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? (http://kerneltrap.org/node/6505 http://lwn.net/Articles/178199/) |
7 |
TODO: aio_fcntl, at least for file-locking |
8 |
TODO: fallocate, if it wouldn't be so useless. |
9 |
TODO: aio_mincore? |
10 |
TODO: getxattr etc.? |
11 |
TODO: MS_xxx, posix_fadvise, posix_madvise |
12 |
TODO: realpath |
13 |
|
14 |
4.0 |
15 |
- INCOMPATIBLE CHANGE: fork is no longer supported (indeed, it never was), |
16 |
see FORK BEHAVIOUR in manpage for details. |
17 |
- passes testsuite on win32 now (activetstate 5.10.1, YMMV). |
18 |
- (libeio) fix a deadlock where a wakeup signal could be missed when |
19 |
a timeout occured at the same time. |
20 |
- (libeio) added realpath. |
21 |
- (libeio) added fallocate. |
22 |
- (libeio) disabling sendfile on darwin unfortunately broke the emulation. |
23 |
- (libeio) do not acquire any locks when forking. |
24 |
- (libeio) use fewer time() syscalls when waiting for new requests. |
25 |
|
26 |
3.93 Wed Jun 29 23:44:18 CEST 2011 |
27 |
- ECB.H WAS MISSING. |
28 |
|
29 |
3.92 Wed Jun 29 14:45:41 CEST 2011 |
30 |
- ecb.h was missing. |
31 |
|
32 |
3.91 Wed Jun 29 13:24:42 CEST 2011 |
33 |
- (libeio) work around a Linux (and likely FreeBSD and other |
34 |
kernels) bug where sendfile would not transfer all the requested |
35 |
bytes on large transfers, using a heuristic. |
36 |
- buggy sendfile caused aio_move/copy to sometimes fail for big |
37 |
files (fortunately it checks that the whole file has been |
38 |
transferred...) |
39 |
- use libecb for higher performance and higher portability. |
40 |
- (libeio) disable sendfile on darwin, it's too broken. |
41 |
- disable fork tests on !linux, as only linux supports mixing |
42 |
pthread and fork in perl. |
43 |
- document the fact that fork doesn't work anymore when using |
44 |
this module, on anything !GNU/Linux. |
45 |
- increase timeout in t/04_fork.t, as too many CPAN-tester setups |
46 |
run on a 0.1MHz cpu. Or so. |
47 |
|
48 |
3.9 Fri May 27 02:43:47 CEST 2011 |
49 |
- (libeio) fix memory corruption in aio_readdirx for the flags |
50 |
combination READDIR_STAT_ORDER | READDIR_DIRS_FIRST. |
51 |
- add lots of newer POSIX and GNU/Linux-specific open |
52 |
flags. |
53 |
|
54 |
3.8 Sun Mar 27 12:25:33 CEST 2011 |
55 |
- use nonstandard but maybe-working-on-bsd fork technique. |
56 |
- support a max_idle value of 0. |
57 |
- support setting of idle timeout value (IO::AIO::idle_timeout). |
58 |
|
59 |
3.72 Fri Feb 11 04:25:38 CET 2011 |
60 |
- use _POSIX_MEMLOCK_RANGE to detect mlock/munlock. |
61 |
- aio_mknod always used a dev_t value of 0. |
62 |
- new treescan option: --grep. |
63 |
- add more S_IF macros, and major/minor/makedev "macros". |
64 |
|
65 |
3.71 Thu Dec 30 08:18:46 CET 2010 |
66 |
- the numerical result value passed to callbacks did not stringify |
67 |
correctly, due to internal reuse and failure to reset the sv flags. |
68 |
- actually test for posix_[mf]advise, as at least uClibc defines |
69 |
_POSIX_ADVISORY_INFO without actually having any of the required |
70 |
funcitonality. ugh. |
71 |
|
72 |
3.7 Mon Nov 1 23:00:34 CET 2010 |
73 |
- implement/add madvise, mmap, munmap, aio_mlockall, munlockall, |
74 |
aio_mlock, munlock, msync and mtouch to @EXPORT_OK. |
75 |
- document the sad state of affairs w.r.t. pthread on many bsds. |
76 |
- do not enable mmap on systems without _POSIX_MAPPED_FILES |
77 |
(openbsd 4.8). |
78 |
- do not leak memory in IO::AIO::mmap when the scalar already |
79 |
had string-data. |
80 |
- add O_RDWR, O_APPEND and O_EXCL symbols and semi-document |
81 |
them. |
82 |
- cache the result IV, for a minor speedup in the common case. |
83 |
- croak when an mmapped-scalar changes location, to detect |
84 |
user-errors better. |
85 |
- fix aio_readlink prototype. |
86 |
|
87 |
3.65 Wed Mar 31 02:45:05 CEST 2010 |
88 |
- actually use PATH_MAX instead of NAME_MAX for readlink, as to not |
89 |
cut off long pathnames in aio_readlink (based on patch by |
90 |
Rasmus Andersson). |
91 |
- a double fork partially killed the event pipe (great testcase |
92 |
by dormando). affects IO::AIO, BDB and Async::Interrupt. |
93 |
- suck steve's dick and rename our symbols to avoid colliding |
94 |
with nonstandard identifier spam on the broken os x pseudo-os. |
95 |
affects IO::AIO and BDB. |
96 |
|
97 |
3.6 Mon Jan 11 00:43:39 CET 2010 |
98 |
- (libeio) more fixes for the freebsd/apple sendfile - broken by |
99 |
manpage standards, but apparently correct accoridng to actual |
100 |
kernel sources. |
101 |
- add IO::AIO::mmap/munmap support. |
102 |
- add IO::AIO::m(un)lockall support. |
103 |
- clean up manpage. |
104 |
|
105 |
3.5 Thu Jan 7 21:25:04 CET 2010 |
106 |
- (liebio) fix freebsd sendfile (Vanilla Hsu). |
107 |
- (libeio) also fix darwin sendfile that suffered from a similar bug. |
108 |
- add aio_statvfs as interface to statvfs/fstatvfs. |
109 |
- work around buggy symlink() on cygwin in the testsuite. |
110 |
- wtf. freebsd-8.0 actually passes the testsuite! I knew |
111 |
they could implement fork and semaphores one day! cheers! |
112 |
|
113 |
3.4 Sat Jan 2 15:13:04 CET 2010 |
114 |
- (libeio) max_poll_time was not properly converted to ticks. |
115 |
- clarify cancel_subs description. |
116 |
- IO::AIO::sendfile did not actually return the return value |
117 |
from the sendfile call. |
118 |
- implement aio_msync, aio_mtouch. |
119 |
- (libeio) tentatively support darwin in sendfile. |
120 |
|
121 |
3.31 Thu Nov 12 02:14:29 CET 2009 |
122 |
- fix result status documentation of aio_copy, aio_move. |
123 |
- speed up object creation considerably by avoiding |
124 |
hash lookups. |
125 |
|
126 |
3.3 Wed Aug 5 13:52:58 CEST 2009 |
127 |
- use common::sense. |
128 |
- use common schmorp.h header. |
129 |
- allow integers (file descriptors) in addition to file handles. |
130 |
- take advantage of linux' eventfd (but it seems slower on smp?). |
131 |
- use poll not select on posix platforms. |
132 |
|
133 |
3.261 Wed Jul 1 10:11:51 CEST 2009 |
134 |
- more 0S X workarounds (patch by Tokuhiro Matsuno). |
135 |
|
136 |
3.26 Tue Jun 30 09:33:26 CEST 2009 |
137 |
- 0S X of course claims to be posix 2008 but lacks posix_fadvise. |
138 |
try to work around this horribly broken OS in a somewhat hackish |
139 |
way. might help other os's too. |
140 |
|
141 |
3.25 Sat Jun 27 05:18:26 CEST 2009 |
142 |
- added IO::AIO::fadvise and IO::AIO::sendfile. |
143 |
- (libeio) replaced quicksort+insertion sort by a tuned radix |
144 |
sort + insertion sort, resulting in comparable runtime (usually |
145 |
faster) to the old version, but without any ill side effects on |
146 |
degenerated (for quicksort) data. |
147 |
- (libeio) correctly sort dirs by inodes if we have perfect type |
148 |
knowledge and DIRS_FIRST is used. |
149 |
- (libeio) internally the wrong DT_* constants were used, but no |
150 |
known system has differing EIO_DT_*/DT_* values, so not an issue. |
151 |
- removed a (harmless) assert that was left over on the code |
152 |
but should not have been. |
153 |
- use more correct types (change IO length from IV to VAL64, |
154 |
which makes a difference on systems without 64 bit perls). |
155 |
|
156 |
3.23 Sat Jun 13 16:57:58 CEST 2009 |
157 |
- fix off-by-one bug in aio_readdir that was introduced in the |
158 |
rushed 3.22 release. |
159 |
|
160 |
3.22 Sat Jun 13 15:32:40 CEST 2009 |
161 |
- speed up readdirx sort algorithm slightly. |
162 |
- bin/treescan was missing from distro tarball. |
163 |
|
164 |
3.21 Fri Jun 12 18:45:53 CEST 2009 |
165 |
- new options --dirs and --files for treescan. |
166 |
- install bin/treescan by default. |
167 |
- (libeio) aio_readdir can now be cancelled while executing. |
168 |
- fix a printf format string for 64 bit systems (could lead |
169 |
to problems on big endian 64 bit systems). |
170 |
- do not use qsort() but our own algorithm: glibc initialises |
171 |
the whole locale and I/O subsystem inside qsort, causing |
172 |
a stack overflow on 32 bit machines. The new sort uses much less |
173 |
stack and is more than twice as fast in typical situations. |
174 |
|
175 |
3.2 Sun Jun 7 20:30:05 CEST 2009 |
176 |
- (libeio) pwrite emulation was even more flawed than intended and did |
177 |
not restore the file offset. |
178 |
- add aio_readdirx, which can return inode and filetype and sort |
179 |
the names in various ways. |
180 |
- unfortunately, utime, chmod, chown on an open file that has just |
181 |
been written can easily block, which caused aio_copy to block the |
182 |
process. no more! |
183 |
- no longer rely on dst path in aio_copy when futime is available. |
184 |
|
185 |
3.19 Tue Apr 21 22:05:21 CEST 2009 |
186 |
- more perl 5.10 workarounds for aio_read and write. |
187 |
- aio_write no longer modifies the sv (if possible). |
188 |
- aio_read now works correctly with magic values. |
189 |
|
190 |
3.18 Sun Apr 19 21:17:32 CEST 2009 |
191 |
- better diagnostics when some aio_* functions get passed an illegal fh. |
192 |
- try to avoid crashes due to incompatible 5.10 API changes. grmbl. |
193 |
|
194 |
3.17 Thu Nov 20 08:45:36 CET 2008 |
195 |
- (libeio) added aio_sync_file_range (untested). |
196 |
- add aio_busy to @IO::AIO::AIO_REQ. |
197 |
|
198 |
3.16 Wed Oct 22 18:28:01 CEST 2008 |
199 |
- use SvREFCNT_inc instead of SvREFCNT_inc_NN in a non-speed critical |
200 |
part to improve portability to perl 5.8 (reported by szymon). |
201 |
|
202 |
3.15 Mon Oct 13 00:39:55 CEST 2008 |
203 |
- automatic removal of feeders was broken. |
204 |
- (libeio) use a more robust method to detect whether a feeder |
205 |
has added something to the group or not. |
206 |
|
207 |
3.1 Thu Oct 2 13:34:40 CEST 2008 |
208 |
- pre-deref the passed callback object, for increased speed |
209 |
and decreased memory usage. |
210 |
- call on_next_submit callback even for aio_group - impact unknown, |
211 |
but seems more correct. |
212 |
- $req->cb now returns the original callback. |
213 |
- (libeio) pass EIO_GROUP/aio_group requests through the queue without |
214 |
requiring a thread switch. |
215 |
- (libeio) status code of aio_futime and aio_utime was always passed 0. |
216 |
- do some other ยต-optimisations. |
217 |
|
218 |
3.07 Sat Aug 2 16:06:13 CEST 2008 |
219 |
- do not include LIBS in autoconf tests. |
220 |
|
221 |
3.06 Tue Jul 15 12:41:32 CEST 2008 |
222 |
- move preadwritelock definition before it's first reference, |
223 |
to make it compile again on broken platforms (they still exist) |
224 |
or badly cofngiured perls (redhat...), reported by Rob Bloodgood. |
225 |
|
226 |
3.05 Thu Jun 19 23:23:52 CEST 2008 |
227 |
- work around perl overriding readdir etc. with thread-unsafe |
228 |
(sic!) versions. |
229 |
|
230 |
3.04 Wed Jun 18 01:35:38 CEST 2008 |
231 |
- (libeio) fix eio_mknod, which confused it's arguments. |
232 |
- (libeio) do not use readdir_r, as Ulrich Drepper pointed out that this |
233 |
is stupid. |
234 |
- (libeio) fix eio__readahead prototype, patch by Jost Krieger. |
235 |
- (libeio) fix a bug that could potentially cause IO::AIO |
236 |
not to get initialised properly. |
237 |
|
238 |
3.03 Thu May 29 05:33:30 CEST 2008 |
239 |
- (libeio) correctly call pthread_attr_destroy (). |
240 |
- (libeio) work around broken bsd headers once more. |
241 |
- reduce shared library size again by not including wrappers. |
242 |
- max_outstanding could cause poll to enter a busy-waiting loop. |
243 |
- document the new IO::AIO::poll_cb result value. |
244 |
|
245 |
3.02 Mon May 12 02:32:02 CEST 2008 |
246 |
- fix a memory leak on aio_readlink. |
247 |
- bring back working fchmod. |
248 |
- nop and busy now set result to 0. |
249 |
- set errno to ENOMEM when allocation fails. |
250 |
|
251 |
3.01 Sun May 11 03:07:03 CEST 2008 |
252 |
- (libeio) make it compile on systems without readahead or readdir_r. |
253 |
- (libeio) improve configure check for readahead availability. |
254 |
- do not try to link against -lrt. |
255 |
- use a separate configure script for IO::AIO (not the libeio one). |
256 |
|
257 |
3.0 Sun May 11 00:57:14 CEST 2008 |
258 |
- added bin/treescan to the distribution. |
259 |
- switched to using libeio. |
260 |
- LOTS OF INCOMPATIBLE CHANGES: |
261 |
- remove signal functionality, it is not worth the effort. |
262 |
- max_outstanding no longer returns the previous number of requests. |
263 |
- poll_cb no longer returns number of requests. |
264 |
|
265 |
2.62 Sat Apr 26 13:59:33 CEST 2008 |
266 |
- port to solaris perls configured for old posix |
267 |
(analysed by Jost Krieger). |
268 |
- keep a reference to the perl filehandle object in aio_close, |
269 |
so it doesn't get closed prematurely, leading to ugly races. |
270 |
|
271 |
2.61 Wed Apr 16 18:45:02 CEST 2008 |
272 |
- fix treescan output duplication and improve output ordering. |
273 |
also display files given on the commandline. |
274 |
- use a different algorithm for aio_close that is faster |
275 |
and probably has even lower chances of blocking. |
276 |
- do our own stack memory management for threads - linux |
277 |
allocates outrageous amounts of VM (not so bad unless you |
278 |
use mlockall...), which severely limits the number of threads |
279 |
on 32-bit arches: stack size is the larger of PTHREAD_STACK_MIN |
280 |
and 4096 * sizeof (long) (usually this is 16..64k). |
281 |
|
282 |
2.6 Sun Mar 30 08:28:11 CEST 2008 |
283 |
- added aio_sync. |
284 |
- added aio_pathsync. |
285 |
- fix prototypes of (void) functions. |
286 |
|
287 |
2.51 Sat Oct 6 16:04:54 CEST 2007 |
288 |
- perlio isn't generally threadsafe, so aio_close could not work reliably. |
289 |
aio_close now tries a safe hack that might not be that asynchronous |
290 |
(see the manpage for details). |
291 |
- discard callback return values as to not grow the stack endlessly |
292 |
when poll_cb handles many requests without returning. |
293 |
- minor code cleanups. |
294 |
|
295 |
2.5 Thu Oct 4 14:49:08 CEST 2007 |
296 |
- replaced _fd2fh with faster xs code. |
297 |
- aio_close will now try to do "the right thing" and thus might |
298 |
work sensibly for the very first time. |
299 |
|
300 |
2.41 Mon Sep 24 21:28:21 CEST 2007 |
301 |
- after fork, invest some work to make sure that the poll_fileno |
302 |
stays stable (by dup2'ing the new fd over the old one), to make |
303 |
it easier for programs/libs that don't control their forking |
304 |
behaviour and cannot recreate their watchers. |
305 |
|
306 |
2.4 Sun Aug 5 18:44:22 CEST 2007 |
307 |
- add aio_truncate, aio_chmod, aio_chown, aio_utime (linux |
308 |
successfully demonstrated that you can block on futimes...). |
309 |
- allow undef as fileoffset for aio_read/write and use read/write(2) |
310 |
internally (useful for sockets or O_APPEND handles). |
311 |
- allow undef for length in aio_write. |
312 |
- negative dataoffsets work as expected now in aio_read/aio_write. |
313 |
- use NV instead of UV for 32 bit perls and file offsets, as NVs |
314 |
have a larger range then. |
315 |
- shared code between BDB and IO::AIO. |
316 |
- aio_busy was completely broken. now fixed. |
317 |
- readahead emulation now returns something. |
318 |
- correctly set errno to ENOSYS on unimplemented functions |
319 |
(should never happen, though). |
320 |
- large changes to make it partially compile and run on win32, |
321 |
but win32 lacks too much functionality, and perl overrides way |
322 |
too many functions with crashing versions. |
323 |
|
324 |
2.33 Tue Jan 23 23:55:41 CET 2007 |
325 |
- fix install path (Andreas J. Koenig). |
326 |
|
327 |
2.32 Mon Jan 22 16:56:23 CET 2007 |
328 |
- added aio_rmtree. |
329 |
- wow, aio_mkdir was missing. |
330 |
- aio_load did return undef on error, not -1. |
331 |
- use prefixexp not prefix in autoconf.pm |
332 |
(suggested by ... rt.cpan.org unreachable). |
333 |
- avoid installing autoconf.pm. |
334 |
|
335 |
2.31 Sat Jan 6 03:46:02 CET 2007 |
336 |
- added aio_load. |
337 |
|
338 |
2.3 Sat Dec 23 05:48:07 CET 2006 |
339 |
- fix off-by-one bug in aio_read, causing memory corruption |
340 |
or crashes. |
341 |
|
342 |
2.21 Fri Dec 22 05:03:38 CET 2006 |
343 |
- minor doc updates. |
344 |
- minor cleanups. |
345 |
- IO::AIO::poll returns request count. |
346 |
- undocumented and unfinished async-signal-reporting. |
347 |
|
348 |
2.2 Tue Oct 31 00:31:00 CET 2006 |
349 |
- minor doc updates. |
350 |
- added aio_readlink. |
351 |
- properly zero-terminate aio_read buffer. |
352 |
- fix aio_read/write with nonzero dataoffset. |
353 |
- reduced size of aio request structure. |
354 |
- better diagnostics when pathnames etc. are not byte-encoded; |
355 |
do not leak request in that case, either. |
356 |
|
357 |
2.1 Sun Oct 29 02:01:57 CET 2006 |
358 |
- INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
359 |
IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
360 |
- fix a bug in aio_move where it would fail everytime |
361 |
it would have to copy files manually. |
362 |
- doc updates. |
363 |
- start threads only on (very low) demand. |
364 |
- end superfluous threads automatically after 10s idling. |
365 |
- add IO::AIO::max_idle to control this. |
366 |
- some tuning applied to random places. |
367 |
- add aio_mknod. |
368 |
|
369 |
2.0 Thu Oct 26 18:27:58 CEST 2006 |
370 |
- minor incompatibility: max_outstanding semantics have changed. |
371 |
- major rewrite, consider all functionality EXPERIMENTAL. |
372 |
- aio requests can be cancelled. |
373 |
- callbacks can be (re-)set later. |
374 |
- aio requests can choose between 9 priorities. |
375 |
- add aio_group: aio requests can be grouped |
376 |
into composite requests. |
377 |
- generator interface for groups. |
378 |
- add IO::AIO::poll_some. |
379 |
- major documentation improvements. |
380 |
- very minor bugfixes. |
381 |
- add aio_nop for dummy requests. |
382 |
- add aio_busy mainly for benchmarking and debugging. |
383 |
- use "fast" mutexes on linux. |
384 |
- use dynamic (and larger) buffers instead of |
385 |
putting them onto the stack. |
386 |
- optimise x86 and amd64 a bit. |
387 |
- better error checking when ENOMEM. |
388 |
- hopefully fix mem and dirp leaks on fork. |
389 |
- warn about broken perl malloc. |
390 |
- compiles on cygwin (but pread is broken on cygwin |
391 |
which is not detected by autoconf, so you have to |
392 |
edit autoconf/config.h yourself to enable emulation). |
393 |
|
394 |
1.8 Sat Jun 24 22:20:44 CEST 2006 |
395 |
- add and/or document aio_rename, aio_link, aio_symlink and aio_move. |
396 |
|
397 |
1.73 Wed Mar 1 22:49:32 CET 2006 |
398 |
- codename "jost" |
399 |
- AIX/Solaris(?) aren't being helpful again, hardcode a value |
400 |
for NAME_MAX and drive with your eyes closed (reported by |
401 |
wcooley@nakedape.cc). |
402 |
|
403 |
1.72 Thu Feb 2 00:40:44 CET 2006 |
404 |
- another minor bug in aio_scandir, again, the callback was called |
405 |
twice when the directory can't be stat'ed. |
406 |
|
407 |
1.71 Mon Dec 26 20:20:03 CET 2005 |
408 |
- grr, another minor bug in aio_scandir, this time resulting |
409 |
in a runtime error. |
410 |
|
411 |
1.7 Mon Dec 26 19:29:48 CET 2005 |
412 |
- aio_scandir errornously called callback twice in some cases. |
413 |
- added AnyEvent example. |
414 |
|
415 |
1.61 Wed Sep 7 19:40:42 CEST 2005 |
416 |
- no longer include $Config{libs}, as distros include too much |
417 |
garbage in there that is not installed on typical systems |
418 |
(for example, debian links perl against -lgdbm and many more |
419 |
libraries despite perl not needing it, slowing down startup and |
420 |
wasting memory). |
421 |
|
422 |
1.6 Tue Aug 30 17:44:44 CEST 2005 |
423 |
- added aio_readdir, aio_scandir. |
424 |
- added aio_sendfile, including sendfile emulation when sendfile |
425 |
isn't available. Linux, FreeBSD and HP-UX might work (sendfilev |
426 |
support is unavailable on my solaris 8 machine, but preliminary |
427 |
code is there. What's lacking is configury stuff and testing...). |
428 |
|
429 |
1.5 Sat Aug 20 02:32:19 CEST 2005 |
430 |
- create the aio threads on demand only (good after forking). |
431 |
- somebody nagged about C89 compatibility, so this release should |
432 |
be C89 compatible, but no guarantees for future ones. |
433 |
- abort when the configure script fails. |
434 |
|
435 |
1.4 Wed Aug 17 08:07:27 CEST 2005 |
436 |
- forgot to recreate the pipe after forking, this could cause |
437 |
deadlocks. |
438 |
|
439 |
1.3 Wed Aug 17 07:25:54 CEST 2005 |
440 |
- properly propagate exceptions from callbacks, instead of silently |
441 |
eating them. |
442 |
- use a different fork algorithm that avoids recursive callback |
443 |
invocations and reduces fork latency in busy processes. |
444 |
- fix a bug where the buffer scalar in aio_read got assigned the wrong |
445 |
length after reading. |
446 |
- fix a bug where the child would process some outstanding requests |
447 |
from the parent. |
448 |
- make a copy of the callback scalar, to avoid some very uncommon |
449 |
(but valid) usages where the callback scalar changes. |
450 |
- the callback was never freed in aio_read/aio_write. |
451 |
- aio_read/aio_write will now set the readonly flag on the data |
452 |
scalar for better error reporting. |
453 |
|
454 |
1.2 Wed Aug 17 01:22:58 CEST 2005 |
455 |
- verified on cygwin, linux/amd64+ia32+alpha, solaris. |
456 |
- use SvPVbyte for filenames now. You *need* to encode your strings to |
457 |
the external filename encoding first. |
458 |
- fix the bug where filehandles could leak (again). |
459 |
- clarify freebsd situation: make it work and tell me how, or go away |
460 |
with your underdocumented, broken and outdated OS. |
461 |
- fork'ing should now be officially supported (where linux and |
462 |
posix contradict each other, I followed posix). |
463 |
- only call read() once per poll, not twice (speedup). |
464 |
- new functions: aio_rmdir, aio_symlink. |
465 |
- force callback arguments to be undef or a reference, for better |
466 |
error catching. |
467 |
|
468 |
1.1 Sun Aug 7 05:27:23 CEST 2005 |
469 |
- added configure script to detect availability of some calls. |
470 |
(imperfect, as we cannot detect things like -lrt due to missing |
471 |
feedback yet). |
472 |
- emulate fdatasync via fsync, pread/pwrite via unsafe read/write |
473 |
and readahead via manual reading on platforms lacking those. |
474 |
- ported t/03_errors.t+framework by Brad Fitzpatrick |
475 |
from Linux::AIO. |
476 |
- removed XOPEN_SOURCE because stupid solaris 9 doesn't |
477 |
compile with it enabled. Probably breaks others. Oh well. |
478 |
- try to be more careful with errno on non-threaded perls. |
479 |
- cygwins perl insisted on a matching open mode when morphing |
480 |
the fd to an fh. |
481 |
- convert sv's to "byte" encoding in aio_read/aio_write. Might |
482 |
do that for pathnames, too, in the future, so make sure you |
483 |
properly encode your pathnames. |
484 |
|
485 |
1.0 Fri Jul 22 10:43:30 CEST 2005 |
486 |
- aio_open did not reliably deliver (open) filehandles. |
487 |
- all aio_*-functions keep a reference to the filehandle so |
488 |
it doesn't go away while the aio requests is working on it. |
489 |
- disable warnings in AIO.pm. |
490 |
|
491 |
0.9 Wed Jul 20 23:55:04 CEST 2005 |
492 |
- allow recursive calls to poll_cb, use less locking calls, too. |
493 |
- fix the Gtk2-example in the SYNOPSIS. |
494 |
- use pread/pwrite not pread64/pwrite64. this was simply a leftover |
495 |
from Linux::AIO. |
496 |
(found by compiling on cygwin. Unfortunately, cygwin declares but |
497 |
does not implement pread/pwrite). |
498 |
- only include <sys/syscall.h> on linux. |
499 |
|
500 |
0.5 Tue Jul 12 13:03:09 CEST 2005 |
501 |
- removed #include <endian.h>, it's not portable, and we don't |
502 |
need it anyway. |
503 |
- link against -lrt, as it might contain goodies such as fdatasync, |
504 |
and often aio_read, which we do not use yet, but it shouldn't hurt. |
505 |
|
506 |
0.4 Mon Jul 11 05:28:25 CEST 2005 |
507 |
- added IO::AIO::flush, IO::AIO::poll. |
508 |
- make callback argument optional for all calls. |
509 |
|
510 |
0.3 Mon Jul 11 01:42:37 CEST 2005 |
511 |
- restructured manpage. |
512 |
- added SYNOPSIS with examples for Event, Glib, Tk. |
513 |
- remove explicit context support so I don't have to maintain |
514 |
it for threaded perls (i.e. it should work now :). |
515 |
|
516 |
0.2 Sun Jul 10 22:57:12 CEST 2005 |
517 |
- return immediately from poll_wait if nreqs==0. |
518 |
- get rid of the request pipe and use faster thread-thread |
519 |
communications for most requests and responses. |
520 |
|
521 |
0.1 |
522 |
- cloned from Linux::AIO, which is very similar. |
523 |
|