| 1 | Revision history for IO::AIO |
1 | Revision history for IO::AIO |
| 2 | |
2 | |
| 3 | 1.3 |
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 |
|
|
8 | |
|
|
9 | - fix treescan output duplication and improve output ordering. |
|
|
10 | also display files given on the commandline. |
|
|
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: stack size is the larger of PTHREAD_STACK_MIN |
|
|
15 | and 4096 * sizeof (long) (usually this is 16..64k). |
|
|
16 | |
|
|
17 | 2.6 Sun Mar 30 08:28:11 CEST 2008 |
|
|
18 | - added aio_sync. |
|
|
19 | - added aio_pathsync. |
|
|
20 | - fix prototypes of (void) functions. |
|
|
21 | |
|
|
22 | 2.51 Sat Oct 6 16:04:54 CEST 2007 |
|
|
23 | - perlio isn't generally threadsafe, so aio_close could not work reliably. |
|
|
24 | aio_close now tries a safe hack that might not be that asynchronous |
|
|
25 | (see the manpage for details). |
|
|
26 | - discard callback return values as to not grow the stack endlessly |
|
|
27 | when poll_cb handles many requests without returning. |
|
|
28 | - minor code cleanups. |
|
|
29 | |
|
|
30 | 2.5 Thu Oct 4 14:49:08 CEST 2007 |
|
|
31 | - replaced _fd2fh with faster xs code. |
|
|
32 | - aio_close will now try to do "the right thing" and thus might |
|
|
33 | work sensibly for the very first time. |
|
|
34 | |
|
|
35 | 2.41 Mon Sep 24 21:28:21 CEST 2007 |
|
|
36 | - after fork, invest some work to make sure that the poll_fileno |
|
|
37 | stays stable (by dup2'ing the new fd over the old one), to make |
|
|
38 | it easier for programs/libs that don't control their forking |
|
|
39 | behaviour and cannot recreate their watchers. |
|
|
40 | |
|
|
41 | 2.4 Sun Aug 5 18:44:22 CEST 2007 |
|
|
42 | - add aio_truncate, aio_chmod, aio_chown, aio_utime (linux |
|
|
43 | successfully demonstrated that you can block on futimes...). |
|
|
44 | - allow undef as fileoffset for aio_read/write and use read/write(2) |
|
|
45 | internally (useful for sockets or O_APPEND handles). |
|
|
46 | - allow undef for length in aio_write. |
|
|
47 | - negative dataoffsets work as expected now in aio_read/aio_write. |
|
|
48 | - use NV instead of UV for 32 bit perls and file offsets, as NVs |
|
|
49 | have a larger range then. |
|
|
50 | - shared code between BDB and IO::AIO. |
|
|
51 | - aio_busy was completely broken. now fixed. |
|
|
52 | - readahead emulation now returns something. |
|
|
53 | - correctly set errno to ENOSYS on unimplemented functions |
|
|
54 | (should never happen, though). |
|
|
55 | - large changes to make it partially compile and run on win32, |
|
|
56 | but win32 lacks too much functionality, and perl overrides way |
|
|
57 | too many functions with crashing versions. |
|
|
58 | |
|
|
59 | 2.33 Tue Jan 23 23:55:41 CET 2007 |
|
|
60 | - fix install path (Andreas J. Koenig). |
|
|
61 | |
|
|
62 | 2.32 Mon Jan 22 16:56:23 CET 2007 |
|
|
63 | - added aio_rmtree. |
|
|
64 | - wow, aio_mkdir was missing. |
|
|
65 | - aio_load did return undef on error, not -1. |
|
|
66 | - use prefixexp not prefix in autoconf.pm |
|
|
67 | (suggested by ... rt.cpan.org unreachable). |
|
|
68 | - avoid installing autoconf.pm. |
|
|
69 | |
|
|
70 | 2.31 Sat Jan 6 03:46:02 CET 2007 |
|
|
71 | - added aio_load. |
|
|
72 | |
|
|
73 | 2.3 Sat Dec 23 05:48:07 CET 2006 |
|
|
74 | - fix off-by-one bug in aio_read, causing memory corruption |
|
|
75 | or crashes. |
|
|
76 | |
|
|
77 | 2.21 Fri Dec 22 05:03:38 CET 2006 |
|
|
78 | - minor doc updates. |
|
|
79 | - minor cleanups. |
|
|
80 | - IO::AIO::poll returns request count. |
|
|
81 | - undocumented and unfinished async-signal-reporting. |
|
|
82 | |
|
|
83 | 2.2 Tue Oct 31 00:31:00 CET 2006 |
|
|
84 | - minor doc updates. |
|
|
85 | - added aio_readlink. |
|
|
86 | - properly zero-terminate aio_read buffer. |
|
|
87 | - fix aio_read/write with nonzero dataoffset. |
|
|
88 | - reduced size of aio request structure. |
|
|
89 | - better diagnostics when pathnames etc. are not byte-encoded; |
|
|
90 | do not leak request in that case, either. |
|
|
91 | |
|
|
92 | 2.1 Sun Oct 29 02:01:57 CET 2006 |
|
|
93 | - INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
|
|
94 | IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
|
|
95 | - fix a bug in aio_move where it would fail everytime |
|
|
96 | it would have to copy files manually. |
|
|
97 | - doc updates. |
|
|
98 | - start threads only on (very low) demand. |
|
|
99 | - end superfluous threads automatically after 10s idling. |
|
|
100 | - add IO::AIO::max_idle to control this. |
|
|
101 | - some tuning applied to random places. |
|
|
102 | - add aio_mknod. |
|
|
103 | |
|
|
104 | 2.0 Thu Oct 26 18:27:58 CEST 2006 |
|
|
105 | - minor incompatibility: max_outstanding semantics have changed. |
|
|
106 | - major rewrite, consider all functionality EXPERIMENTAL. |
|
|
107 | - aio requests can be cancelled. |
|
|
108 | - callbacks can be (re-)set later. |
|
|
109 | - aio requests can choose between 9 priorities. |
|
|
110 | - add aio_group: aio requests can be grouped |
|
|
111 | into composite requests. |
|
|
112 | - generator interface for groups. |
|
|
113 | - add IO::AIO::poll_some. |
|
|
114 | - major documentation improvements. |
|
|
115 | - very minor bugfixes. |
|
|
116 | - add aio_nop for dummy requests. |
|
|
117 | - add aio_busy mainly for benchmarking and debugging. |
|
|
118 | - use "fast" mutexes on linux. |
|
|
119 | - use dynamic (and larger) buffers instead of |
|
|
120 | putting them onto the stack. |
|
|
121 | - optimise x86 and amd64 a bit. |
|
|
122 | - better error checking when ENOMEM. |
|
|
123 | - hopefully fix mem and dirp leaks on fork. |
|
|
124 | - warn about broken perl malloc. |
|
|
125 | - compiles on cygwin (but pread is broken on cygwin |
|
|
126 | which is not detected by autoconf, so you have to |
|
|
127 | edit autoconf/config.h yourself to enable emulation). |
|
|
128 | |
|
|
129 | 1.8 Sat Jun 24 22:20:44 CEST 2006 |
|
|
130 | - add and/or document aio_rename, aio_link, aio_symlink and aio_move. |
|
|
131 | |
|
|
132 | 1.73 Wed Mar 1 22:49:32 CET 2006 |
|
|
133 | - codename "jost" |
|
|
134 | - AIX/Solaris(?) aren't being helpful again, hardcode a value |
|
|
135 | for NAME_MAX and drive with your eyes closed (reported by |
|
|
136 | wcooley@nakedape.cc). |
|
|
137 | |
|
|
138 | 1.72 Thu Feb 2 00:40:44 CET 2006 |
|
|
139 | - another minor bug in aio_scandir, again, the callback was called |
|
|
140 | twice when the directory can't be stat'ed. |
|
|
141 | |
|
|
142 | 1.71 Mon Dec 26 20:20:03 CET 2005 |
|
|
143 | - grr, another minor bug in aio_scandir, this time resulting |
|
|
144 | in a runtime error. |
|
|
145 | |
|
|
146 | 1.7 Mon Dec 26 19:29:48 CET 2005 |
|
|
147 | - aio_scandir errornously called callback twice in some cases. |
|
|
148 | - added AnyEvent example. |
|
|
149 | |
|
|
150 | 1.61 Wed Sep 7 19:40:42 CEST 2005 |
|
|
151 | - no longer include $Config{libs}, as distros include too much |
|
|
152 | garbage in there that is not installed on typical systems |
|
|
153 | (for example, debian links perl against -lgdbm and many more |
|
|
154 | libraries despite perl not needing it, slowing down startup and |
|
|
155 | wasting memory). |
|
|
156 | |
|
|
157 | 1.6 Tue Aug 30 17:44:44 CEST 2005 |
|
|
158 | - added aio_readdir, aio_scandir. |
|
|
159 | - added aio_sendfile, including sendfile emulation when sendfile |
|
|
160 | isn't available. Linux, FreeBSD and HP-UX might work (sendfilev |
|
|
161 | support is unavailable on my solaris 8 machine, but preliminary |
|
|
162 | code is there. What's lacking is configury stuff and testing...). |
|
|
163 | |
|
|
164 | 1.5 Sat Aug 20 02:32:19 CEST 2005 |
|
|
165 | - create the aio threads on demand only (good after forking). |
|
|
166 | - somebody nagged about C89 compatibility, so this release should |
|
|
167 | be C89 compatible, but no guarantees for future ones. |
|
|
168 | - abort when the configure script fails. |
|
|
169 | |
|
|
170 | 1.4 Wed Aug 17 08:07:27 CEST 2005 |
|
|
171 | - forgot to recreate the pipe after forking, this could cause |
|
|
172 | deadlocks. |
|
|
173 | |
|
|
174 | 1.3 Wed Aug 17 07:25:54 CEST 2005 |
| 4 | - properly propagate exceptions from callbacks, instead of silently |
175 | - properly propagate exceptions from callbacks, instead of silently |
| 5 | eating them. |
176 | eating them. |
| 6 | - forking while in an atfork handler is now supported, although |
177 | - use a different fork algorithm that avoids recursive callback |
| 7 | I am pretty sure some pthread implementations will barf - don't |
178 | invocations and reduces fork latency in busy processes. |
| 8 | fork in an AIO callback if that is a concern for you, or synchronize |
179 | - fix a bug where the buffer scalar in aio_read got assigned the wrong |
| 9 | for yourself by clearing the request queue first. |
180 | length after reading. |
|
|
181 | - fix a bug where the child would process some outstanding requests |
|
|
182 | from the parent. |
|
|
183 | - make a copy of the callback scalar, to avoid some very uncommon |
|
|
184 | (but valid) usages where the callback scalar changes. |
|
|
185 | - the callback was never freed in aio_read/aio_write. |
|
|
186 | - aio_read/aio_write will now set the readonly flag on the data |
|
|
187 | scalar for better error reporting. |
| 10 | |
188 | |
| 11 | 1.2 Wed Aug 17 01:22:58 CEST 2005 |
189 | 1.2 Wed Aug 17 01:22:58 CEST 2005 |
| 12 | - verified on cygwin, linux/amd64+ia32+alpha, solaris. |
190 | - verified on cygwin, linux/amd64+ia32+alpha, solaris. |
| 13 | - use SvPVbyte for filenames now. You *need* to encode your strings to |
191 | - use SvPVbyte for filenames now. You *need* to encode your strings to |
| 14 | the external filename encoding first. |
192 | the external filename encoding first. |
