| 1 | Revision history for IO::AIO |
1 | Revision history for IO::AIO |
| 2 | |
2 | |
|
|
3 | TODO: better autoconf.pm that can return LIBS etc. |
|
|
4 | TODO: aio_cptree/mvtree/rmtree |
|
|
5 | TODO: aio_utime, aio_chmod, aio_chown? should simply stat/open first? |
|
|
6 | TODO: reduce condvar fairness: schedule hot-cache-threads first? |
|
|
7 | |
|
|
8 | - minor cleanups. |
|
|
9 | - IO::AIO::poll returns request count. |
|
|
10 | |
|
|
11 | 2.2 Tue Oct 31 00:31:00 CET 2006 |
|
|
12 | - minor doc updates. |
|
|
13 | - added aio_readlink. |
|
|
14 | - properly zero-terminate aio_read buffer. |
|
|
15 | - fix aio_read/write with nonzero dataoffset. |
|
|
16 | - reduced size of aio request structure. |
|
|
17 | - better diagnostics when pathnames etc. are not byte-encoded; |
|
|
18 | do not leak request in that case, either. |
|
|
19 | |
|
|
20 | 2.1 Sun Oct 29 02:01:57 CET 2006 |
|
|
21 | - INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
|
|
22 | IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
|
|
23 | - fix a bug in aio_move where it would fail everytime |
|
|
24 | it would have to copy files manually. |
|
|
25 | - doc updates. |
|
|
26 | - start threads only on (very low) demand. |
|
|
27 | - end superfluous threads automatically after 10s idling. |
|
|
28 | - add IO::AIO::max_idle to control this. |
|
|
29 | - some tuning applied to random places. |
|
|
30 | - add aio_mknod. |
|
|
31 | |
|
|
32 | 2.0 Thu Oct 26 18:27:58 CEST 2006 |
|
|
33 | - minor incompatibility: max_outstanding semantics have changed. |
|
|
34 | - major rewrite, consider all functionality EXPERIMENTAL. |
|
|
35 | - aio requests can be cancelled. |
|
|
36 | - callbacks can be (re-)set later. |
|
|
37 | - aio requests can choose between 9 priorities. |
|
|
38 | - add aio_group: aio requests can be grouped |
|
|
39 | into composite requests. |
|
|
40 | - generator interface for groups. |
|
|
41 | - add IO::AIO::poll_some. |
|
|
42 | - major documentation improvements. |
|
|
43 | - very minor bugfixes. |
|
|
44 | - add aio_nop for dummy requests. |
|
|
45 | - add aio_busy mainly for benchmarking and debugging. |
|
|
46 | - use "fast" mutexes on linux. |
|
|
47 | - use dynamic (and larger) buffers instead of |
|
|
48 | putting them onto the stack. |
|
|
49 | - optimise x86 and amd64 a bit. |
|
|
50 | - better error checking when ENOMEM. |
|
|
51 | - hopefully fix mem and dirp leaks on fork. |
|
|
52 | - warn about broken perl malloc. |
|
|
53 | - compiles on cygwin (but pread is broken on cygwin |
|
|
54 | which is not detected by autoconf, so you have to |
|
|
55 | edit autoconf/config.h yourself to enable emulation). |
|
|
56 | |
|
|
57 | 1.8 Sat Jun 24 22:20:44 CEST 2006 |
|
|
58 | - add and/or document aio_rename, aio_link, aio_symlink and aio_move. |
|
|
59 | |
|
|
60 | 1.73 Wed Mar 1 22:49:32 CET 2006 |
|
|
61 | - codename "jost" |
|
|
62 | - AIX/Solaris(?) aren't being helpful again, hardcode a value |
|
|
63 | for NAME_MAX and drive with your eyes closed (reported by |
|
|
64 | wcooley@nakedape.cc). |
|
|
65 | |
|
|
66 | 1.72 Thu Feb 2 00:40:44 CET 2006 |
|
|
67 | - another minor bug in aio_scandir, again, the callback was called |
|
|
68 | twice when the directory can't be stat'ed. |
|
|
69 | |
|
|
70 | 1.71 Mon Dec 26 20:20:03 CET 2005 |
|
|
71 | - grr, another minor bug in aio_scandir, this time resulting |
|
|
72 | in a runtime error. |
|
|
73 | |
|
|
74 | 1.7 Mon Dec 26 19:29:48 CET 2005 |
|
|
75 | - aio_scandir errornously called callback twice in some cases. |
|
|
76 | - added AnyEvent example. |
|
|
77 | |
|
|
78 | 1.61 Wed Sep 7 19:40:42 CEST 2005 |
|
|
79 | - no longer include $Config{libs}, as distros include too much |
|
|
80 | garbage in there that is not installed on typical systems |
|
|
81 | (for example, debian links perl against -lgdbm and many more |
|
|
82 | libraries despite perl not needing it, slowing down startup and |
|
|
83 | wasting memory). |
|
|
84 | |
|
|
85 | 1.6 Tue Aug 30 17:44:44 CEST 2005 |
|
|
86 | - added aio_readdir, aio_scandir. |
|
|
87 | - added aio_sendfile, including sendfile emulation when sendfile |
|
|
88 | isn't available. Linux, FreeBSD and HP-UX might work (sendfilev |
|
|
89 | support is unavailable on my solaris 8 machine, but preliminary |
|
|
90 | code is there. What's lacking is configury stuff and testing...). |
|
|
91 | |
|
|
92 | 1.5 Sat Aug 20 02:32:19 CEST 2005 |
|
|
93 | - create the aio threads on demand only (good after forking). |
|
|
94 | - somebody nagged about C89 compatibility, so this release should |
|
|
95 | be C89 compatible, but no guarantees for future ones. |
|
|
96 | - abort when the configure script fails. |
|
|
97 | |
|
|
98 | 1.4 Wed Aug 17 08:07:27 CEST 2005 |
|
|
99 | - forgot to recreate the pipe after forking, this could cause |
|
|
100 | deadlocks. |
|
|
101 | |
|
|
102 | 1.3 Wed Aug 17 07:25:54 CEST 2005 |
|
|
103 | - properly propagate exceptions from callbacks, instead of silently |
|
|
104 | eating them. |
|
|
105 | - use a different fork algorithm that avoids recursive callback |
|
|
106 | invocations and reduces fork latency in busy processes. |
|
|
107 | - fix a bug where the buffer scalar in aio_read got assigned the wrong |
|
|
108 | length after reading. |
|
|
109 | - fix a bug where the child would process some outstanding requests |
|
|
110 | from the parent. |
|
|
111 | - make a copy of the callback scalar, to avoid some very uncommon |
|
|
112 | (but valid) usages where the callback scalar changes. |
|
|
113 | - the callback was never freed in aio_read/aio_write. |
|
|
114 | - aio_read/aio_write will now set the readonly flag on the data |
|
|
115 | scalar for better error reporting. |
|
|
116 | |
|
|
117 | 1.2 Wed Aug 17 01:22:58 CEST 2005 |
|
|
118 | - verified on cygwin, linux/amd64+ia32+alpha, solaris. |
|
|
119 | - use SvPVbyte for filenames now. You *need* to encode your strings to |
|
|
120 | the external filename encoding first. |
|
|
121 | - fix the bug where filehandles could leak (again). |
|
|
122 | - clarify freebsd situation: make it work and tell me how, or go away |
|
|
123 | with your underdocumented, broken and outdated OS. |
|
|
124 | - fork'ing should now be officially supported (where linux and |
|
|
125 | posix contradict each other, I followed posix). |
|
|
126 | - only call read() once per poll, not twice (speedup). |
|
|
127 | - new functions: aio_rmdir, aio_symlink. |
|
|
128 | - force callback arguments to be undef or a reference, for better |
|
|
129 | error catching. |
|
|
130 | |
|
|
131 | 1.1 Sun Aug 7 05:27:23 CEST 2005 |
| 3 | - added configure script to detetc availability of some calls. |
132 | - added configure script to detect availability of some calls. |
|
|
133 | (imperfect, as we cannot detect things like -lrt due to missing |
|
|
134 | feedback yet). |
|
|
135 | - emulate fdatasync via fsync, pread/pwrite via unsafe read/write |
|
|
136 | and readahead via manual reading on platforms lacking those. |
| 4 | - ported t/03_errors.t+framework by Brad Fitzpatrick |
137 | - ported t/03_errors.t+framework by Brad Fitzpatrick |
| 5 | from Linux::AIO. |
138 | from Linux::AIO. |
| 6 | - removed XOPEN_SOURCE because stupid solaris 9 doesn't |
139 | - removed XOPEN_SOURCE because stupid solaris 9 doesn't |
| 7 | compile with it enabled. Probably breaks others. Oh well. |
140 | compile with it enabled. Probably breaks others. Oh well. |
| 8 | - emulate fdatasync by fsync on platforms lacking it. |
141 | - try to be more careful with errno on non-threaded perls. |
|
|
142 | - cygwins perl insisted on a matching open mode when morphing |
|
|
143 | the fd to an fh. |
|
|
144 | - convert sv's to "byte" encoding in aio_read/aio_write. Might |
|
|
145 | do that for pathnames, too, in the future, so make sure you |
|
|
146 | properly encode your pathnames. |
| 9 | |
147 | |
| 10 | 1.0 Fri Jul 22 10:43:30 CEST 2005 |
148 | 1.0 Fri Jul 22 10:43:30 CEST 2005 |
| 11 | - aio_open did not reliably deliver (open) filehandles. |
149 | - aio_open did not reliably deliver (open) filehandles. |
| 12 | - all aio_*-functions keep a reference to the filehandle so |
150 | - all aio_*-functions keep a reference to the filehandle so |
| 13 | it doesn't go away while the aio requests is working on it. |
151 | it doesn't go away while the aio requests is working on it. |
