1 | Revision history for IO::AIO |
1 | Revision history for IO::AIO |
2 | |
2 | |
3 | TODO: better autoconf.pm that can return LIBS etc. |
3 | TODO: better autoconf.pm that can return LIBS etc. |
|
|
4 | TODO: aio_cptree/mvtree |
|
|
5 | TODO: reduce condvar fairness: schedule hot-cache-threads first? |
|
|
6 | TODO: splice/tee/vmsplice/sync_file_range? (http://kerneltrap.org/node/6505 http://lwn.net/Articles/178199/) |
|
|
7 | TODO: aio_fcntl, at least for file-locking |
4 | |
8 | |
5 | 2.0 |
9 | - fix treescan output duplication and improve output ordering. |
6 | - INCOMPATIBLE CHANGE: max_outstanding was removed, use |
10 | also display files given on the commandline. |
7 | a group with a feeder instead. |
11 | - do our own stack memory management for threads - linux |
|
|
12 | allocates outrageous amounts of VM (not so bad unless you |
|
|
13 | use mlockall...), which severely limits the number of threads |
|
|
14 | on 32-bit arches. |
|
|
15 | |
|
|
16 | 2.6 Sun Mar 30 08:28:11 CEST 2008 |
|
|
17 | - added aio_sync. |
|
|
18 | - added aio_pathsync. |
|
|
19 | - fix prototypes of (void) functions. |
|
|
20 | |
|
|
21 | 2.51 Sat Oct 6 16:04:54 CEST 2007 |
|
|
22 | - perlio isn't generally threadsafe, so aio_close could not work reliably. |
|
|
23 | aio_close now tries a safe hack that might not be that asynchronous |
|
|
24 | (see the manpage for details). |
|
|
25 | - discard callback return values as to not grow the stack endlessly |
|
|
26 | when poll_cb handles many requests without returning. |
|
|
27 | - minor code cleanups. |
|
|
28 | |
|
|
29 | 2.5 Thu Oct 4 14:49:08 CEST 2007 |
|
|
30 | - replaced _fd2fh with faster xs code. |
|
|
31 | - aio_close will now try to do "the right thing" and thus might |
|
|
32 | work sensibly for the very first time. |
|
|
33 | |
|
|
34 | 2.41 Mon Sep 24 21:28:21 CEST 2007 |
|
|
35 | - after fork, invest some work to make sure that the poll_fileno |
|
|
36 | stays stable (by dup2'ing the new fd over the old one), to make |
|
|
37 | it easier for programs/libs that don't control their forking |
|
|
38 | behaviour and cannot recreate their watchers. |
|
|
39 | |
|
|
40 | 2.4 Sun Aug 5 18:44:22 CEST 2007 |
|
|
41 | - add aio_truncate, aio_chmod, aio_chown, aio_utime (linux |
|
|
42 | successfully demonstrated that you can block on futimes...). |
|
|
43 | - allow undef as fileoffset for aio_read/write and use read/write(2) |
|
|
44 | internally (useful for sockets or O_APPEND handles). |
|
|
45 | - allow undef for length in aio_write. |
|
|
46 | - negative dataoffsets work as expected now in aio_read/aio_write. |
|
|
47 | - use NV instead of UV for 32 bit perls and file offsets, as NVs |
|
|
48 | have a larger range then. |
|
|
49 | - shared code between BDB and IO::AIO. |
|
|
50 | - aio_busy was completely broken. now fixed. |
|
|
51 | - readahead emulation now returns something. |
|
|
52 | - correctly set errno to ENOSYS on unimplemented functions |
|
|
53 | (should never happen, though). |
|
|
54 | - large changes to make it partially compile and run on win32, |
|
|
55 | but win32 lacks too much functionality, and perl overrides way |
|
|
56 | too many functions with crashing versions. |
|
|
57 | |
|
|
58 | 2.33 Tue Jan 23 23:55:41 CET 2007 |
|
|
59 | - fix install path (Andreas J. Koenig). |
|
|
60 | |
|
|
61 | 2.32 Mon Jan 22 16:56:23 CET 2007 |
|
|
62 | - added aio_rmtree. |
|
|
63 | - wow, aio_mkdir was missing. |
|
|
64 | - aio_load did return undef on error, not -1. |
|
|
65 | - use prefixexp not prefix in autoconf.pm |
|
|
66 | (suggested by ... rt.cpan.org unreachable). |
|
|
67 | - avoid installing autoconf.pm. |
|
|
68 | |
|
|
69 | 2.31 Sat Jan 6 03:46:02 CET 2007 |
|
|
70 | - added aio_load. |
|
|
71 | |
|
|
72 | 2.3 Sat Dec 23 05:48:07 CET 2006 |
|
|
73 | - fix off-by-one bug in aio_read, causing memory corruption |
|
|
74 | or crashes. |
|
|
75 | |
|
|
76 | 2.21 Fri Dec 22 05:03:38 CET 2006 |
|
|
77 | - minor doc updates. |
|
|
78 | - minor cleanups. |
|
|
79 | - IO::AIO::poll returns request count. |
|
|
80 | - undocumented and unfinished async-signal-reporting. |
|
|
81 | |
|
|
82 | 2.2 Tue Oct 31 00:31:00 CET 2006 |
|
|
83 | - minor doc updates. |
|
|
84 | - added aio_readlink. |
|
|
85 | - properly zero-terminate aio_read buffer. |
|
|
86 | - fix aio_read/write with nonzero dataoffset. |
|
|
87 | - reduced size of aio request structure. |
|
|
88 | - better diagnostics when pathnames etc. are not byte-encoded; |
|
|
89 | do not leak request in that case, either. |
|
|
90 | |
|
|
91 | 2.1 Sun Oct 29 02:01:57 CET 2006 |
|
|
92 | - INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
|
|
93 | IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
|
|
94 | - fix a bug in aio_move where it would fail everytime |
|
|
95 | it would have to copy files manually. |
|
|
96 | - doc updates. |
|
|
97 | - start threads only on (very low) demand. |
|
|
98 | - end superfluous threads automatically after 10s idling. |
|
|
99 | - add IO::AIO::max_idle to control this. |
|
|
100 | - some tuning applied to random places. |
|
|
101 | - add aio_mknod. |
|
|
102 | |
|
|
103 | 2.0 Thu Oct 26 18:27:58 CEST 2006 |
|
|
104 | - minor incompatibility: max_outstanding semantics have changed. |
8 | - major rewrite, consider all functionality EXPERIMENTAL. |
105 | - major rewrite, consider all functionality EXPERIMENTAL. |
9 | - aio requests can be cancelled. |
106 | - aio requests can be cancelled. |
10 | - callbacks can be (re-)set later. |
107 | - callbacks can be (re-)set later. |
11 | - aio requests can have choose 9 priorities. |
108 | - aio requests can choose between 9 priorities. |
12 | - add aio_group: aio requests can be grouped |
109 | - add aio_group: aio requests can be grouped |
13 | into composite requests. |
110 | into composite requests. |
14 | - generator interface for groups. |
111 | - generator interface for groups. |
15 | - add IO::AIO::poll_some. |
112 | - add IO::AIO::poll_some. |
16 | - major documentation improvements. |
113 | - major documentation improvements. |
17 | - very minor bugfixes. |
114 | - very minor bugfixes. |
18 | - add aio_nop for dummy requests. |
115 | - add aio_nop for dummy requests. |
19 | - add aio_busy mainly for benchmarking and debugging. |
116 | - add aio_busy mainly for benchmarking and debugging. |
20 | - deprecated max_outstanding. |
|
|
21 | - use "fast" mutexes on linux. |
117 | - use "fast" mutexes on linux. |
22 | - use dynamic (and larger) buffers instead of |
118 | - use dynamic (and larger) buffers instead of |
23 | putting them onto the stack. |
119 | putting them onto the stack. |
24 | - optimise x86 and amd64 a bit. |
120 | - optimise x86 and amd64 a bit. |
25 | - better error checking when ENOMEM. |
121 | - better error checking when ENOMEM. |
26 | - hopefully fix mem and dirp leaks on fork. |
122 | - hopefully fix mem and dirp leaks on fork. |
|
|
123 | - warn about broken perl malloc. |
27 | - compiles on cygwin (but pread is broken on cygwin |
124 | - compiles on cygwin (but pread is broken on cygwin |
28 | which is not detected by autoconf, so you have to |
125 | which is not detected by autoconf, so you have to |
29 | edit autoconf/config.h yourself to enable emulation). |
126 | edit autoconf/config.h yourself to enable emulation). |
30 | |
127 | |
31 | 1.8 Sat Jun 24 22:20:44 CEST 2006 |
128 | 1.8 Sat Jun 24 22:20:44 CEST 2006 |
… | |
… | |
68 | - somebody nagged about C89 compatibility, so this release should |
165 | - somebody nagged about C89 compatibility, so this release should |
69 | be C89 compatible, but no guarantees for future ones. |
166 | be C89 compatible, but no guarantees for future ones. |
70 | - abort when the configure script fails. |
167 | - abort when the configure script fails. |
71 | |
168 | |
72 | 1.4 Wed Aug 17 08:07:27 CEST 2005 |
169 | 1.4 Wed Aug 17 08:07:27 CEST 2005 |
73 | - forget to recreate the pipe after forking, this could cause |
170 | - forgot to recreate the pipe after forking, this could cause |
74 | deadlocks. |
171 | deadlocks. |
75 | |
172 | |
76 | 1.3 Wed Aug 17 07:25:54 CEST 2005 |
173 | 1.3 Wed Aug 17 07:25:54 CEST 2005 |
77 | - properly propagate exceptions from callbacks, instead of silently |
174 | - properly propagate exceptions from callbacks, instead of silently |
78 | eating them. |
175 | eating them. |