detection,
423 use AnyEvent::AIO and AnyEvent::BDB. 680 use AnyEvent::AIO and AnyEvent::BDB.
… …
547 804
5484.11 Thu Oct 11 02:40:24 CEST 2007 8054.11 Thu Oct 11 02:40:24 CEST 2007
549 - port to threaded perls. 806 - port to threaded perls.
550 807
5514.1 Thu Oct 11 02:38:16 CEST 2007 8084.1 Thu Oct 11 02:38:16 CEST 2007
552 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 809 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
553 be local to each coro (see Coro::State). 810 be local to each coro (see Coro::State).
554 - incompatible change: for very deep reasons, cede and cede_notself 811 - incompatible change: for very deep reasons, cede and cede_notself
555 cannot return anything, so nothing will be returned. 812 cannot return anything, so nothing will be returned.
556 - possibly bring back 5.10 compatibility (untested). 813 - possibly bring back 5.10 compatibility (untested).
557 - work around stupid (and wrong) warning on 5.10 :(. 814 - work around stupid (and wrong) warning on 5.10 :(.
… …
629 but stateful semantics. 886 but stateful semantics.
630 - fixed a lot of typos in Coro.pm (patch submitted by David 887 - fixed a lot of typos in Coro.pm (patch submitted by David
631 Steinbrunner, which applied flawlessly). 888 Steinbrunner, which applied flawlessly).
632 889
6333.6 Sat Apr 14 17:13:31 CEST 2007 8903.6 Sat Apr 14 17:13:31 CEST 2007
634 - added some bugfixes to get eg/myhttpd working again. 891 - added some bugfixes to get eg/myhttpd working again.
635 - added Coro::Storable for often-cede'ing freeze/thaw. 892 - added Coro::Storable for often-cede'ing freeze/thaw.
636 - try to do a clean exit when a coroutine calls exit 893 - try to do a clean exit when a coroutine calls exit
637 (EXPERIMENTAL). 894 (EXPERIMENTAL).
638 - got rid of indirect call through _coro_init. 895 - got rid of indirect call through _coro_init.
639 - updated the partly antique examples in eg/ to 896 - updated the partly antique examples in eg/ to

Diff Legend

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