1 |
Revision history for IO::AIO |
2 |
|
3 |
TODO: scandir - some dirs mostly contain subdirs - invert logic? |
4 |
TODO: aio_cptree/mvtree |
5 |
TODO: vmsplice? (http://kerneltrap.org/node/6505 http://lwn.net/Articles/178199/) |
6 |
TODO: getxattr, setxattr, listxattr, removexattr etc.? |
7 |
TODO: emulation for splice? |
8 |
TODO: syncfs/sync windows: |
9 |
TODO: posix_fallocate when flags=0 |
10 |
http://stackoverflow.com/questions/65170/how-to-get-name-associated-with-open-handle/5286888#5286888 |
11 |
http://blogs.msdn.com/b/adioltean/archive/2005/04/16/408947.aspx |
12 |
http://msdn.microsoft.com/en-us/library/aa366789%28v=vs.85%29.aspx |
13 |
http://msdn.microsoft.com/en-us/library/windows/desktop/aa366789%28v=vs.85%29.aspx |
14 |
http://msdn.microsoft.com/en-us/library/windows/desktop/aa364425%28v=vs.85%29.aspx |
15 |
http://msdn.microsoft.com/en-us/library/windows/desktop/aa364963%28v=vs.85%29.aspx |
16 |
http://msdn.microsoft.com/en-us/library/windows/desktop/aa364996%28v=vs.85%29.aspx |
17 |
http://msdn.microsoft.com/en-us/library/windows/desktop/aa364994%28v=vs.85%29.aspx |
18 |
TODO: extra socket/tcp constants &c? |
19 |
TODO: execveat? really? |
20 |
TODO: http://lwn.net/Articles/593918/ SHMEM_SET_SEALS, SHMEM_GET_SEALS, SEAL_SHRINK, SEAL_GROW, SEAL_WRITE, int memfd_create(const char *name, u64 size, u64 flags);, MFD_CLOEXEC, int fd = open("/tmp", O_RDWR | O_TMPFILE | O_EXCL, S_IRWXU); |
21 |
TODO: fcntl F_GETLKP/F_SETLKP/F_SETLKPW, http://lwn.net/Articles/586904/ |
22 |
TODO: linkat + AT_EMPTY_PATH |
23 |
TODO: name_to_handle_At + open_by_handle_at = clone fds |
24 |
TODO: lchown |
25 |
TODO: mount/umount2/MS_* flags |
26 |
|
27 |
TODO: rewrite rmtree et al. to support working directories (also speed them up) |
28 |
TODO: maybe IO::AIO leaks fds when requests are cancelled? maybe initialise result to -1? |
29 |
TODO: aio_wd should use O_PATH on linux, due to lacking O_SEARCH (http://comments.gmane.org/gmane.linux.file-systems/33611) |
30 |
http://www.openwall.com/lists/musl/2013/02/23/4 |
31 |
TODO: copy_file_range, copy_file_range |
32 |
TODO: FIDEDUPERANGE, file_ioctl.c |
33 |
TODO: mincore |
34 |
TODO: SOCK_CLOEXEc etc. |
35 |
TODO: unlinkat flags, linkat flags |
36 |
TODO: 4.6 |
37 |
|
38 |
- add a bunch of (mostly linux-specific) constants for use in ioctls |
39 |
(see aio_ioctl docs). |
40 |
- treescan now has a proper manpage and useful --help output. |
41 |
- new option --sync in treescan, to sync everything in a subtree. |
42 |
- changed default for aio_msync flags to MSYNC_SYNC. |
43 |
- document offset/length behaviour of mprotect/madvise. |
44 |
- support linux's renameat2 (via aio_rename2). |
45 |
- add aio_rename2, an aio_rename with flags. |
46 |
- add F_DUPPFD_CLOEXEC, F_OFD_[SG]ETLKW? constants. |
47 |
- add FALLOC_FL_INSERT_RANGE, FALLOC_FL_UNSHARE_RANGE. |
48 |
|
49 |
4.34 Sun May 1 19:18:24 CEST 2016 |
50 |
- def0.h was not properly generated during previous release, causing |
51 |
compile errors on various platforms. |
52 |
- major/minor were accidentally switched (reported by Alexander Lishenyuk). |
53 |
- removed duplicate definition of MAP_HUGETLB, that was found due to |
54 |
a bug in Perl::Tidy (testcase by alaska332@gmail.com). |
55 |
- added (untested!) aio_fcntl, aio_ioctl requests. |
56 |
- (libeio) names set via prctl are truncated to 15 chars + nul, not 16, |
57 |
as manpages-dev originally claimed. |
58 |
|
59 |
4.33 Mon Jan 18 12:50:10 CET 2016 |
60 |
- add IO::AIO::pipe2 function. |
61 |
- added support for FALLOC_FL_COLLAPSE_RANGE and FALLOC_FL_ZERO_RANGE |
62 |
constants. |
63 |
- added support for O_TMPFILE and O_PATH constants. |
64 |
- added support for MAP_FIXED, MAP_GROWSDOWN,MAP_32BIT, MAP_HUGETLB, MAP_STACK |
65 |
consdtants, whether they can be sensibly used or not. |
66 |
- use NO_INIT where applicable. |
67 |
- update libecb. |
68 |
- added stability canary support. |
69 |
- updated linux super magic table to 4.3.3. |
70 |
|
71 |
4.32 Wed Feb 11 20:32:11 CET 2015 |
72 |
- replace off_t by STRLEN where appropriate, should not result in |
73 |
user-visible changes. |
74 |
- update ecb.h for C11 compatibility. |
75 |
|
76 |
4.31 Tue Jun 3 03:29:27 CEST 2014 |
77 |
- work around more 5.20 bugs. backwards compatibility my ass. |
78 |
|
79 |
4.3 Fri Apr 11 06:22:38 CEST 2014 |
80 |
- perl5porters broke Async::Interrupt, BDB, EV, IO::AIO, OpenCL |
81 |
without warning by switching the meaning of USE_SOCKETS_AS_HANDLES |
82 |
in 5.18. What's so attractive about giving a shit about backwards |
83 |
compatibility - I will never understand. |
84 |
|
85 |
4.2 Sat Jan 25 01:13:14 CET 2014 |
86 |
- aio_group could corrupt memory because it didn't restore |
87 |
the stack after req_submit. |
88 |
- be more careful on (e.g. permission) errors in bin/treescan. |
89 |
- work around changes in ExtUtils::MakeMaker. |
90 |
- (libeio) implement aio_realpath for win32. |
91 |
- (xthread) work around compile time bugs in ptw32. |
92 |
- added IO::AIO::pipesize. |
93 |
- (libecb) insignificant update. |
94 |
|
95 |
4.19 Sun Jan 6 12:47:26 CET 2013 |
96 |
- avoid endless loop in fiemap with some XFS files. |
97 |
- in aio_rename and aio_rmdir, specialcase the case of [$wd, "."] |
98 |
and call rename/rmdir instead of renameat/unlinkat. |
99 |
|
100 |
4.18 Thu Oct 11 07:01:26 CEST 2012 |
101 |
- fix unintended xthread_create by intentionalising it :) |
102 |
|
103 |
4.17 Thu Oct 11 05:19:47 CEST 2012 |
104 |
- rename aio_fallocate to aio_allocate, to match documentation. |
105 |
- add list of linux fsid values to aio_statvfs docs. |
106 |
- work around a bug in btrfs' FIEMAP ioctl implementation. |
107 |
- work around AIX bug: statvfs.f_fsid is a struct, not unsigned long |
108 |
as per unix spec. |
109 |
|
110 |
4.16 Tue Aug 14 05:39:03 CEST 2012 |
111 |
- aio_statvfs was wromgly marked for wd emulation, causing |
112 |
it to malfunction for paths. |
113 |
- fix a crash in aio_fiemap, when used on ranges without any |
114 |
extents. |
115 |
- work around linux kernel bug (at least in 3.2): kernel might |
116 |
trash fiemap input arguments. |
117 |
- work around linux kernel bug (at least in 3.2): kernel does not |
118 |
set FIEMAP_EXTENT_LAST on last segment. |
119 |
- work around linux kernel bug (at least in 3.2): kernel silently |
120 |
truncates segment count if a file. |
121 |
- make fh argument of IO::AIO::mmap optional, also add mmap/munmap |
122 |
to quick overview. |
123 |
- splice can corrupt data. |
124 |
- (libeio) remove pread/pwrite emulation. |
125 |
- do not grow mmaped scalars in aio_read. |
126 |
- add FALLOC_FL_PUNCH_HOLE and document fallocate. |
127 |
- provide pread/pwrite for win32, making it link again on native win32. |
128 |
|
129 |
4.15 Tue Apr 10 06:59:00 CEST 2012 |
130 |
- always include linux/types.h for fiemap, for compatibility to |
131 |
ancient systems (Paul Howarth). |
132 |
- experimental support for IO::AIO::splice and ::tee (no aio_...). |
133 |
- provide SEEK_HOLE and SEEK_DATA, if available. |
134 |
- work around (again!) an immensely stupid bug in RHEL, defining |
135 |
autoconf macros in linux system headers (analysed by Paul Howarth). |
136 |
|
137 |
4.14 Sat Apr 7 02:45:18 CEST 2012 |
138 |
- fix stat structure usage on windows, which caused bogus stat results. |
139 |
- (libeio) make readahead emulation behave more like actual readahead by never failing. |
140 |
- new request aio_seek. |
141 |
- new request aio_fiemap. |
142 |
- autogenerate the #ifdef/#define 0 blocks for symbols we export. |
143 |
|
144 |
4.12 Fri Dec 30 08:51:25 CET 2011 |
145 |
- realpath would return a random value if the name was too long, |
146 |
instead of -1. |
147 |
- port to c89. |
148 |
|
149 |
4.11 Mon Oct 10 00:24:11 CEST 2011 |
150 |
- libeio didn't compile on !linux (Paul Howarth). |
151 |
|
152 |
4.1 Sun Oct 9 10:24:11 CEST 2011 |
153 |
- IO::AIO did access uninitialised memory on unsuccessful stats. |
154 |
- (libeio) added syncfs syscall wrapper. |
155 |
- (libeio) set thread name on linux (ps -L/Hcx, top, gdb). |
156 |
- (libeio) support multiple indepenent working directories. |
157 |
- applied speling corrections by Nicholas Bamber. |
158 |
- tune treescan #threads and #outstanding requests a bit. |
159 |
- reduce the number of file descriptors in use at any one time |
160 |
in treescan, to avoid running out of them (and using too much ram). |
161 |
- take advantage of working directoriy abstractions |
162 |
in aio_scandir and treescan. |
163 |
- reduce compiled size by ~10% by not inlining some key functionality. |
164 |
- added --progress switch to treescan. |
165 |
|
166 |
4.0 Mon Jul 18 05:01:10 CEST 2011 |
167 |
- INCOMPATIBLE CHANGE: fork is no longer supported (indeed, it never was), |
168 |
see FORK BEHAVIOUR in manpage for details. |
169 |
- passes testsuite on win32 now and may actually work |
170 |
(activestate 5.10.1 / MSVC6, YMMV). |
171 |
- (libeio) fix a deadlock where a wakeup signal could be missed when |
172 |
a timeout occured at the same time. |
173 |
- (libeio) added realpath. |
174 |
- (libeio) added fallocate. |
175 |
- (libeio) disabling sendfile on darwin unfortunately broke the emulation. |
176 |
- (libeio) do not acquire any locks when forking. |
177 |
- (libeio) use fewer time() syscalls when waiting for new requests. |
178 |
|
179 |
3.93 Wed Jun 29 23:44:18 CEST 2011 |
180 |
- ECB.H WAS MISSING. |
181 |
|
182 |
3.92 Wed Jun 29 14:45:41 CEST 2011 |
183 |
- ecb.h was missing. |
184 |
|
185 |
3.91 Wed Jun 29 13:24:42 CEST 2011 |
186 |
- (libeio) work around a Linux (and likely FreeBSD and other |
187 |
kernels) bug where sendfile would not transfer all the requested |
188 |
bytes on large transfers, using a heuristic. |
189 |
- buggy sendfile caused aio_move/copy to sometimes fail for big |
190 |
files (fortunately it checks that the whole file has been |
191 |
transferred...) |
192 |
- use libecb for higher performance and higher portability. |
193 |
- (libeio) disable sendfile on darwin, it's too broken. |
194 |
- disable fork tests on !linux, as only linux supports mixing |
195 |
pthread and fork in perl. |
196 |
- document the fact that fork doesn't work anymore when using |
197 |
this module, on anything !GNU/Linux. |
198 |
- increase timeout in t/04_fork.t, as too many CPAN-tester setups |
199 |
run on a 0.1MHz cpu. Or so. |
200 |
|
201 |
3.9 Fri May 27 02:43:47 CEST 2011 |
202 |
- (libeio) fix memory corruption in aio_readdirx for the flags |
203 |
combination READDIR_STAT_ORDER | READDIR_DIRS_FIRST. |
204 |
- add lots of newer POSIX and GNU/Linux-specific open |
205 |
flags. |
206 |
|
207 |
3.8 Sun Mar 27 12:25:33 CEST 2011 |
208 |
- use nonstandard but maybe-working-on-bsd fork technique. |
209 |
- support a max_idle value of 0. |
210 |
- support setting of idle timeout value (IO::AIO::idle_timeout). |
211 |
|
212 |
3.72 Fri Feb 11 04:25:38 CET 2011 |
213 |
- use _POSIX_MEMLOCK_RANGE to detect mlock/munlock. |
214 |
- aio_mknod always used a dev_t value of 0. |
215 |
- new treescan option: --grep. |
216 |
- add more S_IF macros, and major/minor/makedev "macros". |
217 |
|
218 |
3.71 Thu Dec 30 08:18:46 CET 2010 |
219 |
- the numerical result value passed to callbacks did not stringify |
220 |
correctly, due to internal reuse and failure to reset the sv flags. |
221 |
- actually test for posix_[mf]advise, as at least uClibc defines |
222 |
_POSIX_ADVISORY_INFO without actually having any of the required |
223 |
funcitonality. ugh. |
224 |
|
225 |
3.7 Mon Nov 1 23:00:34 CET 2010 |
226 |
- implement/add madvise, mmap, munmap, aio_mlockall, munlockall, |
227 |
aio_mlock, munlock, msync and mtouch to @EXPORT_OK. |
228 |
- document the sad state of affairs w.r.t. pthread on many bsds. |
229 |
- do not enable mmap on systems without _POSIX_MAPPED_FILES |
230 |
(openbsd 4.8). |
231 |
- do not leak memory in IO::AIO::mmap when the scalar already |
232 |
had string-data. |
233 |
- add O_RDWR, O_APPEND and O_EXCL symbols and semi-document |
234 |
them. |
235 |
- cache the result IV, for a minor speedup in the common case. |
236 |
- croak when an mmapped-scalar changes location, to detect |
237 |
user-errors better. |
238 |
- fix aio_readlink prototype. |
239 |
|
240 |
3.65 Wed Mar 31 02:45:05 CEST 2010 |
241 |
- actually use PATH_MAX instead of NAME_MAX for readlink, as to not |
242 |
cut off long pathnames in aio_readlink (based on patch by |
243 |
Rasmus Andersson). |
244 |
- a double fork partially killed the event pipe (great testcase |
245 |
by dormando). affects IO::AIO, BDB and Async::Interrupt. |
246 |
- suck steve's dick and rename our symbols to avoid colliding |
247 |
with nonstandard identifier spam on the broken os x pseudo-os. |
248 |
affects IO::AIO and BDB. |
249 |
|
250 |
3.6 Mon Jan 11 00:43:39 CET 2010 |
251 |
- (libeio) more fixes for the freebsd/apple sendfile - broken by |
252 |
manpage standards, but apparently correct accoridng to actual |
253 |
kernel sources. |
254 |
- add IO::AIO::mmap/munmap support. |
255 |
- add IO::AIO::m(un)lockall support. |
256 |
- clean up manpage. |
257 |
|
258 |
3.5 Thu Jan 7 21:25:04 CET 2010 |
259 |
- (liebio) fix freebsd sendfile (Vanilla Hsu). |
260 |
- (libeio) also fix darwin sendfile that suffered from a similar bug. |
261 |
- add aio_statvfs as interface to statvfs/fstatvfs. |
262 |
- work around buggy symlink() on cygwin in the testsuite. |
263 |
- wtf. freebsd-8.0 actually passes the testsuite! I knew |
264 |
they could implement fork and semaphores one day! cheers! |
265 |
|
266 |
3.4 Sat Jan 2 15:13:04 CET 2010 |
267 |
- (libeio) max_poll_time was not properly converted to ticks. |
268 |
- clarify cancel_subs description. |
269 |
- IO::AIO::sendfile did not actually return the return value |
270 |
from the sendfile call. |
271 |
- implement aio_msync, aio_mtouch. |
272 |
- (libeio) tentatively support darwin in sendfile. |
273 |
|
274 |
3.31 Thu Nov 12 02:14:29 CET 2009 |
275 |
- fix result status documentation of aio_copy, aio_move. |
276 |
- speed up object creation considerably by avoiding |
277 |
hash lookups. |
278 |
|
279 |
3.3 Wed Aug 5 13:52:58 CEST 2009 |
280 |
- use common::sense. |
281 |
- use common schmorp.h header. |
282 |
- allow integers (file descriptors) in addition to file handles. |
283 |
- take advantage of linux' eventfd (but it seems slower on smp?). |
284 |
- use poll not select on posix platforms. |
285 |
|
286 |
3.261 Wed Jul 1 10:11:51 CEST 2009 |
287 |
- more 0S X workarounds (patch by Tokuhiro Matsuno). |
288 |
|
289 |
3.26 Tue Jun 30 09:33:26 CEST 2009 |
290 |
- 0S X of course claims to be posix 2008 but lacks posix_fadvise. |
291 |
try to work around this horribly broken OS in a somewhat hackish |
292 |
way. might help other os's too. |
293 |
|
294 |
3.25 Sat Jun 27 05:18:26 CEST 2009 |
295 |
- added IO::AIO::fadvise and IO::AIO::sendfile. |
296 |
- (libeio) replaced quicksort+insertion sort by a tuned radix |
297 |
sort + insertion sort, resulting in comparable runtime (usually |
298 |
faster) to the old version, but without any ill side effects on |
299 |
degenerated (for quicksort) data. |
300 |
- (libeio) correctly sort dirs by inodes if we have perfect type |
301 |
knowledge and DIRS_FIRST is used. |
302 |
- (libeio) internally the wrong DT_* constants were used, but no |
303 |
known system has differing EIO_DT_*/DT_* values, so not an issue. |
304 |
- removed a (harmless) assert that was left over on the code |
305 |
but should not have been. |
306 |
- use more correct types (change IO length from IV to VAL64, |
307 |
which makes a difference on systems without 64 bit perls). |
308 |
|
309 |
3.23 Sat Jun 13 16:57:58 CEST 2009 |
310 |
- fix off-by-one bug in aio_readdir that was introduced in the |
311 |
rushed 3.22 release. |
312 |
|
313 |
3.22 Sat Jun 13 15:32:40 CEST 2009 |
314 |
- speed up readdirx sort algorithm slightly. |
315 |
- bin/treescan was missing from distro tarball. |
316 |
|
317 |
3.21 Fri Jun 12 18:45:53 CEST 2009 |
318 |
- new options --dirs and --files for treescan. |
319 |
- install bin/treescan by default. |
320 |
- (libeio) aio_readdir can now be cancelled while executing. |
321 |
- fix a printf format string for 64 bit systems (could lead |
322 |
to problems on big endian 64 bit systems). |
323 |
- do not use qsort() but our own algorithm: glibc initialises |
324 |
the whole locale and I/O subsystem inside qsort, causing |
325 |
a stack overflow on 32 bit machines. The new sort uses much less |
326 |
stack and is more than twice as fast in typical situations. |
327 |
|
328 |
3.2 Sun Jun 7 20:30:05 CEST 2009 |
329 |
- (libeio) pwrite emulation was even more flawed than intended and did |
330 |
not restore the file offset. |
331 |
- add aio_readdirx, which can return inode and filetype and sort |
332 |
the names in various ways. |
333 |
- unfortunately, utime, chmod, chown on an open file that has just |
334 |
been written can easily block, which caused aio_copy to block the |
335 |
process. no more! |
336 |
- no longer rely on dst path in aio_copy when futime is available. |
337 |
|
338 |
3.19 Tue Apr 21 22:05:21 CEST 2009 |
339 |
- more perl 5.10 workarounds for aio_read and write. |
340 |
- aio_write no longer modifies the sv (if possible). |
341 |
- aio_read now works correctly with magic values. |
342 |
|
343 |
3.18 Sun Apr 19 21:17:32 CEST 2009 |
344 |
- better diagnostics when some aio_* functions get passed an illegal fh. |
345 |
- try to avoid crashes due to incompatible 5.10 API changes. grmbl. |
346 |
|
347 |
3.17 Thu Nov 20 08:45:36 CET 2008 |
348 |
- (libeio) added aio_sync_file_range (untested). |
349 |
- add aio_busy to @IO::AIO::AIO_REQ. |
350 |
|
351 |
3.16 Wed Oct 22 18:28:01 CEST 2008 |
352 |
- use SvREFCNT_inc instead of SvREFCNT_inc_NN in a non-speed critical |
353 |
part to improve portability to perl 5.8 (reported by szymon). |
354 |
|
355 |
3.15 Mon Oct 13 00:39:55 CEST 2008 |
356 |
- automatic removal of feeders was broken. |
357 |
- (libeio) use a more robust method to detect whether a feeder |
358 |
has added something to the group or not. |
359 |
|
360 |
3.1 Thu Oct 2 13:34:40 CEST 2008 |
361 |
- pre-deref the passed callback object, for increased speed |
362 |
and decreased memory usage. |
363 |
- call on_next_submit callback even for aio_group - impact unknown, |
364 |
but seems more correct. |
365 |
- $req->cb now returns the original callback. |
366 |
- (libeio) pass EIO_GROUP/aio_group requests through the queue without |
367 |
requiring a thread switch. |
368 |
- (libeio) status code of aio_futime and aio_utime was always passed 0. |
369 |
- do some other µ-optimisations. |
370 |
|
371 |
3.07 Sat Aug 2 16:06:13 CEST 2008 |
372 |
- do not include LIBS in autoconf tests. |
373 |
|
374 |
3.06 Tue Jul 15 12:41:32 CEST 2008 |
375 |
- move preadwritelock definition before it's first reference, |
376 |
to make it compile again on broken platforms (they still exist) |
377 |
or badly cofngiured perls (redhat...), reported by Rob Bloodgood. |
378 |
|
379 |
3.05 Thu Jun 19 23:23:52 CEST 2008 |
380 |
- work around perl overriding readdir etc. with thread-unsafe |
381 |
(sic!) versions. |
382 |
|
383 |
3.04 Wed Jun 18 01:35:38 CEST 2008 |
384 |
- (libeio) fix eio_mknod, which confused it's arguments. |
385 |
- (libeio) do not use readdir_r, as Ulrich Drepper pointed out that this |
386 |
is stupid. |
387 |
- (libeio) fix eio__readahead prototype, patch by Jost Krieger. |
388 |
- (libeio) fix a bug that could potentially cause IO::AIO |
389 |
not to get initialised properly. |
390 |
|
391 |
3.03 Thu May 29 05:33:30 CEST 2008 |
392 |
- (libeio) correctly call pthread_attr_destroy (). |
393 |
- (libeio) work around broken bsd headers once more. |
394 |
- reduce shared library size again by not including wrappers. |
395 |
- max_outstanding could cause poll to enter a busy-waiting loop. |
396 |
- document the new IO::AIO::poll_cb result value. |
397 |
|
398 |
3.02 Mon May 12 02:32:02 CEST 2008 |
399 |
- fix a memory leak on aio_readlink. |
400 |
- bring back working fchmod. |
401 |
- nop and busy now set result to 0. |
402 |
- set errno to ENOMEM when allocation fails. |
403 |
|
404 |
3.01 Sun May 11 03:07:03 CEST 2008 |
405 |
- (libeio) make it compile on systems without readahead or readdir_r. |
406 |
- (libeio) improve configure check for readahead availability. |
407 |
- do not try to link against -lrt. |
408 |
- use a separate configure script for IO::AIO (not the libeio one). |
409 |
|
410 |
3.0 Sun May 11 00:57:14 CEST 2008 |
411 |
- added bin/treescan to the distribution. |
412 |
- switched to using libeio. |
413 |
- LOTS OF INCOMPATIBLE CHANGES: |
414 |
- remove signal functionality, it is not worth the effort. |
415 |
- max_outstanding no longer returns the previous number of requests. |
416 |
- poll_cb no longer returns number of requests. |
417 |
|
418 |
2.62 Sat Apr 26 13:59:33 CEST 2008 |
419 |
- port to solaris perls configured for old posix |
420 |
(analysed by Jost Krieger). |
421 |
- keep a reference to the perl filehandle object in aio_close, |
422 |
so it doesn't get closed prematurely, leading to ugly races. |
423 |
|
424 |
2.61 Wed Apr 16 18:45:02 CEST 2008 |
425 |
- fix treescan output duplication and improve output ordering. |
426 |
also display files given on the commandline. |
427 |
- use a different algorithm for aio_close that is faster |
428 |
and probably has even lower chances of blocking. |
429 |
- do our own stack memory management for threads - linux |
430 |
allocates outrageous amounts of VM (not so bad unless you |
431 |
use mlockall...), which severely limits the number of threads |
432 |
on 32-bit arches: stack size is the larger of PTHREAD_STACK_MIN |
433 |
and 4096 * sizeof (long) (usually this is 16..64k). |
434 |
|
435 |
2.6 Sun Mar 30 08:28:11 CEST 2008 |
436 |
- added aio_sync. |
437 |
- added aio_pathsync. |
438 |
- fix prototypes of (void) functions. |
439 |
|
440 |
2.51 Sat Oct 6 16:04:54 CEST 2007 |
441 |
- perlio isn't generally threadsafe, so aio_close could not work reliably. |
442 |
aio_close now tries a safe hack that might not be that asynchronous |
443 |
(see the manpage for details). |
444 |
- discard callback return values as to not grow the stack endlessly |
445 |
when poll_cb handles many requests without returning. |
446 |
- minor code cleanups. |
447 |
|
448 |
2.5 Thu Oct 4 14:49:08 CEST 2007 |
449 |
- replaced _fd2fh with faster xs code. |
450 |
- aio_close will now try to do "the right thing" and thus might |
451 |
work sensibly for the very first time. |
452 |
|
453 |
2.41 Mon Sep 24 21:28:21 CEST 2007 |
454 |
- after fork, invest some work to make sure that the poll_fileno |
455 |
stays stable (by dup2'ing the new fd over the old one), to make |
456 |
it easier for programs/libs that don't control their forking |
457 |
behaviour and cannot recreate their watchers. |
458 |
|
459 |
2.4 Sun Aug 5 18:44:22 CEST 2007 |
460 |
- add aio_truncate, aio_chmod, aio_chown, aio_utime (linux |
461 |
successfully demonstrated that you can block on futimes...). |
462 |
- allow undef as fileoffset for aio_read/write and use read/write(2) |
463 |
internally (useful for sockets or O_APPEND handles). |
464 |
- allow undef for length in aio_write. |
465 |
- negative dataoffsets work as expected now in aio_read/aio_write. |
466 |
- use NV instead of UV for 32 bit perls and file offsets, as NVs |
467 |
have a larger range then. |
468 |
- shared code between BDB and IO::AIO. |
469 |
- aio_busy was completely broken. now fixed. |
470 |
- readahead emulation now returns something. |
471 |
- correctly set errno to ENOSYS on unimplemented functions |
472 |
(should never happen, though). |
473 |
- large changes to make it partially compile and run on win32, |
474 |
but win32 lacks too much functionality, and perl overrides way |
475 |
too many functions with crashing versions. |
476 |
|
477 |
2.33 Tue Jan 23 23:55:41 CET 2007 |
478 |
- fix install path (Andreas J. Koenig). |
479 |
|
480 |
2.32 Mon Jan 22 16:56:23 CET 2007 |
481 |
- added aio_rmtree. |
482 |
- wow, aio_mkdir was missing. |
483 |
- aio_load did return undef on error, not -1. |
484 |
- use prefixexp not prefix in autoconf.pm |
485 |
(suggested by ... rt.cpan.org unreachable). |
486 |
- avoid installing autoconf.pm. |
487 |
|
488 |
2.31 Sat Jan 6 03:46:02 CET 2007 |
489 |
- added aio_load. |
490 |
|
491 |
2.3 Sat Dec 23 05:48:07 CET 2006 |
492 |
- fix off-by-one bug in aio_read, causing memory corruption |
493 |
or crashes. |
494 |
|
495 |
2.21 Fri Dec 22 05:03:38 CET 2006 |
496 |
- minor doc updates. |
497 |
- minor cleanups. |
498 |
- IO::AIO::poll returns request count. |
499 |
- undocumented and unfinished async-signal-reporting. |
500 |
|
501 |
2.2 Tue Oct 31 00:31:00 CET 2006 |
502 |
- minor doc updates. |
503 |
- added aio_readlink. |
504 |
- properly zero-terminate aio_read buffer. |
505 |
- fix aio_read/write with nonzero dataoffset. |
506 |
- reduced size of aio request structure. |
507 |
- better diagnostics when pathnames etc. are not byte-encoded; |
508 |
do not leak request in that case, either. |
509 |
|
510 |
2.1 Sun Oct 29 02:01:57 CET 2006 |
511 |
- INCOMPATIBLE CHANGE: replace IO::AIO::poll_some by |
512 |
IO::AIO::max_poll_reqs and IO::AIO::max_poll_time. |
513 |
- fix a bug in aio_move where it would fail everytime |
514 |
it would have to copy files manually. |
515 |
- doc updates. |
516 |
- start threads only on (very low) demand. |
517 |
- end superfluous threads automatically after 10s idling. |
518 |
- add IO::AIO::max_idle to control this. |
519 |
- some tuning applied to random places. |
520 |
- add aio_mknod. |
521 |
|
522 |
2.0 Thu Oct 26 18:27:58 CEST 2006 |
523 |
- minor incompatibility: max_outstanding semantics have changed. |
524 |
- major rewrite, consider all functionality EXPERIMENTAL. |
525 |
- aio requests can be cancelled. |
526 |
- callbacks can be (re-)set later. |
527 |
- aio requests can choose between 9 priorities. |
528 |
- add aio_group: aio requests can be grouped |
529 |
into composite requests. |
530 |
- generator interface for groups. |
531 |
- add IO::AIO::poll_some. |
532 |
- major documentation improvements. |
533 |
- very minor bugfixes. |
534 |
- add aio_nop for dummy requests. |
535 |
- add aio_busy mainly for benchmarking and debugging. |
536 |
- use "fast" mutexes on linux. |
537 |
- use dynamic (and larger) buffers instead of |
538 |
putting them onto the stack. |
539 |
- optimise x86 and amd64 a bit. |
540 |
- better error checking when ENOMEM. |
541 |
- hopefully fix mem and dirp leaks on fork. |
542 |
- warn about broken perl malloc. |
543 |
- compiles on cygwin (but pread is broken on cygwin |
544 |
which is not detected by autoconf, so you have to |
545 |
edit autoconf/config.h yourself to enable emulation). |
546 |
|
547 |
1.8 Sat Jun 24 22:20:44 CEST 2006 |
548 |
- add and/or document aio_rename, aio_link, aio_symlink and aio_move. |
549 |
|
550 |
1.73 Wed Mar 1 22:49:32 CET 2006 |
551 |
- codename "jost" |
552 |
- AIX/Solaris(?) aren't being helpful again, hardcode a value |
553 |
for NAME_MAX and drive with your eyes closed (reported by |
554 |
wcooley@nakedape.cc). |
555 |
|
556 |
1.72 Thu Feb 2 00:40:44 CET 2006 |
557 |
- another minor bug in aio_scandir, again, the callback was called |
558 |
twice when the directory can't be stat'ed. |
559 |
|
560 |
1.71 Mon Dec 26 20:20:03 CET 2005 |
561 |
- grr, another minor bug in aio_scandir, this time resulting |
562 |
in a runtime error. |
563 |
|
564 |
1.7 Mon Dec 26 19:29:48 CET 2005 |
565 |
- aio_scandir errornously called callback twice in some cases. |
566 |
- added AnyEvent example. |
567 |
|
568 |
1.61 Wed Sep 7 19:40:42 CEST 2005 |
569 |
- no longer include $Config{libs}, as distros include too much |
570 |
garbage in there that is not installed on typical systems |
571 |
(for example, debian links perl against -lgdbm and many more |
572 |
libraries despite perl not needing it, slowing down startup and |
573 |
wasting memory). |
574 |
|
575 |
1.6 Tue Aug 30 17:44:44 CEST 2005 |
576 |
- added aio_readdir, aio_scandir. |
577 |
- added aio_sendfile, including sendfile emulation when sendfile |
578 |
isn't available. Linux, FreeBSD and HP-UX might work (sendfilev |
579 |
support is unavailable on my solaris 8 machine, but preliminary |
580 |
code is there. What's lacking is configury stuff and testing...). |
581 |
|
582 |
1.5 Sat Aug 20 02:32:19 CEST 2005 |
583 |
- create the aio threads on demand only (good after forking). |
584 |
- somebody nagged about C89 compatibility, so this release should |
585 |
be C89 compatible, but no guarantees for future ones. |
586 |
- abort when the configure script fails. |
587 |
|
588 |
1.4 Wed Aug 17 08:07:27 CEST 2005 |
589 |
- forgot to recreate the pipe after forking, this could cause |
590 |
deadlocks. |
591 |
|
592 |
1.3 Wed Aug 17 07:25:54 CEST 2005 |
593 |
- properly propagate exceptions from callbacks, instead of silently |
594 |
eating them. |
595 |
- use a different fork algorithm that avoids recursive callback |
596 |
invocations and reduces fork latency in busy processes. |
597 |
- fix a bug where the buffer scalar in aio_read got assigned the wrong |
598 |
length after reading. |
599 |
- fix a bug where the child would process some outstanding requests |
600 |
from the parent. |
601 |
- make a copy of the callback scalar, to avoid some very uncommon |
602 |
(but valid) usages where the callback scalar changes. |
603 |
- the callback was never freed in aio_read/aio_write. |
604 |
- aio_read/aio_write will now set the readonly flag on the data |
605 |
scalar for better error reporting. |
606 |
|
607 |
1.2 Wed Aug 17 01:22:58 CEST 2005 |
608 |
- verified on cygwin, linux/amd64+ia32+alpha, solaris. |
609 |
- use SvPVbyte for filenames now. You *need* to encode your strings to |
610 |
the external filename encoding first. |
611 |
- fix the bug where filehandles could leak (again). |
612 |
- clarify freebsd situation: make it work and tell me how, or go away |
613 |
with your underdocumented, broken and outdated OS. |
614 |
- fork'ing should now be officially supported (where linux and |
615 |
posix contradict each other, I followed posix). |
616 |
- only call read() once per poll, not twice (speedup). |
617 |
- new functions: aio_rmdir, aio_symlink. |
618 |
- force callback arguments to be undef or a reference, for better |
619 |
error catching. |
620 |
|
621 |
1.1 Sun Aug 7 05:27:23 CEST 2005 |
622 |
- added configure script to detect availability of some calls. |
623 |
(imperfect, as we cannot detect things like -lrt due to missing |
624 |
feedback yet). |
625 |
- emulate fdatasync via fsync, pread/pwrite via unsafe read/write |
626 |
and readahead via manual reading on platforms lacking those. |
627 |
- ported t/03_errors.t+framework by Brad Fitzpatrick |
628 |
from Linux::AIO. |
629 |
- removed XOPEN_SOURCE because stupid solaris 9 doesn't |
630 |
compile with it enabled. Probably breaks others. Oh well. |
631 |
- try to be more careful with errno on non-threaded perls. |
632 |
- cygwins perl insisted on a matching open mode when morphing |
633 |
the fd to an fh. |
634 |
- convert sv's to "byte" encoding in aio_read/aio_write. Might |
635 |
do that for pathnames, too, in the future, so make sure you |
636 |
properly encode your pathnames. |
637 |
|
638 |
1.0 Fri Jul 22 10:43:30 CEST 2005 |
639 |
- aio_open did not reliably deliver (open) filehandles. |
640 |
- all aio_*-functions keep a reference to the filehandle so |
641 |
it doesn't go away while the aio requests is working on it. |
642 |
- disable warnings in AIO.pm. |
643 |
|
644 |
0.9 Wed Jul 20 23:55:04 CEST 2005 |
645 |
- allow recursive calls to poll_cb, use less locking calls, too. |
646 |
- fix the Gtk2-example in the SYNOPSIS. |
647 |
- use pread/pwrite not pread64/pwrite64. this was simply a leftover |
648 |
from Linux::AIO. |
649 |
(found by compiling on cygwin. Unfortunately, cygwin declares but |
650 |
does not implement pread/pwrite). |
651 |
- only include <sys/syscall.h> on linux. |
652 |
|
653 |
0.5 Tue Jul 12 13:03:09 CEST 2005 |
654 |
- removed #include <endian.h>, it's not portable, and we don't |
655 |
need it anyway. |
656 |
- link against -lrt, as it might contain goodies such as fdatasync, |
657 |
and often aio_read, which we do not use yet, but it shouldn't hurt. |
658 |
|
659 |
0.4 Mon Jul 11 05:28:25 CEST 2005 |
660 |
- added IO::AIO::flush, IO::AIO::poll. |
661 |
- make callback argument optional for all calls. |
662 |
|
663 |
0.3 Mon Jul 11 01:42:37 CEST 2005 |
664 |
- restructured manpage. |
665 |
- added SYNOPSIS with examples for Event, Glib, Tk. |
666 |
- remove explicit context support so I don't have to maintain |
667 |
it for threaded perls (i.e. it should work now :). |
668 |
|
669 |
0.2 Sun Jul 10 22:57:12 CEST 2005 |
670 |
- return immediately from poll_wait if nreqs==0. |
671 |
- get rid of the request pipe and use faster thread-thread |
672 |
communications for most requests and responses. |
673 |
|
674 |
0.1 |
675 |
- cloned from Linux::AIO, which is very similar. |
676 |
|