3 BUGS ---------------------------------------------------------------
5 There seems to be a bug with hardlinks
7 mbp/2 build$ ls -l /tmp/a /tmp/b -i
10 2568307 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a1
11 2568307 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a2
12 2568307 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a3
13 2568310 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a4
14 2568310 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a5
15 2568310 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b1
16 2568310 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b2
17 2568310 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b3
21 2568309 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a1
22 2568309 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a2
23 2568309 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a3
24 2568311 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a4
25 2568311 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a5
26 2568311 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b1
27 2568311 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b2
28 2568311 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b3
29 mbp/2 build$ rm -r /tmp/b && ./rsync -avH /tmp/a/ /tmp/b
30 building file list ... done
31 created directory /tmp/b
37 wrote 350 bytes read 52 bytes 804.00 bytes/sec
38 total size is 232 speedup is 0.58
39 mbp/2 build$ rm -r /tmp/b
40 mbp/2 build$ ls -l /tmp/b
41 ls: /tmp/b: No such file or directory
42 mbp/2 build$ rm -r /tmp/b && ./rsync -avH /tmp/a/ /tmp/b
43 rm: cannot remove `/tmp/b': No such file or directory
44 mbp/2 build$ rm -f -r /tmp/b && ./rsync -avH /tmp/a/ /tmp/b
45 building file list ... done
46 created directory /tmp/b
52 wrote 350 bytes read 52 bytes 804.00 bytes/sec
53 total size is 232 speedup is 0.58
54 mbp/2 build$ ls -l /tmp/b
56 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a1
57 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a2
58 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a3
59 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a4
60 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a5
61 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b1
62 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b2
63 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b3
64 mbp/2 build$ ls -l /tmp/a
66 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a1
67 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a2
68 -rw-rw-r-- 3 mbp mbp 29 Mar 25 17:30 a3
69 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a4
70 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 a5
71 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b1
72 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b2
73 -rw-rw-r-- 5 mbp mbp 29 Mar 25 17:30 b3
76 Progress indicator can produce corrupt output when transferring directories:
79 main/binary-arm/admin/
81 main/binary-arm/comm/8.56kB/s 0:00:52
82 main/binary-arm/devel/
84 main/binary-arm/editors/
85 main/binary-arm/electronics/s 0:00:53
86 main/binary-arm/games/
87 main/binary-arm/graphics/
88 main/binary-arm/hamradio/
89 main/binary-arm/interpreters/
90 main/binary-arm/libs/6.61kB/s 0:00:54
97 I don't think we handle this properly on systems that don't have the
101 DAEMON --------------------------------------------------------------
103 server-imposed bandwidth limits
107 There are already some patches to do this.
109 BitKeeper uses a server whose login shell is set to bkd. That's
110 probably a reasonable approach.
113 FEATURES ------------------------------------------------------------
116 --dry-run is insufficiently dry
118 Mark Santcroos points out that -n fails to list files which have
119 only metadata changes, though it probably should.
121 There may be a Debian bug about this as well.
126 If the platform doesn't support it, then don't even try.
128 If running as non-root, then don't fail, just give a warning.
129 (There was a thread about this a while ago?)
131 http://lists.samba.org/pipermail/rsync/2001-August/thread.html
132 http://lists.samba.org/pipermail/rsync/2001-September/thread.html
137 Avoids traversal. Better option than a pile of --include statements
138 for people who want to generate the file list using a find(1)
144 Perhaps allow supplementary groups to be specified in rsyncd.conf;
145 then make the first one the primary gid and all the rest be
149 File list structure in memory
151 Rather than one big array, perhaps have a tree in memory mirroring
154 This might make sorting much faster! (I'm not sure it's a big CPU
157 It might also reduce memory use in storing repeated directory names
158 -- again I'm not sure this is a problem.
162 Traverse just one directory at a time. Tridge says it's possible.
164 At the moment rsync reads the whole file list into memory at the
165 start, which makes us use a lot of memory and also not pipeline
166 network access as much as we could.
169 Handling duplicate names
171 We need to be careful of duplicate names getting into the file list.
172 See clean_flist(). This could happen if multiple arguments include
175 I think duplicates are only a problem if they're both flowing
176 through the pipeline at the same time. For example we might have
177 updated the first occurrence after reading the checksums for the
178 second. So possibly we just need to make sure that we don't have
179 both in the pipeline at the same time.
181 Possibly if we did one directory at a time that would be sufficient.
183 Alternatively we could pre-process the arguments to make sure no
184 duplicates will ever be inserted. There could be some bad cases
185 when we're collapsing symlinks.
187 We could have a hash table.
189 The root of the problem is that we do not want more than one file
190 list entry referring to the same file. At first glance there are
191 several ways this could happen: symlinks, hardlinks, and repeated
192 names on the command line.
194 If names are repeated on the command line, they may be present in
195 different forms, perhaps by traversing directory paths in different
196 ways, traversing paths including symlinks. Also we need to allow
197 for expansion of globs by rsync.
199 At the moment, clean_flist() requires having the entire file list in
200 memory. Duplicate names are detected just by a string comparison.
202 We don't need to worry about hard links causing duplicates because
203 files are never updated in place. Similarly for symlinks.
205 I think even if we're using a different symlink mode we don't need
208 Unless we're really clever this will introduce a protocol
209 incompatibility, so we need to be able to accept the old format as
215 At exit, show how much memory was used for the file list, etc.
217 Also we do a wierd exponential-growth allocation in flist.c. I'm
218 not sure this makes sense with modern mallocs. At any rate it will
219 make us allocate a huge amount of memory for large file lists.
224 At the moment hardlink handling is very expensive, so it's off by
225 default. It does not need to be so.
227 Since most of the solutions are rather intertwined with the file
228 list it is probably better to fix that first, although fixing
229 hardlinks is possibly simpler.
231 We can rule out hardlinked directories since they will probably
232 screw us up in all kinds of ways. They simply should not be used.
234 At the moment rsync only cares about hardlinks to regular files. I
235 guess you could also use them for sockets, devices and other beasts,
236 but I have not seen them.
238 When trying to reproduce hard links, we only need to worry about
239 files that have more than one name (nlinks>1 && !S_ISDIR).
241 The basic point of this is to discover alternate names that refer to
242 the same file. All operations, including creating the file and
243 writing modifications to it need only to be done for the first name.
244 For all later names, we just create the link and then leave it
247 If hard links are to be preserved:
249 Before the generator/receiver fork, the list of files is received
250 from the sender (recv_file_list), and a table for detecting hard
253 The generator looks for hard links within the file list and does
254 not send checksums for them, though it does send other metadata.
256 The sender sends the device number and inode with file entries, so
257 that files are uniquely identified.
259 The receiver goes through and creates hard links (do_hard_links)
260 after all data has been written, but before directory permissions
263 At the moment device and inum are sent as 4-byte integers, which
264 will probably cause problems on large filesystems. On Linux the
265 kernel uses 64-bit ino_t's internally, and people will soon have
266 filesystems big enough to use them. We ought to follow NFS4 in
267 using 64-bit device and inode identification, perhaps with a
268 protocol version bump.
270 Once we've seen all the names for a particular file, we no longer
271 need to think about it and we can deallocate the memory.
273 We can also have the case where there are links to a file that are
274 not in the tree being transferred. There's nothing we can do about
275 that. Because we rename the destination into place after writing,
276 any hardlinks to the old file are always going to be orphaned. In
277 fact that is almost necessary because otherwise we'd get really
278 confused if we were generating checksums for one name of a file and
281 At the moment the code seems to make a whole second copy of the file
282 list, which seems unnecessary.
284 We should have a test case that exercises hard links. Since it
285 might be hard to compare ./tls output where the inodes change we
286 might need a little program to check whether several names refer to
291 Implement suggestions from http://www.kame.net/newsletter/19980604/
292 and ftp://ftp.iij.ad.jp/pub/RFC/rfc2553.txt
294 If a host has multiple addresses, then listen try to connect to all
295 in order until we get through. (getaddrinfo may return multiple
296 addresses.) This is kind of implemented already.
298 Possibly also when starting as a server we may need to listen on
299 multiple passive addresses. This might be a bit harder, because we
300 may need to select on all of them. Hm.
302 Define a syntax for IPv6 literal addresses. Since they include
303 colons, they tend to break most naming systems, including ours.
304 Based on the HTTP IPv6 syntax, I think we should use
306 rsync://[::1]/foo/bar
309 which should just take a small change to the parser code.
314 If we hang or get SIGINT, then explain where we were up to. Perhaps
315 have a static buffer that contains the current function name, or
316 some kind of description of what we were trying to do. This is a
317 little easier on people than needing to run strace/truss.
319 "The dungeon collapses! You are killed." Rather than "unexpected
320 eof" give a message that is more detailed if possible and also more
323 If we get an error writing to a socket, then we should perhaps
324 continue trying to read to see if an error message comes across
325 explaining why the socket is closed. I'm not sure if this would
326 work, but it would certainly make our messages more helpful.
328 What happens if a directory is missing -x attributes. Do we lose
329 our load? (Debian #28416) Probably fixed now, but a test case
335 Device major/minor numbers should be at least 32 bits each. See
336 http://lists.samba.org/pipermail/rsync/2001-November/005357.html
338 Transfer ACLs. Need to think of a standard representation.
339 Probably better not to even try to convert between NT and POSIX.
340 Possibly can share some code with Samba.
344 With the current common --include '*/' --exclude '*' pattern, people
345 can end up with many empty directories. We might avoid this by
346 lazily creating such directories.
351 Perhaps don't use our own zlib.
355 - will automatically be up to date with bugfixes in zlib
357 - can leave it out for small rsync on e.g. recovery disks
359 - can use a shared library
361 - avoids people breaking rsync by trying to do this themselves and
364 Should we ship zlib for systems that don't have it, or require
365 people to install it separately?
367 Apparently this will make us incompatible with versions of rsync
368 that use the patched version of rsync. Probably the simplest way to
369 do this is to just disable gzip (with a warning) when talking to old
372 After we get the @RSYNCD greeting from the server, we know it's
373 version but we have not yet sent the command line, so we could just
374 remove the -z option if the server is too old.
376 For ssh invocation it's not so simple, because we actually use the
377 command line to start the remote process. However, we only actually
378 do compression in token.c, and we could therefore once we discover
379 the remote version emit an error if it's too old. I'm not sure if
380 that's a good tradeoff or not.
385 Allow RSYNC_PROXY to be http://user:pass@proxy.foo:3128/, and do
386 HTTP Basic Proxy-Authentication.
388 Multiple schemes are possible, up to and including the insanity that
389 is NTLM, but Basic probably covers most cases.
393 Add --with-socks, and then perhaps a command-line option to put them
394 on or off. This might be more reliable than LD_PRELOAD hacks.
398 <Rasmus> mbp: hey, how about an rsync option that just gives you the
399 summary without the list of files? And perhaps gives more
400 information like the number of new files, number of changed,
402 <mbp> Rasmus: nice idea
403 <mbp> there is --stats
404 <mbp> but at the moment it's very tridge-oriented
405 <mbp> rather than user-friendly
406 <mbp> it would be nice to improve it
407 <mbp> that would also work well with --dryrun
411 Rather than storing the file list in memory, store it in a TDB.
413 This *might* make memory usage lower while building the file list.
415 Hashtable lookup will mean files are not transmitted in order,
418 This would neatly eliminate one of the major post-fork shared data
424 On 12 Mar 2002, Dave Dykstra <dwd@bell-labs.com> wrote:
425 > If we would add an option to do that functionality, I would vote for one
426 > that was more general which could mask off any set of permission bits and
427 > possibly add any set of bits. Perhaps a chmod-like syntax if it could be
428 > implemented simply.
430 I think that would be good too. For example, people uploading files
431 to a web server might like to say
433 rsync -avzP --chmod a+rX ./ sourcefrog.net:/home/www/sourcefrog/
435 Ideally the patch would implement as many of the gnu chmod semantics
436 as possible. I think the mode parser should be a separate function
437 that passes back something like (mask,set) description to the rest of
438 the program. For bonus points there would be a test case for the
446 Allow people to specify the diff command. (Might want to use wdiff,
449 Just diff the temporary file with the destination file, and delete
450 the tmp file rather than moving it into place.
452 Interaction with --partial.
454 Security interactions with daemon mode?
456 (Suggestion from david.e.sewell)
459 Incorrect timestamps (Debian #100295)
461 A bit hard to believe, but apparently it happens.
464 Check "refuse options works"
466 We need a test case for this...
468 Was this broken when we changed to popt?
471 PERFORMANCE ----------------------------------------------------------
475 If we're doing a local transfer, or using -W, then perhaps don't
476 send the file checksum. If we're doing a local transfer, then
477 calculating MD4 checksums uses 90% of CPU and is unlikely to be
480 Indeed for transfers over zlib or ssh we can also rely on the
481 transport to have quite strong protection against corruption.
483 Perhaps we should have an option to disable this, analogous to
484 --whole-file, although it would default to disabled. The file
485 checksum takes up a definite space in the protocol -- we can either
486 set it to 0, or perhaps just leave it out.
490 Perhaps borrow an assembler MD4 from someone?
492 Make sure we call MD4 with properly-sized blocks whenever possible
493 to avoid copying into the residue region?
497 Test whether this is actually faster than just using malloc(). If
498 it's not (anymore), throw it out.
502 PLATFORMS ------------------------------------------------------------
506 Don't detach, because this messes up --srvany.
508 http://sources.redhat.com/ml/cygwin/2001-08/msg00234.html
510 According to "Effective TCP/IP Programming" (??) close() on a socket
511 has incorrect behaviour on Windows -- it sends a RST packet to the
512 other side, which gives a "connection reset by peer" error. On that
513 platform we should probably do shutdown() instead. However, on Unix
514 we are correct to call close(), because shutdown() discards
517 DEVELOPMENT ----------------------------------------------------------
521 Build rsync with SPLINT to try to find security holes. Add
522 annotations as necessary. Keep track of the number of warnings
523 found initially, and see how many of them are real bugs, or real
524 security bugs. Knowing the percentage of likely hits would be
525 really interesting for other projects.
529 Something that just keeps running rsync continuously over a data set
530 likely to generate problems.
534 Run current rsync versions against significant past releases.
538 jra recommends Valgrind:
540 http://devel-home.kde.org/~sewardj/
546 Build tar file; upload
548 Send announcement to mailing list and c.o.l.a.
550 Make freshmeat announcement
556 TESTING --------------------------------------------------------------
560 Part of the regression suite should be making sure that we don't
561 break backwards compatibility: old clients vs new servers and so
562 on. Ideally we would test both up and down from the current release
565 We might need to omit broken old versions, or versions in which
566 particular functionality is broken
568 It might be sufficient to test downloads from well-known public
569 rsync servers running different versions of rsync. This will give
570 some testing and also be the most common case for having different
571 versions and not being able to upgrade.
574 Test on kernel source
576 Download all versions of kernel; unpack, sync between them. Also
577 sync between uncompressed tarballs. Compare directories after
580 Use local mode; ssh; daemon; --whole-file and --no-whole-file.
582 Use awk to pull out the 'speedup' number for each transfer. Make
588 Sparse and non-sparse
592 Insert bytes, delete bytes, swap blocks, ...
594 configure option to enable dangerous tests
596 If tests are skipped, say why.
598 Test daemon feature to disallow particular options.
600 Pipe program that makes slow/jerky connections.
602 Versions of read() and write() that corrupt the stream, or abruptly fail
604 Separate makefile target to run rough tests -- or perhaps just run
608 DOCUMENTATION --------------------------------------------------------
612 Keep list of open issues and todos on the web site
614 Update web site from CVS
617 Perhaps redo manual as SGML
619 The man page is getting rather large, and there is more information
620 that ought to be added.
622 TexInfo source is probably a dying format.
624 Linuxdoc looks like the most likely contender. I know DocBook is
625 favoured by some people, but it's so bloody verbose, even with emacs
629 BUILD FARM -----------------------------------------------------------
633 AMDAHL UTS (Dave Dykstra)
635 Cygwin (on different versions of Win32?)
637 HP-UX variants (via HP?)
642 LOGGING --------------------------------------------------------------
644 Perhaps flush stdout after each filename, so that people trying to
645 monitor progress in a log file can do so more easily. See
646 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=48108
648 At the connections that just get a list of modules are not logged,
651 If a child of the rsync daemon dies with a signal, we should notice
652 that when we reap it and log a message.
654 Keep stderr and stdout properly separated (Debian #23626)
656 Use a separate function for reporting errors; prefix it with
657 "rsync:" or "rsync(remote)", or perhaps even "rsync(local
662 Indicate whether files are new, updated, or deleted
664 At end of transfer, show how many files were or were not transferred
669 Explain *why* every file is transferred or not (e.g. "local mtime
670 123123 newer than 1283198")
675 Add an rsyncd.conf parameter to turn on debugging on the server.
679 NICE -----------------------------------------------------------------
681 --no-detach and --no-fork options
683 Very useful for debugging. Also good when running under a
684 daemon-monitoring process that tries to restart the service when the
687 hang/timeout friendliness
691 Change to using gettext(). Probably need to ship this for platforms
694 Solicit translations.
696 Does anyone care? Before we bother modifying the code, we ought to
697 get the manual translated first, because that's possibly more useful
698 and at any rate demonstrates desire.
703 Write a small emulation of interactive ftp as a Pythonn program
704 that calls rsync. Commands such as "cd", "ls", "ls *.c" etc map
705 fairly directly into rsync commands: it just needs to remember the
706 current host, directory and so on. We can probably even do
707 completion of remote filenames.
710 RELATED PROJECTS -----------------------------------------------------
712 http://rsync.samba.org/rsync-and-debian/
716 Exhaustive, tortuous testing
720 rsyncsplit as alternative to real integration with gzip?
722 reverse rsync over HTTP Range
724 Goswin Brederlow suggested this on Debian; I think tridge and I
725 talked about it previous in relation to rproxy.