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

Comparing Coro/Changes (file contents):
Revision 1.534 by root, Sun Apr 11 19:23:18 2010 UTC vs.
Revision 1.641 by root, Sun Jun 21 17:02:49 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)?
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?
7 9
8TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 10 - avoid segfaulting (or worse) when tracing a canceled thread.
11 - the Event module silently broke it's public hook API, causing
12 failures on perls compiled with -Duselongdouble. Requiring
13 the latest version of Event because I don't know which release
14 changed this (it's not mentioned in the ChangeLog).
9 15
166.43 Thu Jun 4 15:38:14 CEST 2015
17 - use stability canary.
18 - port to stableperl-5.22.0-1.001.
19 - update libecb.
20
216.42 Wed Feb 11 20:29:52 CET 2015
22 - Coro::SemaphoreSet->try did not actually work (analyzed by
23 SATO Kentaro).
24 - upgrade libecb to be C11 compliant.
25
266.41 Sat Sep 6 22:08:46 CEST 2014
27 - restore portability to perl 5.8.x.
28 - give new Coro's a valid GvHV(PL_hintgv) - this is slow and takes
29 up some memory, but fixes "use feature" and similar modules
30 when used inside a Coro.
31 - allow zero as argument to Coro::Channel to mean the same thing
32 as no argument. this works with older versions as well,
33 but wasn't legal until now.
34 - slightly better c header file detection.
35
366.39 Mon Jun 2 00:00:08 CEST 2014
37 - work around more incompatible changes in 5.20.
38
396.38 Sun Jun 1 21:54:23 CEST 2014
40 - check that perl slots actually have enough space to hold
41 interpreter variables.
42 - untested port to perl 5.19 (64 bit tmps indices) (reported
43 by Andreas König).
44 - croak when cancel is called without a thread context.
45
466.37 Tue Mar 4 13:27:33 CET 2014
47 - *sigh*, removed leftover debugging code from debugging a
48 perl bug, of all things.
49
506.36 Tue Mar 4 07:11:59 CET 2014
51 - semaphores would not clear the destroy hook when interrupted
52 by ->throw, causing segfaults or worse.
53 - ->throw on a thread waiting for a semaphore did not acquire
54 the semaphore, but also didn't wake up other waiters,
55 possibly causing a deadlock.
56 - "FATAL: $Coro::IDLE blocked itself" will now use Carp::confess
57 to report a full stacktrace, which should help find out
58 where the actual call is.
59 - "atomically" destroy data in slf_destroy, because it is
60 the right thing to do, just in case.
61 - disable ecb.h libm dependency, to compile on ancient systems
62 or under adverse conditions.
63
646.33 Mon Nov 18 11:26:27 CET 2013
65 - do not crash when freeing padlists with holes (in 5.18).
66 - tentative SVt_BIND 5.19 port/fix.
67
686.32 Tue Nov 5 15:35:35 CET 2013
69 - use a new algorithm to derive padlists for perl 5.18. The old
70 one could lead to 0-pointer accesses inside perl (reported
71 by Darin McBride).
72
736.31 Thu May 9 07:39:48 CEST 2013
74 - Coro::AIO requests would crash if the thread was ready'd
75 while the request was ongoing.
76
776.29 Wed May 8 02:55:18 CEST 2013
78 - when an on_destroy handler destructs the coro currently being
79 destructed a perl scalar could be accessed after being freed,
80 likely causing a crash.
81
826.28 Wed Mar 6 06:58:02 CET 2013
83 - clean remnants of existing __DIE__ and __WARN__ handlers so
84 they lose their magic and will not cause segfaults later
85 (testcase by Andrey Sagulin).
86 - improved Coro::State documentation a bit.
87 - Coro::Debug::command now flushes the output.
88 - add hack detection code for x32 abi, because the braindead slugs
89 who designed that made it look exactly like x86_64 without
90 providing proper compile time symbols to test for it. as a result,
91 this detection cannot work reliably.
92 - valgrind stack registering was broken.
93 - do not rely on Time::HiRes anymore in Coro::Debug.
94
956.23 Fri Dec 7 23:36:37 CET 2012
96 - use experimental fiber implementation on native windows
97 perls.
98 - use sizeof (void *) as multiplication factor for stack sizes,
99 to accomodate the totally braindamaged microsoft 64 bit "os".
100 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
101 activeperl 5.16 crashes when PerlIO_define_layer is called due
102 to some bug in the perl dll, strawberry perl at least passes
103 the testsuite.
104 - implement Coro::Handle->peeraddr/host/port, for slightly
105 improved compatibility with LWP.
106 - implement 5.17 compatibility by almost blindly applying a
107 good-looking patch by Father Chrysostomos.
108 - move stack management functions into libcoro 3.
109 - libcoro version 3 "released".
110 - support magic values as timed_io_once args.
111 - recommend AnyEvent 7+ or EV 4+, also require EV
112 version 4 or newer for Coro::EV.
113
1146.10 Tue Oct 9 01:14:27 CEST 2012
115 - updated ecb.h, it had a typo that caused it to not compile on many
116 big endian systems (reported by many people).
117 - disable memory fences in ecb.h to improve portability.
118
1196.09 Sat Oct 6 23:25:02 CEST 2012
120 - Coro::EV I/O watchers were not interruptible by exceptions
121 (Coro::State::throw) (testcase by sten).
122 - ->throw now puts threads into the ready queue, as this seems to
123 be expected by existing code, and code that doesn't cope with spurious
124 wakeups needs fixing anyway.
125 - use fd -1 in mmap.
126 - cast I32 to int in error message printf.
127 - warn about broken so-called "hardened" kernels.
128
1296.08 Fri Apr 13 12:05:47 CEST 2012
130 - be more aggressive about exiting like perl does - formerly,
131 exiting from the non-main thread would not execute END blocks.
132
1336.07 Fri Nov 11 21:21:48 CET 2011
134 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
135 - update ecb.h.
136
1376.06 Mon Aug 8 23:59:48 CEST 2011
138 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
139 backend, which is an order of magnitude slower. unfortunately, cygwins
140 pthread implementation isn't very complete either, so allocate the stack
141 twice just to be sure.
142 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
143 in its ithreaded perl - best recompile cygwin and use the 'w'indows
144 backend for much better performance. also disable ithreads for
145 even better performance...).
146
1476.05 Thu Aug 4 21:36:36 CEST 2011
148 - blush, condvar values would not be propagated from send to recv anymore
149 (reported by Chip Salzenberg).
150 - use exponential increase for the readline buffer length in
151 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
152
1536.04 Wed Aug 3 17:15:45 CEST 2011
154 - use even more efficient and more compatible condvars for
155 compatibility to AnyEvent 6.x :)
156 - more inconsequential ecb.h updates.
157
1586.03 Wed Aug 3 11:41:30 CEST 2011
159 - change how Coro patches AnyEvent condvars for compatibility to
160 AnyEvent 6.x.
161 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
162 hasn't been included.
163
1646.02 Wed Jul 13 04:35:19 CEST 2011
165 - "improve portability to Gentoo" - gentoo manages to put perl variables
166 in memory areas that are farther than 2gb apart, which the jit couldn't
167 handle and barfed. now it's just a bit slower on gentoo and similar
168 systems.
169
1706.01 Sun Jul 3 12:31:14 CEST 2011
171 - workarounds are good, but the test for whether pthreads are used
172 was not good. this one should be better.
173 - check differently whether gcc generates cfi instructions itself.
174
1756.0 Wed Jun 29 19:43:35 CEST 2011
176 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
177 destroyed and cleaned up automatically (e.g. async { schedule }).
178 - implement a JIT compiler for part of the thread switch code,
179 which gives a 50% speed improvement on threaded perls, and
180 about 4% on non-threaded perls (so threaded perls now finally
181 reach about half the speed of non-threaded perls).
182 - slightly modernise Coro::Intro, add section about rouse functions.
183 - avoid DEFSV and ERRSV, giving another 10% improvement
184 in thread switching.
185 - Coro::State->is_destroyed is now called is_zombie.
186 - implement a Coro->safe_cancel method that might fail, but
187 cancels in a "safer" way if it succeeds.
188 - add preliminary support for DEBUGGING perls.
189 - get rid of two hash-accesses when initialising a new Coro - this
190 speeds up coro creation by almost a factor of two.
191 - croak when a coro that is being cancelled tries to block
192 (e.g. while executing a guard block), instead of crashing or
193 deadlocking.
194 - use a more robust and also faster method to identify Coro::State
195 objects - speeds up everything a bit.
196 - implement Coro->cancel in XS for a 20% speed improvement, and to
197 be able to implement mutual cancellation.
198 - speed up context switches by a percent or two by more efficiently
199 allocating context stack entries.
200 - implement Coro->join and Coro->on_destroy in XS for a speedup and
201 a reduction in memory use.
202 - cancelling a coro while it itself is cancelling another coro is
203 now supported and working, instead of triggering an assertion.
204 - be a bit more crash-resistant when calling (buggy) on_destroy
205 callbacks (best effort).
206 - move on_destroy into the slf_frame, to allow extension slf
207 functions to have destructors.
208 - get rid if coro refcounting - simply crash in other interpreter
209 threads by nulling the pointers on clone.
210 - simplify warn/die hook handling when loading Coro - the convoluted
211 logic seems to be no longer neccessary.
212 - use libecb instead of our own home-grown gcc hacks.
213 - document alternatives to Coro::LWP. Please use them :)
214 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
215 sigaltstack. Really. wine suffers from it, erlang suffers from it,
216 and it's known since at least 2006.
217
2185.372 Wed Feb 23 06:14:30 CET 2011
219 - apparently mingw doesn't provide a working gettimeofday, try to
220 work around that by relying on Time::HiRes (indirectly brought to
221 my attention by Max Maischein).
222 - fix some portability issues when Time::HiRes was used.
223
2245.371 Mon Feb 21 14:36:08 CET 2011
225 - backport to windows process emulation code again.
226
2275.37 Sat Feb 19 07:49:44 CET 2011
228 - add a big "Coro thread life cycle" section to "man Coro".
229 - try a tentative workaround against the breakage that 5.13 has
230 introduced without depreciation period. sigh.
231 - no longer use Time::HiRes if gettimeofday is available, which
232 saves quite a lot of memory.
233
2345.36 Sun Feb 13 05:33:41 CET 2011
235 - automatically load Coro::Channel, Coro::RWLock, Coro::Semaphore,
236 Coro::SemaphoreSet, Coro::Signal and Coro::Specific on first "new"
237 method call.
238 - undocument Coro::Timer::sleep and obsolete whole module.
239 - optimise Coro::Timer::timeout memory and cpu usage.
240 - slightly updated Coro::Intro for recent changes.
241 - do not initialise PL_dirty anymore.
242
2435.25 Thu Nov 11 01:08:39 CET 2010
244 - try a different approach on netbsd - netbsd 5 finally has marginally
245 working pthreads, but still broken ucontext/sigaltstack.
246 - openbsd 4.8 finally got their act together, Coro works out of the box
247 with asm, setjmp and pthreads (no change, just informational).
248
2495.24 Sat Oct 23 11:27:12 CEST 2010
250 - port to the EV 4.0 API.
251 - work around bugs in mingw32, making strawberry perl work
252 out of the box.
253 - correctly modify Coro::AIO function prototypes
254 so that they reflect the "no optional parameters" rule.
255 - "ported" libcoro to C++.
256
2575.23 Mon May 17 18:50:42 CEST 2010
258 - be more resistant to ordering changes when initialising
259 Coro::AnyEvent, Coro::EV and Coro::Event (reported by Matthias
260 Waldorf).
261 - document that perl 5.12 deliberately removed support for cloning.
262
2635.22 Wed Apr 14 03:55:35 CEST 2010
264 - correctly return udnef on errors in Coro::Handle::read/write
265 (testcase by Marc Mims).
10 - convert Coro::Util into a "perl compatibility wrapper" - the functions 266 - convert Coro::Util into a "perl compatibility wrapper" - the functions
11 are less useful now, but are drop-in replacements for existing 267 are less useful now, but are drop-in replacements for existing
12 functions, listing better alternatives in the documentation. This also 268 functions, listing better alternatives in the documentation. This also
13 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton 269 fixes a bug in Coro::LWP which naively substituted Socket::inet_aton
14 with Coro::Util::inet_aton. 270 with Coro::Util::inet_aton.
… …
283 - ->throw is now supported on Coro::State objects. 539 - ->throw is now supported on Coro::State objects.
284 - clean up cctx creation code a bit. 540 - clean up cctx creation code a bit.
285 - entersub is actually a UNOP, not a LOGOP (not a bugfix). 541 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
286 542
2874.9 Sat Nov 8 17:45:27 CET 2008 5434.9 Sat Nov 8 17:45:27 CET 2008
288 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there). 544 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
289 - (libcoro) no longer leak threads in the experimental pthread backend, 545 - (libcoro) no longer leak threads in the experimental pthread backend,
290 also speed it up considerably. 546 also speed it up considerably.
291 - (libcoro) do not rely on makecontext passing void *'s unscathed. 547 - (libcoro) do not rely on makecontext passing void *'s unscathed.
292 - fix compiletime dependencies on libcoro in the Makefile. 548 - fix compiletime dependencies on libcoro in the Makefile.
293 - cctx_count wasn't always updated properly. 549 - cctx_count wasn't always updated properly.
… …
355 - remove debugging code related to MgPV_nolen_const, also try to 611 - remove debugging code related to MgPV_nolen_const, also try to
356 make it compile with perl 5.8.6 (yes, apple apparently loves 612 make it compile with perl 5.8.6 (yes, apple apparently loves
357 outdated software). Reported by John S. 613 outdated software). Reported by John S.
358 614
3594.744 Tue Jul 8 22:06:35 CEST 2008 6154.744 Tue Jul 8 22:06:35 CEST 2008
360 - correctly provide default DIE/WARN handlers as documented. 616 - correctly provide default DIE/WARN handlers as documented.
361 - also overwrite PL_vtbl_sigelem.svt_clear, even though current 617 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
362 implementations inside perl work fine for us. 618 implementations inside perl work fine for us.
363 619
3644.743 Mon Jun 16 00:21:57 CEST 2008 6204.743 Mon Jun 16 00:21:57 CEST 2008
365 - when using Coro::EV without running EV::loop it could 621 - when using Coro::EV without running EV::loop it could
… …
387 - sprinkle "no warnings" freely over everything, also suppress 643 - sprinkle "no warnings" freely over everything, also suppress
388 warnings for some other modules. 644 warnings for some other modules.
389 - fix typo in WSAEWOULDBLOCK. 645 - fix typo in WSAEWOULDBLOCK.
390 646
3914.72 Sun May 25 05:14:36 CEST 2008 6474.72 Sun May 25 05:14:36 CEST 2008
392 - tweak META.yaml a bit, unfortunately, there is no documented way 648 - tweak META.yaml a bit, unfortunately, there is no documented way
393 to have optional dependencies with CPAN. doh :( 649 to have optional dependencies with CPAN. doh :(
394 - avoid running some tests on windows because they would fail due to 650 - avoid running some tests on windows because they would fail due to
395 perl bug (broken fork, broken pipes...). 651 perl bug (broken fork, broken pipes...).
396 - work around perl on windows bugs where perl returns undocumented 652 - work around perl on windows bugs where perl returns undocumented
397 error codes for sysread, syswrite etc. by taking advantage 653 error codes for sysread, syswrite etc. by taking advantage
… …
401 - use unix domain sockets in testsuite to work around 657 - use unix domain sockets in testsuite to work around
402 common perl implementation bugs on widows (they are emulated by 658 common perl implementation bugs on widows (they are emulated by
403 tcp sockets on windows. ugh.) 659 tcp sockets on windows. ugh.)
404 660
4054.71 Sat May 24 20:01:27 CEST 2008 6614.71 Sat May 24 20:01:27 CEST 2008
406 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()"). 662 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
407 - take advantage of async name resolution of AnyEvent::Util. 663 - take advantage of async name resolution of AnyEvent::Util.
408 - work around brutal inet_aton override in Coro::LWP. 664 - work around brutal inet_aton override in Coro::LWP.
409 - take advantage of the readyhook in Coro::EV, for smoother 665 - take advantage of the readyhook in Coro::EV, for smoother
410 scheduling. 666 scheduling.
411 667
4124.7 Sun May 11 00:32:19 CEST 2008 6684.7 Sun May 11 00:32:19 CEST 2008
413 - completely reworked the Coro manpage. 669 - completely reworked the Coro manpage.
414 - added Coro::AnyEvent, generic event loop integration. 670 - added Coro::AnyEvent, generic event loop integration.
415 - implement cancel, ready and kill commands in Coro::Debug. 671 - implement cancel, ready and kill commands in Coro::Debug.
416 - document find_coro in Coro::Debug. 672 - document find_coro in Coro::Debug.
417 - incompatible change: rename has_stack to has_cctx. 673 - incompatible change: rename has_stack to has_cctx.
418 - Coro::AIO and Coro::BDB no longer force event model detection, 674 - Coro::AIO and Coro::BDB no longer force event model detection,
419 use AnyEvent::AIO and AnyEvent::BDB. 675 use AnyEvent::AIO and AnyEvent::BDB.
… …
543 799
5444.11 Thu Oct 11 02:40:24 CEST 2007 8004.11 Thu Oct 11 02:40:24 CEST 2007
545 - port to threaded perls. 801 - port to threaded perls.
546 802
5474.1 Thu Oct 11 02:38:16 CEST 2007 8034.1 Thu Oct 11 02:38:16 CEST 2007
548 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now 804 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
549 be local to each coro (see Coro::State). 805 be local to each coro (see Coro::State).
550 - incompatible change: for very deep reasons, cede and cede_notself 806 - incompatible change: for very deep reasons, cede and cede_notself
551 cannot return anything, so nothing will be returned. 807 cannot return anything, so nothing will be returned.
552 - possibly bring back 5.10 compatibility (untested). 808 - possibly bring back 5.10 compatibility (untested).
553 - work around stupid (and wrong) warning on 5.10 :(. 809 - work around stupid (and wrong) warning on 5.10 :(.
… …
625 but stateful semantics. 881 but stateful semantics.
626 - fixed a lot of typos in Coro.pm (patch submitted by David 882 - fixed a lot of typos in Coro.pm (patch submitted by David
627 Steinbrunner, which applied flawlessly). 883 Steinbrunner, which applied flawlessly).
628 884
6293.6 Sat Apr 14 17:13:31 CEST 2007 8853.6 Sat Apr 14 17:13:31 CEST 2007
630 - added some bugfixes to get eg/myhttpd working again. 886 - added some bugfixes to get eg/myhttpd working again.
631 - added Coro::Storable for often-cede'ing freeze/thaw. 887 - added Coro::Storable for often-cede'ing freeze/thaw.
632 - try to do a clean exit when a coroutine calls exit 888 - try to do a clean exit when a coroutine calls exit
633 (EXPERIMENTAL). 889 (EXPERIMENTAL).
634 - got rid of indirect call through _coro_init. 890 - got rid of indirect call through _coro_init.
635 - updated the partly antique examples in eg/ to 891 - updated the partly antique examples in eg/ to

Diff Legend

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