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 IMPORTANT ------------------------------------------------------------
81 If the platform doesn't support it, then don't even try.
83 If running as non-root, then don't fail, just give a warning.
84 (There was a thread about this a while ago?)
86 http://lists.samba.org/pipermail/rsync/2001-August/thread.html
87 http://lists.samba.org/pipermail/rsync/2001-September/thread.html
91 Avoids traversal. Better option than a pile of --include statements
92 for people who want to generate the file list using a find(1)
95 File list structure in memory
97 Rather than one big array, perhaps have a tree in memory mirroring
100 This might make sorting much faster! (I'm not sure it's a big CPU
103 It might also reduce memory use in storing repeated directory names
104 -- again I'm not sure this is a problem.
108 Traverse just one directory at a time. Tridge says it's possible.
110 At the moment rsync reads the whole file list into memory at the
111 start, which makes us use a lot of memory and also not pipeline
112 network access as much as we could.
115 Handling duplicate names
117 We need to be careful of duplicate names getting into the file list.
118 See clean_flist(). This could happen if multiple arguments include
121 I think duplicates are only a problem if they're both flowing
122 through the pipeline at the same time. For example we might have
123 updated the first occurrence after reading the checksums for the
124 second. So possibly we just need to make sure that we don't have
125 both in the pipeline at the same time.
127 Possibly if we did one directory at a time that would be sufficient.
129 Alternatively we could pre-process the arguments to make sure no
130 duplicates will ever be inserted. There could be some bad cases
131 when we're collapsing symlinks.
133 We could have a hash table.
135 The root of the problem is that we do not want more than one file
136 list entry referring to the same file. At first glance there are
137 several ways this could happen: symlinks, hardlinks, and repeated
138 names on the command line.
140 If names are repeated on the command line, they may be present in
141 different forms, perhaps by traversing directory paths in different
142 ways, traversing paths including symlinks. Also we need to allow
143 for expansion of globs by rsync.
145 At the moment, clean_flist() requires having the entire file list in
146 memory. Duplicate names are detected just by a string comparison.
148 We don't need to worry about hard links causing duplicates because
149 files are never updated in place. Similarly for symlinks.
151 I think even if we're using a different symlink mode we don't need
154 Unless we're really clever this will introduce a protocol
155 incompatibility, so we need to be able to accept the old format as
161 At exit, show how much memory was used for the file list, etc.
163 Also we do a wierd exponential-growth allocation in flist.c. I'm
164 not sure this makes sense with modern mallocs. At any rate it will
165 make us allocate a huge amount of memory for large file lists.
170 At the moment hardlink handling is very expensive, so it's off by
171 default. It does not need to be so.
173 Since most of the solutions are rather intertwined with the file
174 list it is probably better to fix that first, although fixing
175 hardlinks is possibly simpler.
177 We can rule out hardlinked directories since they will probably
178 screw us up in all kinds of ways. They simply should not be used.
180 At the moment rsync only cares about hardlinks to regular files. I
181 guess you could also use them for sockets, devices and other beasts,
182 but I have not seen them.
184 When trying to reproduce hard links, we only need to worry about
185 files that have more than one name (nlinks>1 && !S_ISDIR).
187 The basic point of this is to discover alternate names that refer to
188 the same file. All operations, including creating the file and
189 writing modifications to it need only to be done for the first name.
190 For all later names, we just create the link and then leave it
193 If hard links are to be preserved:
195 Before the generator/receiver fork, the list of files is received
196 from the sender (recv_file_list), and a table for detecting hard
199 The generator looks for hard links within the file list and does
200 not send checksums for them, though it does send other metadata.
202 The sender sends the device number and inode with file entries, so
203 that files are uniquely identified.
205 The receiver goes through and creates hard links (do_hard_links)
206 after all data has been written, but before directory permissions
209 At the moment device and inum are sent as 4-byte integers, which
210 will probably cause problems on large filesystems. On Linux the
211 kernel uses 64-bit ino_t's internally, and people will soon have
212 filesystems big enough to use them. We ought to follow NFS4 in
213 using 64-bit device and inode identification, perhaps with a
214 protocol version bump.
216 Once we've seen all the names for a particular file, we no longer
217 need to think about it and we can deallocate the memory.
219 We can also have the case where there are links to a file that are
220 not in the tree being transferred. There's nothing we can do about
221 that. Because we rename the destination into place after writing,
222 any hardlinks to the old file are always going to be orphaned. In
223 fact that is almost necessary because otherwise we'd get really
224 confused if we were generating checksums for one name of a file and
227 At the moment the code seems to make a whole second copy of the file
228 list, which seems unnecessary.
230 We should have a test case that exercises hard links. Since it
231 might be hard to compare ./tls output where the inodes change we
232 might need a little program to check whether several names refer to
237 Implement suggestions from http://www.kame.net/newsletter/19980604/
238 and ftp://ftp.iij.ad.jp/pub/RFC/rfc2553.txt
240 If a host has multiple addresses, then listen try to connect to all
241 in order until we get through. (getaddrinfo may return multiple
242 addresses.) This is kind of implemented already.
244 Possibly also when starting as a server we may need to listen on
245 multiple passive addresses. This might be a bit harder, because we
246 may need to select on all of them. Hm.
248 Define a syntax for IPv6 literal addresses. Since they include
249 colons, they tend to break most naming systems, including ours.
250 Based on the HTTP IPv6 syntax, I think we should use
252 rsync://[::1]/foo/bar
255 which should just take a small change to the parser code.
260 If we hang or get SIGINT, then explain where we were up to. Perhaps
261 have a static buffer that contains the current function name, or
262 some kind of description of what we were trying to do. This is a
263 little easier on people than needing to run strace/truss.
265 "The dungeon collapses! You are killed." Rather than "unexpected
266 eof" give a message that is more detailed if possible and also more
269 If we get an error writing to a socket, then we should perhaps
270 continue trying to read to see if an error message comes across
271 explaining why the socket is closed. I'm not sure if this would
272 work, but it would certainly make our messages more helpful.
274 What happens if a directory is missing -x attributes. Do we lose
275 our load? (Debian #28416) Probably fixed now, but a test case
281 Device major/minor numbers should be at least 32 bits each. See
282 http://lists.samba.org/pipermail/rsync/2001-November/005357.html
284 Transfer ACLs. Need to think of a standard representation.
285 Probably better not to even try to convert between NT and POSIX.
286 Possibly can share some code with Samba.
290 With the current common --include '*/' --exclude '*' pattern, people
291 can end up with many empty directories. We might avoid this by
292 lazily creating such directories.
297 Perhaps don't use our own zlib.
301 - will automatically be up to date with bugfixes in zlib
303 - can leave it out for small rsync on e.g. recovery disks
305 - can use a shared library
307 - avoids people breaking rsync by trying to do this themselves and
310 Should we ship zlib for systems that don't have it, or require
311 people to install it separately?
313 Apparently this will make us incompatible with versions of rsync
314 that use the patched version of rsync. Probably the simplest way to
315 do this is to just disable gzip (with a warning) when talking to old
321 Perhaps flush stdout after each filename, so that people trying to
322 monitor progress in a log file can do so more easily. See
323 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=48108
325 At the connections that just get a list of modules are not logged,
328 If a child of the rsync daemon dies with a signal, we should notice
329 that when we reap it and log a message.
331 Keep stderr and stdout properly separated (Debian #23626)
333 Use a separate function for reporting errors; prefix it with
334 "rsync:" or "rsync(remote)", or perhaps even "rsync(local
340 There are already some patches to do this.
344 Allow RSYNC_PROXY to be http://user:pass@proxy.foo:3128/, and do
345 HTTP Basic Proxy-Authentication.
347 Multiple schemes are possible, up to and including the insanity that
348 is NTLM, but Basic probably covers most cases.
352 Add --with-socks, and then perhaps a command-line option to put them
353 on or off. This might be more reliable than LD_PRELOAD hacks.
357 <Rasmus> mbp: hey, how about an rsync option that just gives you the
358 summary without the list of files? And perhaps gives more
359 information like the number of new files, number of changed,
361 <mbp> Rasmus: nice idea
362 <mbp> there is --stats
363 <mbp> but at the moment it's very tridge-oriented
364 <mbp> rather than user-friendly
365 <mbp> it would be nice to improve it
366 <mbp> that would also work well with --dryrun
370 Rather than storing the file list in memory, store it in a TDB.
372 This *might* make memory usage lower while building the file list.
374 Hashtable lookup will mean files are not transmitted in order,
377 This would neatly eliminate one of the major post-fork shared data
383 On 12 Mar 2002, Dave Dykstra <dwd@bell-labs.com> wrote:
384 > If we would add an option to do that functionality, I would vote for one
385 > that was more general which could mask off any set of permission bits and
386 > possibly add any set of bits. Perhaps a chmod-like syntax if it could be
387 > implemented simply.
389 I think that would be good too. For example, people uploading files
390 to a web server might like to say
392 rsync -avzP --chmod a+rX ./ sourcefrog.net:/home/www/sourcefrog/
394 Ideally the patch would implement as many of the gnu chmod semantics
395 as possible. I think the mode parser should be a separate function
396 that passes back something like (mask,set) description to the rest of
397 the program. For bonus points there would be a test case for the
405 Allow people to specify the diff command. (Might want to use wdiff,
408 Just diff the temporary file with the destination file, and delete
409 the tmp file rather than moving it into place.
411 Interaction with --partial.
413 Security interactions with daemon mode?
415 (Suggestion from david.e.sewell)
418 Incorrect timestamps (Debian #100295)
420 A bit hard to believe, but apparently it happens.
423 Check "refuse options works"
425 We need a test case for this...
427 Was this broken when we changed to popt?
432 Test whether this is actually faster than just using malloc(). If
433 it's not (anymore), throw it out.
437 PLATFORMS ------------------------------------------------------------
441 Don't detach, because this messes up --srvany.
443 http://sources.redhat.com/ml/cygwin/2001-08/msg00234.html
445 According to "Effective TCP/IP Programming" (??) close() on a socket
446 has incorrect behaviour on Windows -- it sends a RST packet to the
447 other side, which gives a "connection reset by peer" error. On that
448 platform we should probably do shutdown() instead. However, on Unix
449 we are correct to call close(), because shutdown() discards
452 DEVELOPMENT ----------------------------------------------------------
456 Build rsync with SPLINT to try to find security holes. Add
457 annotations as necessary. Keep track of the number of warnings
458 found initially, and see how many of them are real bugs, or real
459 security bugs. Knowing the percentage of likely hits would be
460 really interesting for other projects.
464 Something that just keeps running rsync continuously over a data set
465 likely to generate problems.
469 Run current rsync versions against significant past releases.
473 jra recommends Valgrind:
475 http://devel-home.kde.org/~sewardj/
477 TESTING --------------------------------------------------------------
481 Part of the regression suite should be making sure that we don't
482 break backwards compatibility: old clients vs new servers and so
483 on. Ideally we would test the cross product of versions.
485 It might be sufficient to test downloads from well-known public
486 rsync servers running different versions of rsync. This will give
487 some testing and also be the most common case for having different
488 versions and not being able to upgrade.
492 Sparse and non-sparse
496 Insert bytes, delete bytes, swap blocks, ...
498 configure option to enable dangerous tests
500 If tests are skipped, say why.
502 Test daemon feature to disallow particular options.
504 Pipe program that makes slow/jerky connections.
506 Versions of read() and write() that corrupt the stream, or abruptly fail
508 Separate makefile target to run rough tests -- or perhaps just run
512 DOCUMENTATION --------------------------------------------------------
516 Keep list of open issues and todos on the web site
518 Update web site from CVS
520 BUILD FARM -----------------------------------------------------------
524 AMDAHL UTS (Dave Dykstra)
526 Cygwin (on different versions of Win32?)
528 HP-UX variants (via HP?)
532 NICE -----------------------------------------------------------------
534 --no-detach and --no-fork options
536 Very useful for debugging. Also good when running under a
537 daemon-monitoring process that tries to restart the service when the
540 hang/timeout friendliness
544 Indicate whether files are new, updated, or deleted
546 At end of transfer, show how many files were or were not transferred
551 Change to using gettext(). Probably need to ship this for platforms
554 Solicit translations.
560 Write a small emulation of interactive ftp as a Pythonn program
561 that calls rsync. Commands such as "cd", "ls", "ls *.c" etc map
562 fairly directly into rsync commands: it just needs to remember the
563 current host, directory and so on. We can probably even do
564 completion of remote filenames.