3 BUGS ---------------------------------------------------------------
4 Fix progress indicator to not corrupt log
6 Do not rely on having a group called "nobody"
7 Incorrect timestamps (Debian #100295)
10 FEATURES ------------------------------------------------------------
11 server-imposed bandwidth limits
13 Use chroot only if supported
14 Allow supplementary groups in rsyncd.conf 2002/04/09
15 Handling IPv6 on old machines
17 Add ACL support 2001/12/02
18 Lazy directory creation
19 Conditional -z for old protocols
20 proxy authentication 2002/01/23
23 Allow forcing arbitrary permissions 2002/03/12
24 --diff david.e.sewell 2002/03/15
25 Add daemon --no-detach and --no-fork options
26 Create more granular verbosity jw 2003/05/15
28 DOCUMENTATION --------------------------------------------------------
30 Keep list of open issues and todos on the web site
31 Update web site from CVS
32 Perhaps redo manual as SGML
34 LOGGING --------------------------------------------------------------
35 Make dry run list all updates 2002/04/03
37 Improve error messages
38 Better statistics: Rasmus 2002/03/08
39 Perhaps flush stdout like syslog
40 Log deamon sessions that just list modules
41 Log child death on signal
42 Keep stderr and stdout properly separated (Debian #23626)
43 Log errors with function that reports process of origin
44 verbose output David Stein 2001/12/20
45 Add reason for transfer to file logging
46 debugging of daemon 2002/04/08
49 DEVELOPMENT --------------------------------------------------------
50 Handling duplicate names
51 Use generic zlib 2002/02/25
56 Add machines to build farm
58 PERFORMANCE ----------------------------------------------------------
59 File list structure in memory
60 Traverse just one directory at a time
62 Allow skipping MD4 file_sum 2002/04/08
66 TESTING --------------------------------------------------------------
68 Cross-test versions 2001/08/22
71 Create mutator program for testing
72 Create configure option to enable dangerous tests
73 If tests are skipped, say why.
74 Test daemon feature to disallow particular options.
75 Create pipe program for testing
76 Create test makefile target for some tests
77 Test "refuse options" works
79 RELATED PROJECTS -----------------------------------------------------
81 http://rsync.samba.org/rsync-and-debian/
83 rsyncsplit as alternative to real integration with gzip?
84 reverse rsync over HTTP Range
88 BUGS ---------------------------------------------------------------
90 Fix progress indicator to not corrupt log
92 Progress indicator can produce corrupt output when transferring directories:
95 main/binary-arm/admin/
97 main/binary-arm/comm/8.56kB/s 0:00:52
98 main/binary-arm/devel/
100 main/binary-arm/editors/
101 main/binary-arm/electronics/s 0:00:53
102 main/binary-arm/games/
103 main/binary-arm/graphics/
104 main/binary-arm/hamradio/
105 main/binary-arm/interpreters/
106 main/binary-arm/libs/6.61kB/s 0:00:54
107 main/binary-arm/mail/
108 main/binary-arm/math/
109 main/binary-arm/misc/
116 I don't think we handle this properly on systems that don't have the
117 call. Are there any such?
122 Do not rely on having a group called "nobody"
124 http://www.linuxbase.org/spec/refspecs/LSB_1.1.0/gLSB/usernames.html
126 On Debian it's "nogroup"
131 Incorrect timestamps (Debian #100295)
133 A bit hard to believe, but apparently it happens.
140 Don't detach, because this messes up --srvany.
142 http://sources.redhat.com/ml/cygwin/2001-08/msg00234.html
148 FEATURES ------------------------------------------------------------
150 server-imposed bandwidth limits
157 There are already some patches to do this.
159 BitKeeper uses a server whose login shell is set to bkd. That's
160 probably a reasonable approach.
165 Use chroot only if supported
167 If the platform doesn't support it, then don't even try.
169 If running as non-root, then don't fail, just give a warning.
170 (There was a thread about this a while ago?)
172 http://lists.samba.org/pipermail/rsync/2001-August/thread.html
173 http://lists.samba.org/pipermail/rsync/2001-September/thread.html
178 Allow supplementary groups in rsyncd.conf 2002/04/09
180 Perhaps allow supplementary groups to be specified in rsyncd.conf;
181 then make the first one the primary gid and all the rest be
187 Handling IPv6 on old machines
189 The KAME IPv6 patch is nice in theory but has proved a bit of a
190 nightmare in practice. The basic idea of their patch is that rsync
191 is rewritten to use the new getaddrinfo()/getnameinfo() interface,
192 rather than gethostbyname()/gethostbyaddr() as in rsync 2.4.6.
193 Systems that don't have the new interface are handled by providing
194 our own implementation in lib/, which is selectively linked in.
196 The problem with this is that it is really hard to get right on
197 platforms that have a half-working implementation, so redefining
198 these functions clashes with system headers, and leaving them out
199 breaks. This affects at least OSF/1, RedHat 5, and Cobalt, which
200 are moderately improtant.
202 Perhaps the simplest solution would be to have two different files
203 implementing the same interface, and choose either the new or the
204 old API. This is probably necessary for systems that e.g. have
205 IPv6, but gethostbyaddr() can't handle it. The Linux manpage claims
206 this is currently the case.
208 In fact, our internal sockets interface (things like
209 open_socket_out(), etc) is much narrower than the getaddrinfo()
210 interface, and so probably simpler to get right. In addition, the
211 old code is known to work well on old machines.
213 We could drop the rather large lib/getaddrinfo files.
220 Implement suggestions from http://www.kame.net/newsletter/19980604/
221 and ftp://ftp.iij.ad.jp/pub/RFC/rfc2553.txt
223 If a host has multiple addresses, then listen try to connect to all
224 in order until we get through. (getaddrinfo may return multiple
225 addresses.) This is kind of implemented already.
227 Possibly also when starting as a server we may need to listen on
228 multiple passive addresses. This might be a bit harder, because we
229 may need to select on all of them. Hm.
231 Define a syntax for IPv6 literal addresses. Since they include
232 colons, they tend to break most naming systems, including ours.
233 Based on the HTTP IPv6 syntax, I think we should use
235 rsync://[::1]/foo/bar [::1]::bar
237 which should just take a small change to the parser code.
242 Add ACL support 2001/12/02
244 Transfer ACLs. Need to think of a standard representation.
245 Probably better not to even try to convert between NT and POSIX.
246 Possibly can share some code with Samba.
251 Lazy directory creation
253 With the current common --include '*/' --exclude '*' pattern, people
254 can end up with many empty directories. We might avoid this by
255 lazily creating such directories.
260 Conditional -z for old protocols
262 After we get the @RSYNCD greeting from the server, we know it's
263 version but we have not yet sent the command line, so we could just
264 remove the -z option if the server is too old.
266 For ssh invocation it's not so simple, because we actually use the
267 command line to start the remote process. However, we only actually
268 do compression in token.c, and we could therefore once we discover
269 the remote version emit an error if it's too old. I'm not sure if
270 that's a good tradeoff or not.
275 proxy authentication 2002/01/23
277 Allow RSYNC_PROXY to be http://user:pass@proxy.foo:3128/, and do
278 HTTP Basic Proxy-Authentication.
280 Multiple schemes are possible, up to and including the insanity that
281 is NTLM, but Basic probably covers most cases.
288 Add --with-socks, and then perhaps a command-line option to put them
289 on or off. This might be more reliable than LD_PRELOAD hacks.
296 rsync to a FAT partition on a Unix machine doesn't work very well at
297 the moment. I think we get errors about invalid filenames and
298 perhaps also trying to do atomic renames.
300 I guess the code to do this is currently #ifdef'd on Windows;
301 perhaps we ought to intelligently fall back to it on Unix too.
306 Allow forcing arbitrary permissions 2002/03/12
308 On 12 Mar 2002, Dave Dykstra <dwd@bell-labs.com> wrote:
309 > If we would add an option to do that functionality, I
310 > would vote for one that was more general which could mask
311 > off any set of permission bits and possibly add any set of
312 > bits. Perhaps a chmod-like syntax if it could be
313 > implemented simply.
315 I think that would be good too. For example, people uploading files
316 to a web server might like to say
318 rsync -avzP --chmod a+rX ./ sourcefrog.net:/home/www/sourcefrog/
320 Ideally the patch would implement as many of the gnu chmod semantics
321 as possible. I think the mode parser should be a separate function
322 that passes back something like (mask,set) description to the rest
323 of the program. For bonus points there would be a test case for the
326 Possibly also --chown
333 --diff david.e.sewell 2002/03/15
335 Allow people to specify the diff command. (Might want to use wdiff,
338 Just diff the temporary file with the destination file, and delete
339 the tmp file rather than moving it into place.
341 Interaction with --partial.
343 Security interactions with daemon mode?
348 Add daemon --no-detach and --no-fork options
350 Very useful for debugging. Also good when running under a
351 daemon-monitoring process that tries to restart the service when the
357 Create more granular verbosity jw 2003/05/15
359 Control output with the --report option.
361 The option takes as a single argument (no whitespace) a
362 comma delimited lists of keywords.
364 This would separate debugging from "logging" as well as
365 fine grained selection of statistical reporting and what
368 http://lists.samba.org/archive/rsync/2003-May/006059.html
372 DOCUMENTATION --------------------------------------------------------
379 Keep list of open issues and todos on the web site
384 Update web site from CVS
389 Perhaps redo manual as SGML
391 The man page is getting rather large, and there is more information
392 that ought to be added.
394 TexInfo source is probably a dying format.
396 Linuxdoc looks like the most likely contender. I know DocBook is
397 favoured by some people, but it's so bloody verbose, even with emacs
402 LOGGING --------------------------------------------------------------
404 Make dry run list all updates 2002/04/03
408 Mark Santcroos points out that -n fails to list files which have
409 only metadata changes, though it probably should.
411 There may be a Debian bug about this as well.
418 At exit, show how much memory was used for the file list, etc.
420 Also we do a wierd exponential-growth allocation in flist.c. I'm
421 not sure this makes sense with modern mallocs. At any rate it will
422 make us allocate a huge amount of memory for large file lists.
427 Improve error messages
429 If we hang or get SIGINT, then explain where we were up to. Perhaps
430 have a static buffer that contains the current function name, or
431 some kind of description of what we were trying to do. This is a
432 little easier on people than needing to run strace/truss.
434 "The dungeon collapses! You are killed." Rather than "unexpected
435 eof" give a message that is more detailed if possible and also more
438 If we get an error writing to a socket, then we should perhaps
439 continue trying to read to see if an error message comes across
440 explaining why the socket is closed. I'm not sure if this would
441 work, but it would certainly make our messages more helpful.
443 What happens if a directory is missing -x attributes. Do we lose
444 our load? (Debian #28416) Probably fixed now, but a test case would
452 Better statistics: Rasmus 2002/03/08
455 hey, how about an rsync option that just gives you the
456 summary without the list of files? And perhaps gives
457 more information like the number of new files, number
458 of changed, deleted, etc. ?
461 nice idea there is --stats but at the moment it's very
462 tridge-oriented rather than user-friendly it would be
463 nice to improve it that would also work well with
469 Perhaps flush stdout like syslog
471 Perhaps flush stdout after each filename, so that people trying to
472 monitor progress in a log file can do so more easily. See
473 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=48108
478 Log deamon sessions that just list modules
480 At the connections that just get a list of modules are not logged,
486 Log child death on signal
488 If a child of the rsync daemon dies with a signal, we should notice
489 that when we reap it and log a message.
494 Keep stderr and stdout properly separated (Debian #23626)
499 Log errors with function that reports process of origin
501 Use a separate function for reporting errors; prefix it with
502 "rsync:" or "rsync(remote)", or perhaps even "rsync(local
508 verbose output David Stein 2001/12/20
510 Indicate whether files are new, updated, or deleted
512 At end of transfer, show how many files were or were not transferred
518 Add reason for transfer to file logging
520 Explain *why* every file is transferred or not (e.g. "local mtime
521 123123 newer than 1283198")
526 debugging of daemon 2002/04/08
528 Add an rsyncd.conf parameter to turn on debugging on the server.
535 Change to using gettext(). Probably need to ship this for platforms
538 Solicit translations.
540 Does anyone care? Before we bother modifying the code, we ought to
541 get the manual translated first, because that's possibly more useful
542 and at any rate demonstrates desire.
546 DEVELOPMENT --------------------------------------------------------
548 Handling duplicate names
550 We need to be careful of duplicate names getting into the file list.
551 See clean_flist(). This could happen if multiple arguments include
554 I think duplicates are only a problem if they're both flowing
555 through the pipeline at the same time. For example we might have
556 updated the first occurrence after reading the checksums for the
557 second. So possibly we just need to make sure that we don't have
558 both in the pipeline at the same time.
560 Possibly if we did one directory at a time that would be sufficient.
562 Alternatively we could pre-process the arguments to make sure no
563 duplicates will ever be inserted. There could be some bad cases
564 when we're collapsing symlinks.
566 We could have a hash table.
568 The root of the problem is that we do not want more than one file
569 list entry referring to the same file. At first glance there are
570 several ways this could happen: symlinks, hardlinks, and repeated
571 names on the command line.
573 If names are repeated on the command line, they may be present in
574 different forms, perhaps by traversing directory paths in different
575 ways, traversing paths including symlinks. Also we need to allow
576 for expansion of globs by rsync.
578 At the moment, clean_flist() requires having the entire file list in
579 memory. Duplicate names are detected just by a string comparison.
581 We don't need to worry about hard links causing duplicates because
582 files are never updated in place. Similarly for symlinks.
584 I think even if we're using a different symlink mode we don't need
587 Unless we're really clever this will introduce a protocol
588 incompatibility, so we need to be able to accept the old format as
594 Use generic zlib 2002/02/25
596 Perhaps don't use our own zlib.
600 - will automatically be up to date with bugfixes in zlib
602 - can leave it out for small rsync on e.g. recovery disks
604 - can use a shared library
606 - avoids people breaking rsync by trying to do this themselves and
609 Should we ship zlib for systems that don't have it, or require
610 people to install it separately?
612 Apparently this will make us incompatible with versions of rsync
613 that use the patched version of rsync. Probably the simplest way to
614 do this is to just disable gzip (with a warning) when talking to old
622 Rather than storing the file list in memory, store it in a TDB.
624 This *might* make memory usage lower while building the file list.
626 Hashtable lookup will mean files are not transmitted in order,
629 This would neatly eliminate one of the major post-fork shared data
637 Build rsync with SPLINT to try to find security holes. Add
638 annotations as necessary. Keep track of the number of warnings
639 found initially, and see how many of them are real bugs, or real
640 security bugs. Knowing the percentage of likely hits would be
641 really interesting for other projects.
648 jra recommends Valgrind:
650 http://devel-home.kde.org/~sewardj/
655 Create release script
661 Build tar file; upload
663 Send announcement to mailing list and c.o.l.a.
665 Make freshmeat announcement
672 Add machines to build farm
674 Cygwin (on different versions of Win32?)
676 HP-UX variants (via HP?)
684 PERFORMANCE ----------------------------------------------------------
686 File list structure in memory
688 Rather than one big array, perhaps have a tree in memory mirroring
691 This might make sorting much faster! (I'm not sure it's a big CPU
694 It might also reduce memory use in storing repeated directory names
695 -- again I'm not sure this is a problem.
700 Traverse just one directory at a time
702 Traverse just one directory at a time. Tridge says it's possible.
704 At the moment rsync reads the whole file list into memory at the
705 start, which makes us use a lot of memory and also not pipeline
706 network access as much as we could.
713 At the moment hardlink handling is very expensive, so it's off by
714 default. It does not need to be so.
716 Since most of the solutions are rather intertwined with the file
717 list it is probably better to fix that first, although fixing
718 hardlinks is possibly simpler.
720 We can rule out hardlinked directories since they will probably
721 screw us up in all kinds of ways. They simply should not be used.
723 At the moment rsync only cares about hardlinks to regular files. I
724 guess you could also use them for sockets, devices and other beasts,
725 but I have not seen them.
727 When trying to reproduce hard links, we only need to worry about
728 files that have more than one name (nlinks>1 && !S_ISDIR).
730 The basic point of this is to discover alternate names that refer to
731 the same file. All operations, including creating the file and
732 writing modifications to it need only to be done for the first name.
733 For all later names, we just create the link and then leave it
736 If hard links are to be preserved:
738 Before the generator/receiver fork, the list of files is received
739 from the sender (recv_file_list), and a table for detecting hard
742 The generator looks for hard links within the file list and does
743 not send checksums for them, though it does send other metadata.
745 The sender sends the device number and inode with file entries, so
746 that files are uniquely identified.
748 The receiver goes through and creates hard links (do_hard_links)
749 after all data has been written, but before directory permissions
752 At the moment device and inum are sent as 4-byte integers, which
753 will probably cause problems on large filesystems. On Linux the
754 kernel uses 64-bit ino_t's internally, and people will soon have
755 filesystems big enough to use them. We ought to follow NFS4 in
756 using 64-bit device and inode identification, perhaps with a
757 protocol version bump.
759 Once we've seen all the names for a particular file, we no longer
760 need to think about it and we can deallocate the memory.
762 We can also have the case where there are links to a file that are
763 not in the tree being transferred. There's nothing we can do about
764 that. Because we rename the destination into place after writing,
765 any hardlinks to the old file are always going to be orphaned. In
766 fact that is almost necessary because otherwise we'd get really
767 confused if we were generating checksums for one name of a file and
770 At the moment the code seems to make a whole second copy of the file
771 list, which seems unnecessary.
773 We should have a test case that exercises hard links. Since it
774 might be hard to compare ./tls output where the inodes change we
775 might need a little program to check whether several names refer to
781 Allow skipping MD4 file_sum 2002/04/08
783 If we're doing a local transfer, or using -W, then perhaps don't
784 send the file checksum. If we're doing a local transfer, then
785 calculating MD4 checksums uses 90% of CPU and is unlikely to be
788 Indeed for transfers over zlib or ssh we can also rely on the
789 transport to have quite strong protection against corruption.
791 Perhaps we should have an option to disable this,
792 analogous to --whole-file, although it would default to
793 disabled. The file checksum takes up a definite space in
794 the protocol -- we can either set it to 0, or perhaps just
802 Perhaps borrow an assembler MD4 from someone?
804 Make sure we call MD4 with properly-sized blocks whenever possible
805 to avoid copying into the residue region?
812 Test whether this is actually faster than just using malloc(). If
813 it's not (anymore), throw it out.
817 TESTING --------------------------------------------------------------
821 Something that just keeps running rsync continuously over a data set
822 likely to generate problems.
827 Cross-test versions 2001/08/22
829 Part of the regression suite should be making sure that we
830 don't break backwards compatibility: old clients vs new
831 servers and so on. Ideally we would test both up and down
832 from the current release to all old versions.
834 Run current rsync versions against significant past releases.
836 We might need to omit broken old versions, or versions in which
837 particular functionality is broken
839 It might be sufficient to test downloads from well-known public
840 rsync servers running different versions of rsync. This will give
841 some testing and also be the most common case for having different
842 versions and not being able to upgrade.
844 The new --protocol option may help in this.
849 Test on kernel source
851 Download all versions of kernel; unpack, sync between them. Also
852 sync between uncompressed tarballs. Compare directories after
855 Use local mode; ssh; daemon; --whole-file and --no-whole-file.
857 Use awk to pull out the 'speedup' number for each transfer. Make
865 Sparse and non-sparse
870 Create mutator program for testing
872 Insert bytes, delete bytes, swap blocks, ...
877 Create configure option to enable dangerous tests
882 If tests are skipped, say why.
887 Test daemon feature to disallow particular options.
892 Create pipe program for testing
894 Create pipe program that makes slow/jerky connections for
895 testing Versions of read() and write() that corrupt the
896 stream, or abruptly fail
901 Create test makefile target for some tests
903 Separate makefile target to run rough tests -- or perhaps
904 just run them every time?
909 Test "refuse options" works
911 What about for --recursive?
913 If you specify an unrecognized option here, you should get an error.
915 We need a test case for this...
917 Was this broken when we changed to popt?
921 RELATED PROJECTS -----------------------------------------------------
925 Write a small emulation of interactive ftp as a Pythonn program
926 that calls rsync. Commands such as "cd", "ls", "ls *.c" etc map
927 fairly directly into rsync commands: it just needs to remember the
928 current host, directory and so on. We can probably even do
929 completion of remote filenames.
934 http://rsync.samba.org/rsync-and-debian/
942 Exhaustive, tortuous testing
949 rsyncsplit as alternative to real integration with gzip?
954 reverse rsync over HTTP Range
956 Goswin Brederlow suggested this on Debian; I think tridge and I
957 talked about it previous in relation to rproxy.