1 |
Revision history for IO::AIO |
2 |
|
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 |
2.62 Sat Apr 26 13:59:33 CEST 2008 |
10 |
- port to solaris perls configured for old posix |
11 |
(analysed by Jost Krieger). |
12 |
- keep a reference to the perl filehandle object in aio_close, |
13 |
so it doesn't get closed prematurely, leading to ugly races. |
14 |
|
15 |
2.61 Wed Apr 16 18:45:02 CEST 2008 |
16 |
- fix treescan output duplication and improve output ordering. |
17 |
also display files given on the commandline. |
18 |
- use a different algorithm for aio_close that is faster |
19 |
and probably has even lower chances of blocking. |
20 |
- do our own stack memory management for threads - linux |
21 |
allocates outrageous amounts of VM (not so bad unless you |
22 |
use mlockall...), which severely limits the number of threads |
23 |
on 32-bit arches: stack size is the larger of PTHREAD_STACK_MIN |
24 |
and 4096 * sizeof (long) (usually this is 16..64k). |
25 |
|
26 |
2.6 Sun Mar 30 08:28:11 CEST 2008 |
27 |
- added aio_sync. |
28 |
- added aio_pathsync. |
29 |
- fix prototypes of (void) functions. |
30 |
|
31 |
2.51 Sat Oct 6 16:04:54 CEST 2007 |
32 |
- perlio isn't generally threadsafe, so aio_close could not work reliably. |
33 |
aio_close now tries a safe hack that might not be that asynchronous |
34 |
(see the manpage for details). |
35 |
- discard callback return values as to not grow the stack endlessly |
36 |
when poll_cb handles many requests without returning. |
37 |
- minor code cleanups. |
38 |
|
39 |
2.5 Thu Oct 4 14:49:08 CEST 2007 |
40 |
- replaced _fd2fh with faster xs code. |
41 |
- aio_close will now try to do "the right thing" and thus might |
42 |
work sensibly for the very first time. |
43 |
|
44 |
2.41 Mon Sep 24 21:28:21 CEST 2007 |
45 |
- after fork, invest some work to make sure that the poll_fileno |
46 |
stays stable (by dup2'ing the new fd over the old one), to make |
47 |
it easier for programs/libs that don't control their forking |
48 |
behaviour and cannot recreate their watchers. |
49 |
|
50 |
2.4 Sun Aug 5 18:44:22 CEST 2007 |
51 |
- add aio_truncate, aio_chmod, aio_chown, aio_utime (linux |
52 |
successfully demonstrated that you can block on futimes...). |
53 |
- allow undef as fileoffset for aio_read/write and use read/write(2) |
54 |
internally (useful for sockets or O_APPEND handles). |
55 |
- allow undef for length in aio_write. |
56 |
- negative dataoffsets work as expected now in aio_read/aio_write. |
57 |
- use NV instead of UV for 32 bit perls and file offsets, as NVs |
58 |
have a larger range then. |
59 |
- shared code between BDB and IO::AIO. |
60 |
- aio_busy was completely broken. now fixed. |
61 |
- readahead emulation now returns something. |
62 |
- correctly set errno to ENOSYS on unimplemented functions |
63 |
(should never happen, though). |
64 |
- large changes to make it partially compile and run on win32, |
65 |
but win32 lacks too much functionality, and perl overrides way |
66 |
too many functions with crashing versions. |
67 |
|
68 |
2.33 Tue Jan 23 23:55:41 CET 2007 |
69 |
- fix install path (Andreas J. Koenig). |
70 |
|
71 |
2.32 Mon Jan 22 16:56:23 CET 2007 |
72 |
- added aio_rmtree. |
73 |
- wow, aio_mkdir was missing. |
74 |
- aio_load did return undef on error, not -1. |
75 |
- use prefixexp not prefix in autoconf.pm |
76 |
(suggested by ... rt.cpan.org unreachable). |
77 |
- avoid installing autoconf.pm. |
78 |
|
79 |
2.31 Sat Jan 6 03:46:02 CET 2007 |
80 |
- added aio_load. |
81 |
|
82 |
2.3 Sat Dec 23 05:48:07 CET 2006 |
83 |
- fix off-by-one bug in aio_read, causing memory corruption |
84 |
or crashes. |
85 |
|
86 |
2.21 Fri Dec 22 05:03:38 CET 2006 |
87 |
- minor doc updates. |
88 |
- minor cleanups. |
89 |
- IO::AIO::poll returns request count. |
90 |
- undocumented and unfinished async-signal-reporting. |
91 |
|
92 |
2.2 Tue Oct 31 00:31:00 CET 2006 |
93 |
- minor doc updates. |
94 |
- added aio_readlink. |
95 |
- properly zero-terminate aio_read buffer. |
96 |
- fix aio_read/write with nonzero dataoffset. |
97 |
- reduced size of aio request structure. |
98 |
- better diagnostics when pathnames etc. are not byte-encoded; |
99 |
do not leak request in that case, either. |
100 |
|
101 |
2.1 Sun Oct 29 02:01:57 CET 2006 |
102 |
- INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
103 |
IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
104 |
- fix a bug in aio_move where it would fail everytime |
105 |
it would have to copy files manually. |
106 |
- doc updates. |
107 |
- start threads only on (very low) demand. |
108 |
- end superfluous threads automatically after 10s idling. |
109 |
- add IO::AIO::max_idle to control this. |
110 |
- some tuning applied to random places. |
111 |
- add aio_mknod. |
112 |
|
113 |
2.0 Thu Oct 26 18:27:58 CEST 2006 |
114 |
- minor incompatibility: max_outstanding semantics have changed. |
115 |
- major rewrite, consider all functionality EXPERIMENTAL. |
116 |
- aio requests can be cancelled. |
117 |
- callbacks can be (re-)set later. |
118 |
- aio requests can choose between 9 priorities. |
119 |
- add aio_group: aio requests can be grouped |
120 |
into composite requests. |
121 |
- generator interface for groups. |
122 |
- add IO::AIO::poll_some. |
123 |
- major documentation improvements. |
124 |
- very minor bugfixes. |
125 |
- add aio_nop for dummy requests. |
126 |
- add aio_busy mainly for benchmarking and debugging. |
127 |
- use "fast" mutexes on linux. |
128 |
- use dynamic (and larger) buffers instead of |
129 |
putting them onto the stack. |
130 |
- optimise x86 and amd64 a bit. |
131 |
- better error checking when ENOMEM. |
132 |
- hopefully fix mem and dirp leaks on fork. |
133 |
- warn about broken perl malloc. |
134 |
- compiles on cygwin (but pread is broken on cygwin |
135 |
which is not detected by autoconf, so you have to |
136 |
edit autoconf/config.h yourself to enable emulation). |
137 |
|
138 |
1.8 Sat Jun 24 22:20:44 CEST 2006 |
139 |
- add and/or document aio_rename, aio_link, aio_symlink and aio_move. |
140 |
|
141 |
1.73 Wed Mar 1 22:49:32 CET 2006 |
142 |
- codename "jost" |
143 |
- AIX/Solaris(?) aren't being helpful again, hardcode a value |
144 |
for NAME_MAX and drive with your eyes closed (reported by |
145 |
wcooley@nakedape.cc). |
146 |
|
147 |
1.72 Thu Feb 2 00:40:44 CET 2006 |
148 |
- another minor bug in aio_scandir, again, the callback was called |
149 |
twice when the directory can't be stat'ed. |
150 |
|
151 |
1.71 Mon Dec 26 20:20:03 CET 2005 |
152 |
- grr, another minor bug in aio_scandir, this time resulting |
153 |
in a runtime error. |
154 |
|
155 |
1.7 Mon Dec 26 19:29:48 CET 2005 |
156 |
- aio_scandir errornously called callback twice in some cases. |
157 |
- added AnyEvent example. |
158 |
|
159 |
1.61 Wed Sep 7 19:40:42 CEST 2005 |
160 |
- no longer include $Config{libs}, as distros include too much |
161 |
garbage in there that is not installed on typical systems |
162 |
(for example, debian links perl against -lgdbm and many more |
163 |
libraries despite perl not needing it, slowing down startup and |
164 |
wasting memory). |
165 |
|
166 |
1.6 Tue Aug 30 17:44:44 CEST 2005 |
167 |
- added aio_readdir, aio_scandir. |
168 |
- added aio_sendfile, including sendfile emulation when sendfile |
169 |
isn't available. Linux, FreeBSD and HP-UX might work (sendfilev |
170 |
support is unavailable on my solaris 8 machine, but preliminary |
171 |
code is there. What's lacking is configury stuff and testing...). |
172 |
|
173 |
1.5 Sat Aug 20 02:32:19 CEST 2005 |
174 |
- create the aio threads on demand only (good after forking). |
175 |
- somebody nagged about C89 compatibility, so this release should |
176 |
be C89 compatible, but no guarantees for future ones. |
177 |
- abort when the configure script fails. |
178 |
|
179 |
1.4 Wed Aug 17 08:07:27 CEST 2005 |
180 |
- forgot to recreate the pipe after forking, this could cause |
181 |
deadlocks. |
182 |
|
183 |
1.3 Wed Aug 17 07:25:54 CEST 2005 |
184 |
- properly propagate exceptions from callbacks, instead of silently |
185 |
eating them. |
186 |
- use a different fork algorithm that avoids recursive callback |
187 |
invocations and reduces fork latency in busy processes. |
188 |
- fix a bug where the buffer scalar in aio_read got assigned the wrong |
189 |
length after reading. |
190 |
- fix a bug where the child would process some outstanding requests |
191 |
from the parent. |
192 |
- make a copy of the callback scalar, to avoid some very uncommon |
193 |
(but valid) usages where the callback scalar changes. |
194 |
- the callback was never freed in aio_read/aio_write. |
195 |
- aio_read/aio_write will now set the readonly flag on the data |
196 |
scalar for better error reporting. |
197 |
|
198 |
1.2 Wed Aug 17 01:22:58 CEST 2005 |
199 |
- verified on cygwin, linux/amd64+ia32+alpha, solaris. |
200 |
- use SvPVbyte for filenames now. You *need* to encode your strings to |
201 |
the external filename encoding first. |
202 |
- fix the bug where filehandles could leak (again). |
203 |
- clarify freebsd situation: make it work and tell me how, or go away |
204 |
with your underdocumented, broken and outdated OS. |
205 |
- fork'ing should now be officially supported (where linux and |
206 |
posix contradict each other, I followed posix). |
207 |
- only call read() once per poll, not twice (speedup). |
208 |
- new functions: aio_rmdir, aio_symlink. |
209 |
- force callback arguments to be undef or a reference, for better |
210 |
error catching. |
211 |
|
212 |
1.1 Sun Aug 7 05:27:23 CEST 2005 |
213 |
- added configure script to detect availability of some calls. |
214 |
(imperfect, as we cannot detect things like -lrt due to missing |
215 |
feedback yet). |
216 |
- emulate fdatasync via fsync, pread/pwrite via unsafe read/write |
217 |
and readahead via manual reading on platforms lacking those. |
218 |
- ported t/03_errors.t+framework by Brad Fitzpatrick |
219 |
from Linux::AIO. |
220 |
- removed XOPEN_SOURCE because stupid solaris 9 doesn't |
221 |
compile with it enabled. Probably breaks others. Oh well. |
222 |
- try to be more careful with errno on non-threaded perls. |
223 |
- cygwins perl insisted on a matching open mode when morphing |
224 |
the fd to an fh. |
225 |
- convert sv's to "byte" encoding in aio_read/aio_write. Might |
226 |
do that for pathnames, too, in the future, so make sure you |
227 |
properly encode your pathnames. |
228 |
|
229 |
1.0 Fri Jul 22 10:43:30 CEST 2005 |
230 |
- aio_open did not reliably deliver (open) filehandles. |
231 |
- all aio_*-functions keep a reference to the filehandle so |
232 |
it doesn't go away while the aio requests is working on it. |
233 |
- disable warnings in AIO.pm. |
234 |
|
235 |
0.9 Wed Jul 20 23:55:04 CEST 2005 |
236 |
- allow recursive calls to poll_cb, use less locking calls, too. |
237 |
- fix the Gtk2-example in the SYNOPSIS. |
238 |
- use pread/pwrite not pread64/pwrite64. this was simply a leftover |
239 |
from Linux::AIO. |
240 |
(found by compiling on cygwin. Unfortunately, cygwin declares but |
241 |
does not implement pread/pwrite). |
242 |
- only include <sys/syscall.h> on linux. |
243 |
|
244 |
0.5 Tue Jul 12 13:03:09 CEST 2005 |
245 |
- removed #include <endian.h>, it's not portable, and we don't |
246 |
need it anyway. |
247 |
- link against -lrt, as it might contain goodies such as fdatasync, |
248 |
and often aio_read, which we do not use yet, but it shouldn't hurt. |
249 |
|
250 |
0.4 Mon Jul 11 05:28:25 CEST 2005 |
251 |
- added IO::AIO::flush, IO::AIO::poll. |
252 |
- make callback argument optional for all calls. |
253 |
|
254 |
0.3 Mon Jul 11 01:42:37 CEST 2005 |
255 |
- restructured manpage. |
256 |
- added SYNOPSIS with examples for Event, Glib, Tk. |
257 |
- remove explicit context support so I don't have to maintain |
258 |
it for threaded perls (i.e. it should work now :). |
259 |
|
260 |
0.2 Sun Jul 10 22:57:12 CEST 2005 |
261 |
- return immediately from poll_wait if nreqs==0. |
262 |
- get rid of the request pipe and use faster thread-thread |
263 |
communications for most requests and responses. |
264 |
|
265 |
0.1 |
266 |
- cloned from Linux::AIO, which is very similar. |
267 |
|