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 |
4 | TODO: aio_cptree/mvtree |
5 | TODO: reduce condvar fairness: schedule hot-cache-threads first? |
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/) |
6 | TODO: splice/tee/vmsplice? (http://kerneltrap.org/node/6505 http://lwn.net/Articles/178199/) |
7 | TODO: aio_fcntl, at least for file-locking |
7 | TODO: aio_fcntl, at least for file-locking |
|
|
8 | TODO: fallocate, if it wouldn't be so useless. |
|
|
9 | TODO: IO::AIO::mmap? aio_mincore? |
8 | |
10 | |
|
|
11 | - (libeio) more fixes for the freebsd/apple sendfile - broken by |
|
|
12 | manpage standards, but apparently correct accoridng to actual |
|
|
13 | kernel sources. |
|
|
14 | - add IO::AIO::mmap/munmap support. |
|
|
15 | - add IO::AIO::m(un)lockall support. |
|
|
16 | - clean up manpage. |
|
|
17 | |
|
|
18 | 3.5 Thu Jan 7 21:25:04 CET 2010 |
|
|
19 | - (liebio) fix freebsd sendfile (Vanilla Hsu). |
|
|
20 | - (libeio) also fix darwin sendfile that suffered from a similar bug. |
|
|
21 | - add aio_statvfs as interface to statvfs/fstatvfs. |
|
|
22 | - work around buggy symlink() on cygwin in the testsuite. |
|
|
23 | - wtf. freebsd-8.0 actually passes the testsuite! I knew |
|
|
24 | they could implement fork and semaphores one day! cheers! |
|
|
25 | |
|
|
26 | 3.4 Sat Jan 2 15:13:04 CET 2010 |
|
|
27 | - (libeio) max_poll_time was not properly converted to ticks. |
|
|
28 | - clarify cancel_subs description. |
|
|
29 | - IO::AIO::sendfile did not actually return the return value |
|
|
30 | from the sendfile call. |
|
|
31 | - implement aio_msync, aio_mtouch. |
|
|
32 | - (libeio) tentatively support darwin in sendfile. |
|
|
33 | |
|
|
34 | 3.31 Thu Nov 12 02:14:29 CET 2009 |
|
|
35 | - fix result status documentation of aio_copy, aio_move. |
|
|
36 | - speed up object creation considerably by avoiding |
|
|
37 | hash lookups. |
|
|
38 | |
|
|
39 | 3.3 Wed Aug 5 13:52:58 CEST 2009 |
|
|
40 | - use common::sense. |
|
|
41 | - use common schmorp.h header. |
|
|
42 | - allow integers (file descriptors) in addition to file handles. |
|
|
43 | - take advantage of linux' eventfd (but it seems slower on smp?). |
|
|
44 | - use poll not select on posix platforms. |
|
|
45 | |
|
|
46 | 3.261 Wed Jul 1 10:11:51 CEST 2009 |
|
|
47 | - more 0S X workarounds (patch by Tokuhiro Matsuno). |
|
|
48 | |
|
|
49 | 3.26 Tue Jun 30 09:33:26 CEST 2009 |
|
|
50 | - 0S X of course claims to be posix 2008 but lacks posix_fadvise. |
|
|
51 | try to work around this horribly broken OS in a somewhat hackish |
|
|
52 | way. might help other os's too. |
|
|
53 | |
|
|
54 | 3.25 Sat Jun 27 05:18:26 CEST 2009 |
|
|
55 | - added IO::AIO::fadvise and IO::AIO::sendfile. |
|
|
56 | - (libeio) replaced quicksort+insertion sort by a tuned radix |
|
|
57 | sort + insertion sort, resulting in comparable runtime (usually |
|
|
58 | faster) to the old version, but without any ill side effects on |
|
|
59 | degenerated (for quicksort) data. |
|
|
60 | - (libeio) correctly sort dirs by inodes if we have perfect type |
|
|
61 | knowledge and DIRS_FIRST is used. |
|
|
62 | - (libeio) internally the wrong DT_* constants were used, but no |
|
|
63 | known system has differing EIO_DT_*/DT_* values, so not an issue. |
|
|
64 | - removed a (harmless) assert that was left over on the code |
|
|
65 | but should not have been. |
|
|
66 | - use more correct types (change IO length from IV to VAL64, |
|
|
67 | which makes a difference on systems without 64 bit perls). |
|
|
68 | |
|
|
69 | 3.23 Sat Jun 13 16:57:58 CEST 2009 |
|
|
70 | - fix off-by-one bug in aio_readdir that was introduced in the |
|
|
71 | rushed 3.22 release. |
|
|
72 | |
|
|
73 | 3.22 Sat Jun 13 15:32:40 CEST 2009 |
|
|
74 | - speed up readdirx sort algorithm slightly. |
|
|
75 | - bin/treescan was missing from distro tarball. |
|
|
76 | |
|
|
77 | 3.21 Fri Jun 12 18:45:53 CEST 2009 |
|
|
78 | - new options --dirs and --files for treescan. |
|
|
79 | - install bin/treescan by default. |
|
|
80 | - (libeio) aio_readdir can now be cancelled while executing. |
|
|
81 | - fix a printf format string for 64 bit systems (could lead |
|
|
82 | to problems on big endian 64 bit systems). |
|
|
83 | - do not use qsort() but our own algorithm: glibc initialises |
|
|
84 | the whole locale and I/O subsystem inside qsort, causing |
|
|
85 | a stack overflow on 32 bit machines. The new sort uses much less |
|
|
86 | stack and is more than twice as fast in typical situations. |
|
|
87 | |
|
|
88 | 3.2 Sun Jun 7 20:30:05 CEST 2009 |
|
|
89 | - (libeio) pwrite emulation was even more flawed than intended and did |
|
|
90 | not restore the file offset. |
|
|
91 | - add aio_readdirx, which can return inode and filetype and sort |
|
|
92 | the names in various ways. |
|
|
93 | - unfortunately, utime, chmod, chown on an open file that has just |
|
|
94 | been written can easily block, which caused aio_copy to block the |
|
|
95 | process. no more! |
|
|
96 | - no longer rely on dst path in aio_copy when futime is available. |
|
|
97 | |
|
|
98 | 3.19 Tue Apr 21 22:05:21 CEST 2009 |
|
|
99 | - more perl 5.10 workarounds for aio_read and write. |
|
|
100 | - aio_write no longer modifies the sv (if possible). |
|
|
101 | - aio_read now works correctly with magic values. |
|
|
102 | |
|
|
103 | 3.18 Sun Apr 19 21:17:32 CEST 2009 |
|
|
104 | - better diagnostics when some aio_* functions get passed an illegal fh. |
|
|
105 | - try to avoid crashes due to incompatible 5.10 API changes. grmbl. |
|
|
106 | |
|
|
107 | 3.17 Thu Nov 20 08:45:36 CET 2008 |
|
|
108 | - (libeio) added aio_sync_file_range (untested). |
|
|
109 | - add aio_busy to @IO::AIO::AIO_REQ. |
|
|
110 | |
|
|
111 | 3.16 Wed Oct 22 18:28:01 CEST 2008 |
|
|
112 | - use SvREFCNT_inc instead of SvREFCNT_inc_NN in a non-speed critical |
|
|
113 | part to improve portability to perl 5.8 (reported by szymon). |
|
|
114 | |
|
|
115 | 3.15 Mon Oct 13 00:39:55 CEST 2008 |
|
|
116 | - automatic removal of feeders was broken. |
|
|
117 | - (libeio) use a more robust method to detect whether a feeder |
|
|
118 | has added something to the group or not. |
|
|
119 | |
|
|
120 | 3.1 Thu Oct 2 13:34:40 CEST 2008 |
|
|
121 | - pre-deref the passed callback object, for increased speed |
|
|
122 | and decreased memory usage. |
|
|
123 | - call on_next_submit callback even for aio_group - impact unknown, |
|
|
124 | but seems more correct. |
|
|
125 | - $req->cb now returns the original callback. |
|
|
126 | - (libeio) pass EIO_GROUP/aio_group requests through the queue without |
|
|
127 | requiring a thread switch. |
|
|
128 | - (libeio) status code of aio_futime and aio_utime was always passed 0. |
|
|
129 | - do some other ยต-optimisations. |
|
|
130 | |
|
|
131 | 3.07 Sat Aug 2 16:06:13 CEST 2008 |
|
|
132 | - do not include LIBS in autoconf tests. |
|
|
133 | |
|
|
134 | 3.06 Tue Jul 15 12:41:32 CEST 2008 |
9 | - move preadwritelock definition before it's first reference, |
135 | - move preadwritelock definition before it's first reference, |
10 | to make it compile again on broken platforms (they still exist). |
136 | to make it compile again on broken platforms (they still exist) |
11 | reported by Rob Bloodgood. |
137 | or badly cofngiured perls (redhat...), reported by Rob Bloodgood. |
12 | |
138 | |
13 | 3.05 Thu Jun 19 23:23:52 CEST 2008 |
139 | 3.05 Thu Jun 19 23:23:52 CEST 2008 |
14 | - work around perl overriding readdir etc. with thread-unsafe |
140 | - work around perl overriding readdir etc. with thread-unsafe |
15 | (sic!) versions. |
141 | (sic!) versions. |
16 | |
142 | |