| … | |
… | |
| 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? (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 | |
| 9 | 3.2 |
11 | 3.5 Thu Jan 7 21:25:04 CET 2010 |
| 10 | TODO: aio_scandir should short-circuit the case of no DT_UNKNOWN. |
12 | - (liebio) fix freebsd sendfile (Vanilla Hsu). |
|
|
13 | - (libeio) also fix darwin sendfile that suffered from a similar bug. |
|
|
14 | - add aio_statvfs as interface to statvfs/fstatvfs. |
|
|
15 | - work around buggy symlink() on cygwin in the testsuite. |
|
|
16 | - wtf. freebsd-8.0 actually passes the testsuite! I knew |
|
|
17 | they could implement fork and semaphores one day! cheers! |
|
|
18 | |
|
|
19 | 3.4 Sat Jan 2 15:13:04 CET 2010 |
|
|
20 | - (libeio) max_poll_time was not properly converted to ticks. |
|
|
21 | - clarify cancel_subs description. |
|
|
22 | - IO::AIO::sendfile did not actually return the return value |
|
|
23 | from the sendfile call. |
|
|
24 | - implement aio_msync, aio_mtouch. |
|
|
25 | - (libeio) tentatively support darwin in sendfile. |
|
|
26 | |
|
|
27 | 3.31 Thu Nov 12 02:14:29 CET 2009 |
|
|
28 | - fix result status documentation of aio_copy, aio_move. |
|
|
29 | - speed up object creation considerably by avoiding |
|
|
30 | hash lookups. |
|
|
31 | |
|
|
32 | 3.3 Wed Aug 5 13:52:58 CEST 2009 |
|
|
33 | - use common::sense. |
|
|
34 | - use common schmorp.h header. |
|
|
35 | - allow integers (file descriptors) in addition to file handles. |
|
|
36 | - take advantage of linux' eventfd (but it seems slower on smp?). |
|
|
37 | - use poll not select on posix platforms. |
|
|
38 | |
|
|
39 | 3.261 Wed Jul 1 10:11:51 CEST 2009 |
|
|
40 | - more 0S X workarounds (patch by Tokuhiro Matsuno). |
|
|
41 | |
|
|
42 | 3.26 Tue Jun 30 09:33:26 CEST 2009 |
|
|
43 | - 0S X of course claims to be posix 2008 but lacks posix_fadvise. |
|
|
44 | try to work around this horribly broken OS in a somewhat hackish |
|
|
45 | way. might help other os's too. |
|
|
46 | |
|
|
47 | 3.25 Sat Jun 27 05:18:26 CEST 2009 |
|
|
48 | - added IO::AIO::fadvise and IO::AIO::sendfile. |
|
|
49 | - (libeio) replaced quicksort+insertion sort by a tuned radix |
|
|
50 | sort + insertion sort, resulting in comparable runtime (usually |
|
|
51 | faster) to the old version, but without any ill side effects on |
|
|
52 | degenerated (for quicksort) data. |
|
|
53 | - (libeio) correctly sort dirs by inodes if we have perfect type |
|
|
54 | knowledge and DIRS_FIRST is used. |
|
|
55 | - (libeio) internally the wrong DT_* constants were used, but no |
|
|
56 | known system has differing EIO_DT_*/DT_* values, so not an issue. |
|
|
57 | - removed a (harmless) assert that was left over on the code |
|
|
58 | but should not have been. |
|
|
59 | - use more correct types (change IO length from IV to VAL64, |
|
|
60 | which makes a difference on systems without 64 bit perls). |
|
|
61 | |
|
|
62 | 3.23 Sat Jun 13 16:57:58 CEST 2009 |
|
|
63 | - fix off-by-one bug in aio_readdir that was introduced in the |
|
|
64 | rushed 3.22 release. |
|
|
65 | |
|
|
66 | 3.22 Sat Jun 13 15:32:40 CEST 2009 |
|
|
67 | - speed up readdirx sort algorithm slightly. |
|
|
68 | - bin/treescan was missing from distro tarball. |
|
|
69 | |
|
|
70 | 3.21 Fri Jun 12 18:45:53 CEST 2009 |
|
|
71 | - new options --dirs and --files for treescan. |
|
|
72 | - install bin/treescan by default. |
|
|
73 | - (libeio) aio_readdir can now be cancelled while executing. |
|
|
74 | - fix a printf format string for 64 bit systems (could lead |
|
|
75 | to problems on big endian 64 bit systems). |
|
|
76 | - do not use qsort() but our own algorithm: glibc initialises |
|
|
77 | the whole locale and I/O subsystem inside qsort, causing |
|
|
78 | a stack overflow on 32 bit machines. The new sort uses much less |
|
|
79 | stack and is more than twice as fast in typical situations. |
|
|
80 | |
|
|
81 | 3.2 Sun Jun 7 20:30:05 CEST 2009 |
| 11 | - (libeio) pwrite emulation was even more flawed than intended and did |
82 | - (libeio) pwrite emulation was even more flawed than intended and did |
| 12 | not restore the file offset. |
83 | not restore the file offset. |
| 13 | - add aio_readdirx, which can return inode and filetype and sort |
84 | - add aio_readdirx, which can return inode and filetype and sort |
| 14 | the names in various ways. |
85 | the names in various ways. |
| 15 | - unfortunately, utime, chmod, chown on an open file that has just |
86 | - unfortunately, utime, chmod, chown on an open file that has just |
