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

Comparing Coro/Changes (file contents):
Revision 1.565 by root, Fri May 13 19:39:26 2011 UTC vs.
Revision 1.615 by root, Wed May 8 00:44:44 2013 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: channel->maxsize(newsize)? 4TODO: channel->maxsize(newsize)?
5TODO: http://www.microsoft.com/msj/archive/s2ce.aspx 5TODO: __GCC_HAVE_DWARF2_CFI_ASM
6 6
7TODO: tutorial 76.29
8 - when an on_destroy handler destructs the coro currently being
9 destructed a perl scalar could be accessed after being freed,
10 likely causing a crash.
11
126.28 Wed Mar 6 06:58:02 CET 2013
13 - clean remnants of existing __DIE__ and __WARN__ handlers so
14 they lose their magic and will not cause segfaults later
15 (testcase by Andrey Sagulin).
16 - improved Coro::State documentation a bit.
17 - Coro::Debug::command now flushes the output.
18 - add hack detection code for x32 abi, because the braindead slugs
19 who designed that made it look exactly like x86_64 without
20 providing proper compile time symbols to test for it. as a result,
21 this detection cannot work reliably.
22 - valgrind stack registering was broken.
23 - do not rely on Time::HiRes anymore in Coro::Debug.
24
256.23 Fri Dec 7 23:36:37 CET 2012
26 - use experimental fiber implementation on native windows
27 perls.
28 - use sizeof (void *) as multiplication factor for stack sizes,
29 to accomodate the totally braindamaged microsoft 64 bit "os".
30 - changed verifier host from win2k-ap510-32 to win7-sp516-32/64.
31 activeperl 5.16 crashes when PerlIO_define_layer is called due
32 to some bug in the perl dll, strawberry perl at least passes
33 the testsuite.
34 - implement Coro::Handle->peeraddr/host/port, for slightly
35 improved compatibility with LWP.
36 - implement 5.17 compatibility by almost blindly applying a
37 good-looking patch by Father Chrysostomos.
38 - move stack management functions into libcoro 3.
39 - libcoro version 3 "released".
40 - support magic values as timed_io_once args.
41 - recommend AnyEvent 7+ or EV 4+, also require EV
42 version 4 or newer for Coro::EV.
43
446.10 Tue Oct 9 01:14:27 CEST 2012
45 - updated ecb.h, it had a typo that caused it to not compile on many
46 big endian systems (reported by many people).
47 - disable memory fences in ecb.h to improve portability.
48
496.09 Sat Oct 6 23:25:02 CEST 2012
50 - Coro::EV I/O watchers were not interruptible by exceptions
51 (Coro::State::throw) (testcase by sten).
52 - ->throw now puts threads into the ready queue, as this seems to
53 be expected by existing code, and code that doesn't cope with spurious
54 wakeups needs fixing anyway.
55 - use fd -1 in mmap.
56 - cast I32 to int in error message printf.
57 - warn about broken so-called "hardened" kernels.
58
596.08 Fri Apr 13 12:05:47 CEST 2012
60 - be more aggressive about exiting like perl does - formerly,
61 exiting from the non-main thread would not execute END blocks.
62
636.07 Fri Nov 11 21:21:48 CET 2011
64 - work around a bug in PerlIO (setting $SIG{__WARN__} to a PVCV).
65 - update ecb.h.
66
676.06 Mon Aug 8 23:59:48 CEST 2011
68 - cygwin unfortunately patches the stack at runtime, so we use the pthreads
69 backend, which is an order of magnitude slower. unfortunately, cygwins
70 pthread implementation isn't very complete either, so allocate the stack
71 twice just to be sure.
72 (note: cygwin also enables mymalloc, which is NOT THREADSAFE ON WINDOWS,
73 in its ithreaded perl - best recompile cygwin and use the 'w'indows
74 backend for much better performance. also disable ithreads for
75 even better performance...).
76
776.05 Thu Aug 4 21:36:36 CEST 2011
78 - blush, condvar values would not be propagated from send to recv anymore
79 (reported by Chip Salzenberg).
80 - use exponential increase for the readline buffer length in
81 Coro::Handle. also reduce initial allocation to 1020 from 4096 bytes.
82
836.04 Wed Aug 3 17:15:45 CEST 2011
84 - use even more efficient and more compatible condvars for
85 compatibility to AnyEvent 6.x :)
86 - more inconsequential ecb.h updates.
87
886.03 Wed Aug 3 11:41:30 CEST 2011
89 - change how Coro patches AnyEvent condvars for compatibility to
90 AnyEvent 6.x.
91 - update ecb.h, to no longer include <pthread.h> in case WinNT.h
92 hasn't been included.
93
946.02 Wed Jul 13 04:35:19 CEST 2011
95 - "improve portability to Gentoo" - gentoo manages to put perl variables
96 in memory areas that are farther than 2gb apart, which the jit couldn't
97 handle and barfed. now it's just a bit slower on gentoo and similar
98 systems.
99
1006.01 Sun Jul 3 12:31:14 CEST 2011
101 - workarounds are good, but the test for whether pthreads are used
102 was not good. thisone should be better.
103 - check differently whether gcc generates cfi instructions itself.
104
1056.0 Wed Jun 29 19:43:35 CEST 2011
8 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be 106 - INCOMPATIBLE CHANGE: unreferenced coro objects will now be
9 destroyed and cleaned up automatically (e.g. async { schedule }). 107 destroyed and cleaned up automatically (e.g. async { schedule }).
108 - implement a JIT compiler for part of the thread switch code,
109 which gives a 50% speed improvement on threaded perls, and
110 about 4% on non-thraeded perls (so threaded perls now finally
111 reach about half the speed of non-threaded perls).
112 - slightly modernise Coro::Intro, add section about rouse functions.
113 - avoid DEFSV and ERRSV, giving another 10% improvement
114 in thread switching.
10 - Coro::State->is_destroyed is now called is_zombie. 115 - Coro::State->is_destroyed is now called is_zombie.
11 - implement a Coro->safe_cancel method that might fail, but 116 - implement a Coro->safe_cancel method that might fail, but
12 cancels in a "safer" way if it succeeds. 117 cancels in a "safer" way if it succeeds.
118 - add preliminary support for DEBUGGING perls.
119 - get rid of two hash-accesses when initialising a new Coro - this
120 speeds up coro creation by almost a factor of two.
13 - croak when a coro that is being cancelled tries to block 121 - croak when a coro that is being cancelled tries to block
14 (e.g. while executing a guard block), instead of crashing or 122 (e.g. while executing a guard block), instead of crashing or
15 deadlocking. 123 deadlocking.
16 - use a more robust and also faster method to identify Coro::State 124 - use a more robust and also faster method to identify Coro::State
17 objects - speeds up everything a bit. 125 objects - speeds up everything a bit.
… …
27 callbacks (best effort). 135 callbacks (best effort).
28 - move on_destroy into the slf_frame, to allow extension slf 136 - move on_destroy into the slf_frame, to allow extension slf
29 functions to have destructors. 137 functions to have destructors.
30 - get rid if coro refcounting - simply crash in other interpreter 138 - get rid if coro refcounting - simply crash in other interpreter
31 threads by nulling the pointers on clone. 139 threads by nulling the pointers on clone.
140 - simplify warn/die hook handling when loading Coro - the convoluted
141 logic seems to be no longer neccessary.
142 - use libecb instead of our own home-grown gcc hacks.
143 - document alternatives to Coro::LWP. Please use them :)
144 - work around another mindless idiotic NEEDLESS bug in openbsd/mirbsds
145 sigaltstack. Really. wine suffers from it, erlang suffers from it,
146 and it's known since at least 2006.
32 147
335.372 Wed Feb 23 06:14:30 CET 2011 1485.372 Wed Feb 23 06:14:30 CET 2011
34 - apparently mingw doesn't provide a working gettimeofday, try to 149 - apparently mingw doesn't provide a working gettimeofday, try to
35 work around that by relying on Time::HiRes (indirectly brought to 150 work around that by relying on Time::HiRes (indirectly brought to
36 my attention by Max Maischein). 151 my attention by Max Maischein).

Diff Legend

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