| 1 | Revision history for IO::AIO |
1 | Revision history for IO::AIO |
| 2 | |
2 | |
| 3 | TODO: better autoconf.pm that can return LIBS etc. |
3 | TODO: scandir - some dirs mostly contain subdirs - invert logic? |
| 4 | TODO: aio_cptree/mvtree |
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/) |
5 | TODO: vmsplice? (http://kerneltrap.org/node/6505 http://lwn.net/Articles/178199/) |
| 7 | TODO: aio_fcntl, at least for file-locking |
6 | TODO: getxattr, setxattr, listxattr, removexattr etc.? |
| 8 | TODO: fallocate, if it wouldn't be so useless. |
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: name_to_handle_At + open_by_handle_at = clone fds |
|
|
20 | TODO: lchown |
|
|
21 | TODO: mount/umount2/MS_* flags |
|
|
22 | TODO: fsmount/fsconfig et al. https://lwn.net/Articles/759499/ |
|
|
23 | |
|
|
24 | TODO: rewrite rmtree et al. to support working directories (also speed them up) |
|
|
25 | TODO: maybe IO::AIO leaks fds when requests are cancelled? maybe initialise result to -1? |
|
|
26 | TODO: aio_wd should use O_PATH on linux, due to lacking O_SEARCH (http://comments.gmane.org/gmane.linux.file-systems/33611) |
|
|
27 | http://www.openwall.com/lists/musl/2013/02/23/4 |
|
|
28 | TODO: FIDEDUPERANGE, file_ioctl.c |
| 9 | TODO: aio_mincore? |
29 | TODO: mincore |
| 10 | TODO: add? also linux-specific stuff? |
30 | TODO: SOCK_CLOEXEc etc. |
| 11 | TODO: O_EXEC O_SEARCH O_CLOEXEC O_DIRECTORY O_DSYNC O_NOCTTY O_NOFOLLOW O_NONBLOCK O_RSYNC O_SYNC O_TTY_INIT O_DIRECT O_LARGEFILE O_NOATIME O_ASYNC |
31 | TODO: unlinkat flags, linkat flags, statat, renameat2 flags |
|
|
32 | TODO: https://kernelnewbies.org/Linux_4.11 |
|
|
33 | TODO: define useful rules for [[$wd, $path], $path] &c and implement them |
|
|
34 | TODO: https://github.com/coreutils/coreutils/blob/master/src/stat.c maybe IO::AIO::FSID |
|
|
35 | TODO: mlock2, MLOCK_ONFAULT |
|
|
36 | TODO: dup3, pipe2, accept4 |
|
|
37 | TODO: dump config.log when automated testign and configure fails |
|
|
38 | TODO: ns time resolution in utime |
|
|
39 | TODO: statx |
|
|
40 | TODO: pidfd_open/pidfd_send_signal (open/poll /proc?) |
|
|
41 | TODO: aio_fchdir/chdir? emulate chdirat? |
|
|
42 | TODO: rt-4.0.18-20109-1573826674-1229.126277-5-0@rt.cpan.org |
|
|
43 | TODO: aio_splice, aio_tee |
|
|
44 | TODO: copy_file_range |
|
|
45 | TODO: write more abpout the thread model, maybe bump some defaults? |
|
|
46 | TODO: io_uring, but it's so broken :/ |
|
|
47 | TODO: allow import of fds as aio_wd |
|
|
48 | TODO: fsopen/fspick/fsconfig, fsmount |
|
|
49 | TODO: close_range |
|
|
50 | TODO: move_mount, open_tree |
| 12 | |
51 | |
|
|
52 | - added pidfd_open, pidfd_getfd and pidfd_send_signal functions. |
|
|
53 | - rework bin/treescan to avoid recursion. This fixes two bugs |
|
|
54 | where a deep directory traversal or a lot of command line |
|
|
55 | arguments could cause it to crash. |
|
|
56 | - added (but not documented) open_tree, AT_* and move_mount flags. |
|
|
57 | - support defining syscall numbers in gendef0. |
|
|
58 | |
|
|
59 | 4.73 Fri Dec 4 02:08:37 CET 2020 |
|
|
60 | - def0.h was not uptodate, running into musl problems |
|
|
61 | (reported by B Fraser). |
|
|
62 | - IO::AIO::splice and IO::AIO::tee didn't properly return |
|
|
63 | 64 bit values. |
|
|
64 | - added IO::AIO::accept4. |
|
|
65 | - add various F_SEAL related fcntl constants. |
|
|
66 | - removed experimental marker for fdlimit functions. |
|
|
67 | - fiemap now includes the last segment even if it overflows |
|
|
68 | the end offset, which is arguably the correct behaviour. |
|
|
69 | |
|
|
70 | 4.72 Wed Apr 3 04:56:03 CEST 2019 |
|
|
71 | - (libeio) if fd 0 is available do not use it for aio_wd, as it |
|
|
72 | collides with IO::AIO::CWD. |
|
|
73 | - added IO::AIO::memfd_create. |
|
|
74 | - correctly include <sys/uio.h> in the vmsplice test. |
|
|
75 | - reduce codesize by ~7% on amd64 by declaring more functions |
|
|
76 | as noinline. |
|
|
77 | - documentation fixes and updates. |
|
|
78 | - experimental and undocumented preliminary support for synchronous statx. |
|
|
79 | |
|
|
80 | 4.71 Sun Mar 10 13:09:26 CET 2019 |
|
|
81 | - due to an error in the linux manpages, the configure tests |
|
|
82 | for readahead, sync_file_range, splice etc. failed. This has been fixed. |
|
|
83 | |
|
|
84 | 4.7 Mon Mar 4 11:28:13 CET 2019 |
|
|
85 | - majorly speed up scandir for the very special case of a non-POSIX |
|
|
86 | filesystem that nevertheless reports valid dt_type information. |
|
|
87 | The only known filesystem of this type is currently btrfs, which |
|
|
88 | didn't get it's act together to implement POSIX semantics in all |
|
|
89 | these years, and I am tired of waiting. |
|
|
90 | - add IO::AIO::MCL_ONFAULT for mlockall, add IO::AIO::mlockall. |
|
|
91 | - neither sys/mkdev.h nor sys/sysmacros.h were included, even when they |
|
|
92 | were detected by autoconf. |
|
|
93 | |
|
|
94 | 4.6 Sat Aug 25 21:58:36 CEST 2018 |
|
|
95 | - add st_btime, st_btimesec, st_btimensec and st_gen accessors, |
|
|
96 | mostly for bsds which expose birthtime and st_gen members (netbsd, freebsd |
|
|
97 | but only st_gen on openbsd because they compiled their perl without support |
|
|
98 | for birthtimes). |
|
|
99 | - fix madvise and munlock to properly support negative offsets. |
|
|
100 | - allocate fd for the aio_close at boot time, to guarantee it working |
|
|
101 | later, rather than calling abort when it fails. also avoids |
|
|
102 | close-on-exec race issues after module load. |
|
|
103 | - #undef utime on win32, maybe this helps some reports of utime hangs. |
|
|
104 | - minor doc improvements. |
|
|
105 | - minor configure cleanups. |
|
|
106 | - use $Config{perllibs} instead of libs for configure. might help people |
|
|
107 | who didn't install all perl dependencies (might break things as well). |
|
|
108 | |
|
|
109 | 4.54 Tue Aug 14 13:54:48 CEST 2018 |
|
|
110 | - include sys/mkdev.h or sys/sysmacros.h if available. |
|
|
111 | - further tweaks to configure invocation for systems requiring --rpath. |
|
|
112 | - no longer rely on custom paths on win32 platforms. |
|
|
113 | - try to work around buggy PAGESIZE macro on solaris. |
|
|
114 | |
|
|
115 | 4.53 Tue Aug 14 11:27:50 CEST 2018 |
|
|
116 | - add $Config{libs} to LIBS for configure, to work around systems |
|
|
117 | with broken library dependencies (... openbsd). |
|
|
118 | |
|
|
119 | 4.52 Sun Aug 12 08:09:45 CEST 2018 |
|
|
120 | - config.h.in was missing in distribution. |
|
|
121 | |
|
|
122 | 4.51 Sun Aug 12 07:24:14 CEST 2018 |
|
|
123 | - complete rework of the autoconf framework: IO::AIO now uses its own |
|
|
124 | config.h, separate from libeio, and tries to test the actual perl |
|
|
125 | environment, not the standard system environment. |
|
|
126 | - provide nanosecond-accracy stat time accessors for both perl and IO::AIO |
|
|
127 | stat functions. |
|
|
128 | - removed non-portable C++ syntax from eio.c. |
|
|
129 | - try to fix readdir tests on cygwin spuriously failing. |
|
|
130 | |
|
|
131 | 4.5 Wed Aug 1 00:23:55 CEST 2018 |
|
|
132 | - aio_mtouch touch all pages as requested, not just the first page in most cases. |
|
|
133 | - new function: IO::AIO::mremap, linux-specific mremap, with constants MREMAP_MAYMOVE |
|
|
134 | and MREMAP_FIXED. |
|
|
135 | - add O_ACCMODE. |
|
|
136 | - add (undocumented) MSG_CMSG_CLOEXEC and SOCK_CLOEXEC constants. |
|
|
137 | |
|
|
138 | 4.4 Tue Feb 20 07:54:12 CET 2018 |
|
|
139 | - new aio_slurp request. |
|
|
140 | - add (experimental) IO::AIO::eventfd and timerfd* function. |
|
|
141 | - add (experimental) IO::AIO::get_fdlimit and IO::AIO::min_fdlimit. |
|
|
142 | - point out that aio_copy overwrites existing files. |
|
|
143 | - removed filesystem magic number table from aio_statvfs docs as |
|
|
144 | statvfs doesn't actually return this info at all (statfs does). |
|
|
145 | - add a bunch of (mostly linux-specific) constants for use in ioctls |
|
|
146 | (see aio_ioctl docs). |
|
|
147 | - treescan now has a proper manpage and useful --help output. |
|
|
148 | - new option --sync in treescan, to sync everything in a subtree. |
|
|
149 | - changed default for aio_msync flags to MSYNC_SYNC. |
|
|
150 | - document offset/length behaviour of mprotect/madvise. |
|
|
151 | - support linux's renameat2 (via aio_rename2). |
|
|
152 | - add aio_rename2, an aio_rename with flags. |
|
|
153 | - add F_DUPPFD_CLOEXEC, F_OFD_[SG]ETLKW? constants. |
|
|
154 | - add FALLOC_FL_INSERT_RANGE, FALLOC_FL_UNSHARE_RANGE. |
|
|
155 | - (libeio) use posix_close if available. |
|
|
156 | - (libeio) internal close() calls no longer disturb errno. |
|
|
157 | - add IO:AIO::FALLOC_FL_UNSHARE. |
|
|
158 | - update schmorp.h for removal of USE_SOCKETS_AS_HANDLES |
|
|
159 | and better eventfd detection. |
|
|
160 | |
|
|
161 | 4.34 Sun May 1 19:18:24 CEST 2016 |
|
|
162 | - def0.h was not properly generated during previous release, causing |
|
|
163 | compile errors on various platforms. |
|
|
164 | - major/minor were accidentally switched (reported by Alexander Lishenyuk). |
|
|
165 | - removed duplicate definition of MAP_HUGETLB, that was found due to |
|
|
166 | a bug in Perl::Tidy (testcase by alaska332@gmail.com). |
|
|
167 | - added (untested!) aio_fcntl, aio_ioctl requests. |
|
|
168 | - (libeio) names set via prctl are truncated to 15 chars + nul, not 16, |
|
|
169 | as manpages-dev originally claimed. |
|
|
170 | |
|
|
171 | 4.33 Mon Jan 18 12:50:10 CET 2016 |
|
|
172 | - add IO::AIO::pipe2 function. |
|
|
173 | - added support for FALLOC_FL_COLLAPSE_RANGE and FALLOC_FL_ZERO_RANGE |
|
|
174 | constants. |
|
|
175 | - added support for O_TMPFILE and O_PATH constants. |
|
|
176 | - added support for MAP_FIXED, MAP_GROWSDOWN,MAP_32BIT, MAP_HUGETLB, MAP_STACK |
|
|
177 | consdtants, whether they can be sensibly used or not. |
|
|
178 | - use NO_INIT where applicable. |
|
|
179 | - update libecb. |
|
|
180 | - added stability canary support. |
|
|
181 | - updated linux super magic table to 4.3.3. |
|
|
182 | |
|
|
183 | 4.32 Wed Feb 11 20:32:11 CET 2015 |
|
|
184 | - replace off_t by STRLEN where appropriate, should not result in |
|
|
185 | user-visible changes. |
|
|
186 | - update ecb.h for C11 compatibility. |
|
|
187 | |
|
|
188 | 4.31 Tue Jun 3 03:29:27 CEST 2014 |
|
|
189 | - work around more 5.20 bugs. backwards compatibility my ass. |
|
|
190 | |
|
|
191 | 4.3 Fri Apr 11 06:22:38 CEST 2014 |
|
|
192 | - perl5porters broke Async::Interrupt, BDB, EV, IO::AIO, OpenCL |
|
|
193 | without warning by switching the meaning of USE_SOCKETS_AS_HANDLES |
|
|
194 | in 5.18. What's so attractive about giving a shit about backwards |
|
|
195 | compatibility - I will never understand. |
|
|
196 | |
|
|
197 | 4.2 Sat Jan 25 01:13:14 CET 2014 |
|
|
198 | - aio_group could corrupt memory because it didn't restore |
|
|
199 | the stack after req_submit. |
|
|
200 | - be more careful on (e.g. permission) errors in bin/treescan. |
|
|
201 | - work around changes in ExtUtils::MakeMaker. |
|
|
202 | - (libeio) implement aio_realpath for win32. |
|
|
203 | - (xthread) work around compile time bugs in ptw32. |
|
|
204 | - added IO::AIO::pipesize. |
|
|
205 | - (libecb) insignificant update. |
|
|
206 | |
|
|
207 | 4.19 Sun Jan 6 12:47:26 CET 2013 |
|
|
208 | - avoid endless loop in fiemap with some XFS files. |
|
|
209 | - in aio_rename and aio_rmdir, specialcase the case of [$wd, "."] |
|
|
210 | and call rename/rmdir instead of renameat/unlinkat. |
|
|
211 | |
|
|
212 | 4.18 Thu Oct 11 07:01:26 CEST 2012 |
|
|
213 | - fix unintended xthread_create by intentionalising it :) |
|
|
214 | |
|
|
215 | 4.17 Thu Oct 11 05:19:47 CEST 2012 |
|
|
216 | - rename aio_fallocate to aio_allocate, to match documentation. |
|
|
217 | - add list of linux fsid values to aio_statvfs docs. |
|
|
218 | - work around a bug in btrfs' FIEMAP ioctl implementation. |
|
|
219 | - work around AIX bug: statvfs.f_fsid is a struct, not unsigned long |
|
|
220 | as per unix spec. |
|
|
221 | |
|
|
222 | 4.16 Tue Aug 14 05:39:03 CEST 2012 |
|
|
223 | - aio_statvfs was wromgly marked for wd emulation, causing |
|
|
224 | it to malfunction for paths. |
|
|
225 | - fix a crash in aio_fiemap, when used on ranges without any |
|
|
226 | extents. |
|
|
227 | - work around linux kernel bug (at least in 3.2): kernel might |
|
|
228 | trash fiemap input arguments. |
|
|
229 | - work around linux kernel bug (at least in 3.2): kernel does not |
|
|
230 | set FIEMAP_EXTENT_LAST on last segment. |
|
|
231 | - work around linux kernel bug (at least in 3.2): kernel silently |
|
|
232 | truncates segment count if a file. |
|
|
233 | - make fh argument of IO::AIO::mmap optional, also add mmap/munmap |
|
|
234 | to quick overview. |
|
|
235 | - splice can corrupt data. |
|
|
236 | - (libeio) remove pread/pwrite emulation. |
|
|
237 | - do not grow mmaped scalars in aio_read. |
|
|
238 | - add FALLOC_FL_PUNCH_HOLE and document fallocate. |
|
|
239 | - provide pread/pwrite for win32, making it link again on native win32. |
|
|
240 | |
|
|
241 | 4.15 Tue Apr 10 06:59:00 CEST 2012 |
|
|
242 | - always include linux/types.h for fiemap, for compatibility to |
|
|
243 | ancient systems (Paul Howarth). |
|
|
244 | - experimental support for IO::AIO::splice and ::tee (no aio_...). |
|
|
245 | - provide SEEK_HOLE and SEEK_DATA, if available. |
|
|
246 | - work around (again!) an immensely stupid bug in RHEL, defining |
|
|
247 | autoconf macros in linux system headers (analysed by Paul Howarth). |
|
|
248 | |
|
|
249 | 4.14 Sat Apr 7 02:45:18 CEST 2012 |
|
|
250 | - fix stat structure usage on windows, which caused bogus stat results. |
|
|
251 | - (libeio) make readahead emulation behave more like actual readahead by never failing. |
|
|
252 | - new request aio_seek. |
|
|
253 | - new request aio_fiemap. |
|
|
254 | - autogenerate the #ifdef/#define 0 blocks for symbols we export. |
|
|
255 | |
|
|
256 | 4.12 Fri Dec 30 08:51:25 CET 2011 |
|
|
257 | - realpath would return a random value if the name was too long, |
|
|
258 | instead of -1. |
|
|
259 | - port to c89. |
|
|
260 | |
|
|
261 | 4.11 Mon Oct 10 00:24:11 CEST 2011 |
|
|
262 | - libeio didn't compile on !linux (Paul Howarth). |
|
|
263 | |
|
|
264 | 4.1 Sun Oct 9 10:24:11 CEST 2011 |
|
|
265 | - IO::AIO did access uninitialised memory on unsuccessful stats. |
|
|
266 | - (libeio) added syncfs syscall wrapper. |
|
|
267 | - (libeio) set thread name on linux (ps -L/Hcx, top, gdb). |
|
|
268 | - (libeio) support multiple indepenent working directories. |
|
|
269 | - applied speling corrections by Nicholas Bamber. |
|
|
270 | - tune treescan #threads and #outstanding requests a bit. |
|
|
271 | - reduce the number of file descriptors in use at any one time |
|
|
272 | in treescan, to avoid running out of them (and using too much ram). |
|
|
273 | - take advantage of working directoriy abstractions |
|
|
274 | in aio_scandir and treescan. |
|
|
275 | - reduce compiled size by ~10% by not inlining some key functionality. |
|
|
276 | - added --progress switch to treescan. |
|
|
277 | |
|
|
278 | 4.0 Mon Jul 18 05:01:10 CEST 2011 |
|
|
279 | - INCOMPATIBLE CHANGE: fork is no longer supported (indeed, it never was), |
|
|
280 | see FORK BEHAVIOUR in manpage for details. |
|
|
281 | - passes testsuite on win32 now and may actually work |
|
|
282 | (activestate 5.10.1 / MSVC6, YMMV). |
|
|
283 | - (libeio) fix a deadlock where a wakeup signal could be missed when |
|
|
284 | a timeout occured at the same time. |
|
|
285 | - (libeio) added realpath. |
|
|
286 | - (libeio) added fallocate. |
|
|
287 | - (libeio) disabling sendfile on darwin unfortunately broke the emulation. |
|
|
288 | - (libeio) do not acquire any locks when forking. |
|
|
289 | - (libeio) use fewer time() syscalls when waiting for new requests. |
|
|
290 | |
|
|
291 | 3.93 Wed Jun 29 23:44:18 CEST 2011 |
|
|
292 | - ECB.H WAS MISSING. |
|
|
293 | |
|
|
294 | 3.92 Wed Jun 29 14:45:41 CEST 2011 |
|
|
295 | - ecb.h was missing. |
|
|
296 | |
|
|
297 | 3.91 Wed Jun 29 13:24:42 CEST 2011 |
|
|
298 | - (libeio) work around a Linux (and likely FreeBSD and other |
|
|
299 | kernels) bug where sendfile would not transfer all the requested |
|
|
300 | bytes on large transfers, using a heuristic. |
|
|
301 | - buggy sendfile caused aio_move/copy to sometimes fail for big |
|
|
302 | files (fortunately it checks that the whole file has been |
|
|
303 | transferred...) |
|
|
304 | - use libecb for higher performance and higher portability. |
|
|
305 | - (libeio) disable sendfile on darwin, it's too broken. |
|
|
306 | - disable fork tests on !linux, as only linux supports mixing |
|
|
307 | pthread and fork in perl. |
|
|
308 | - document the fact that fork doesn't work anymore when using |
|
|
309 | this module, on anything !GNU/Linux. |
|
|
310 | - increase timeout in t/04_fork.t, as too many CPAN-tester setups |
|
|
311 | run on a 0.1MHz cpu. Or so. |
|
|
312 | |
|
|
313 | 3.9 Fri May 27 02:43:47 CEST 2011 |
|
|
314 | - (libeio) fix memory corruption in aio_readdirx for the flags |
|
|
315 | combination READDIR_STAT_ORDER | READDIR_DIRS_FIRST. |
|
|
316 | - add lots of newer POSIX and GNU/Linux-specific open |
|
|
317 | flags. |
|
|
318 | |
|
|
319 | 3.8 Sun Mar 27 12:25:33 CEST 2011 |
|
|
320 | - use nonstandard but maybe-working-on-bsd fork technique. |
|
|
321 | - support a max_idle value of 0. |
|
|
322 | - support setting of idle timeout value (IO::AIO::idle_timeout). |
|
|
323 | |
|
|
324 | 3.72 Fri Feb 11 04:25:38 CET 2011 |
| 13 | - use _POSIX_MEMLOCK_RANGE to detetc mlock/munlock. |
325 | - use _POSIX_MEMLOCK_RANGE to detect mlock/munlock. |
|
|
326 | - aio_mknod always used a dev_t value of 0. |
|
|
327 | - new treescan option: --grep. |
|
|
328 | - add more S_IF macros, and major/minor/makedev "macros". |
| 14 | |
329 | |
| 15 | 3.71 Thu Dec 30 08:18:46 CET 2010 |
330 | 3.71 Thu Dec 30 08:18:46 CET 2010 |
| 16 | - the numerical result value passed to callbacks did not stringify |
331 | - the numerical result value passed to callbacks did not stringify |
| 17 | correctly, due to internal reuse and failure to reset the sv flags. |
332 | correctly, due to internal reuse and failure to reset the sv flags. |
| 18 | - actually test for posix_[mf]advise, as at least uClibc defines |
333 | - actually test for posix_[mf]advise, as at least uClibc defines |
