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 proxy authentication 2002/01/23
22 Allow forcing arbitrary permissions 2002/03/12
23 --diff david.e.sewell 2002/03/15
24 Add daemon --no-detach and --no-fork options
25 Create more granular verbosity jw 2003/05/15
27 DOCUMENTATION --------------------------------------------------------
29 Keep list of open issues and todos on the web site
30 Update web site from CVS
31 Perhaps redo manual as SGML
33 LOGGING --------------------------------------------------------------
34 Make dry run list all updates 2002/04/03
36 Improve error messages
37 Better statistics: Rasmus 2002/03/08
38 Perhaps flush stdout like syslog
39 Log deamon sessions that just list modules
40 Log child death on signal
41 Keep stderr and stdout properly separated (Debian #23626)
42 Log errors with function that reports process of origin
43 verbose output David Stein 2001/12/20
44 Add reason for transfer to file logging
45 debugging of daemon 2002/04/08
48 DEVELOPMENT --------------------------------------------------------
49 Handling duplicate names
50 Use generic zlib 2002/02/25
55 Add machines to build farm
57 PERFORMANCE ----------------------------------------------------------
58 File list structure in memory
59 Traverse just one directory at a time
60 Allow skipping MD4 file_sum 2002/04/08
63 TESTING --------------------------------------------------------------
65 Cross-test versions 2001/08/22
68 Create mutator program for testing
69 Create configure option to enable dangerous tests
70 If tests are skipped, say why.
71 Test daemon feature to disallow particular options.
72 Create pipe program for testing
73 Create test makefile target for some tests
75 RELATED PROJECTS -----------------------------------------------------
77 http://rsync.samba.org/rsync-and-debian/
79 rsyncsplit as alternative to real integration with gzip?
80 reverse rsync over HTTP Range
84 BUGS ---------------------------------------------------------------
86 Fix progress indicator to not corrupt log
88 Progress indicator can produce corrupt output when transferring directories:
91 main/binary-arm/admin/
93 main/binary-arm/comm/8.56kB/s 0:00:52
94 main/binary-arm/devel/
96 main/binary-arm/editors/
97 main/binary-arm/electronics/s 0:00:53
98 main/binary-arm/games/
99 main/binary-arm/graphics/
100 main/binary-arm/hamradio/
101 main/binary-arm/interpreters/
102 main/binary-arm/libs/6.61kB/s 0:00:54
103 main/binary-arm/mail/
104 main/binary-arm/math/
105 main/binary-arm/misc/
112 I don't think we handle this properly on systems that don't have the
113 call. Are there any such?
118 Do not rely on having a group called "nobody"
120 http://www.linuxbase.org/spec/refspecs/LSB_1.1.0/gLSB/usernames.html
122 On Debian it's "nogroup"
127 Incorrect timestamps (Debian #100295)
129 A bit hard to believe, but apparently it happens.
136 Don't detach, because this messes up --srvany.
138 http://sources.redhat.com/ml/cygwin/2001-08/msg00234.html
144 FEATURES ------------------------------------------------------------
146 server-imposed bandwidth limits
153 There are already some patches to do this.
155 BitKeeper uses a server whose login shell is set to bkd. That's
156 probably a reasonable approach.
161 Use chroot only if supported
163 If the platform doesn't support it, then don't even try.
165 If running as non-root, then don't fail, just give a warning.
166 (There was a thread about this a while ago?)
168 http://lists.samba.org/pipermail/rsync/2001-August/thread.html
169 http://lists.samba.org/pipermail/rsync/2001-September/thread.html
174 Allow supplementary groups in rsyncd.conf 2002/04/09
176 Perhaps allow supplementary groups to be specified in rsyncd.conf;
177 then make the first one the primary gid and all the rest be
183 Handling IPv6 on old machines
185 The KAME IPv6 patch is nice in theory but has proved a bit of a
186 nightmare in practice. The basic idea of their patch is that rsync
187 is rewritten to use the new getaddrinfo()/getnameinfo() interface,
188 rather than gethostbyname()/gethostbyaddr() as in rsync 2.4.6.
189 Systems that don't have the new interface are handled by providing
190 our own implementation in lib/, which is selectively linked in.
192 The problem with this is that it is really hard to get right on
193 platforms that have a half-working implementation, so redefining
194 these functions clashes with system headers, and leaving them out
195 breaks. This affects at least OSF/1, RedHat 5, and Cobalt, which
196 are moderately improtant.
198 Perhaps the simplest solution would be to have two different files
199 implementing the same interface, and choose either the new or the
200 old API. This is probably necessary for systems that e.g. have
201 IPv6, but gethostbyaddr() can't handle it. The Linux manpage claims
202 this is currently the case.
204 In fact, our internal sockets interface (things like
205 open_socket_out(), etc) is much narrower than the getaddrinfo()
206 interface, and so probably simpler to get right. In addition, the
207 old code is known to work well on old machines.
209 We could drop the rather large lib/getaddrinfo files.
216 Implement suggestions from http://www.kame.net/newsletter/19980604/
217 and ftp://ftp.iij.ad.jp/pub/RFC/rfc2553.txt
219 If a host has multiple addresses, then listen try to connect to all
220 in order until we get through. (getaddrinfo may return multiple
221 addresses.) This is kind of implemented already.
223 Possibly also when starting as a server we may need to listen on
224 multiple passive addresses. This might be a bit harder, because we
225 may need to select on all of them. Hm.
227 Define a syntax for IPv6 literal addresses. Since they include
228 colons, they tend to break most naming systems, including ours.
229 Based on the HTTP IPv6 syntax, I think we should use
231 rsync://[::1]/foo/bar [::1]::bar
233 which should just take a small change to the parser code.
238 Add ACL support 2001/12/02
240 Transfer ACLs. Need to think of a standard representation.
241 Probably better not to even try to convert between NT and POSIX.
242 Possibly can share some code with Samba.
247 Lazy directory creation
249 With the current common --include '*/' --exclude '*' pattern, people
250 can end up with many empty directories. We might avoid this by
251 lazily creating such directories.
256 proxy authentication 2002/01/23
258 Allow RSYNC_PROXY to be http://user:pass@proxy.foo:3128/, and do
259 HTTP Basic Proxy-Authentication.
261 Multiple schemes are possible, up to and including the insanity that
262 is NTLM, but Basic probably covers most cases.
269 Add --with-socks, and then perhaps a command-line option to put them
270 on or off. This might be more reliable than LD_PRELOAD hacks.
277 rsync to a FAT partition on a Unix machine doesn't work very well at
278 the moment. I think we get errors about invalid filenames and
279 perhaps also trying to do atomic renames.
281 I guess the code to do this is currently #ifdef'd on Windows;
282 perhaps we ought to intelligently fall back to it on Unix too.
287 Allow forcing arbitrary permissions 2002/03/12
289 On 12 Mar 2002, Dave Dykstra <dwd@bell-labs.com> wrote:
290 > If we would add an option to do that functionality, I
291 > would vote for one that was more general which could mask
292 > off any set of permission bits and possibly add any set of
293 > bits. Perhaps a chmod-like syntax if it could be
294 > implemented simply.
296 I think that would be good too. For example, people uploading files
297 to a web server might like to say
299 rsync -avzP --chmod a+rX ./ sourcefrog.net:/home/www/sourcefrog/
301 Ideally the patch would implement as many of the gnu chmod semantics
302 as possible. I think the mode parser should be a separate function
303 that passes back something like (mask,set) description to the rest
304 of the program. For bonus points there would be a test case for the
307 Possibly also --chown
314 --diff david.e.sewell 2002/03/15
316 Allow people to specify the diff command. (Might want to use wdiff,
319 Just diff the temporary file with the destination file, and delete
320 the tmp file rather than moving it into place.
322 Interaction with --partial.
324 Security interactions with daemon mode?
329 Add daemon --no-detach and --no-fork options
331 Very useful for debugging. Also good when running under a
332 daemon-monitoring process that tries to restart the service when the
338 Create more granular verbosity jw 2003/05/15
340 Control output with the --report option.
342 The option takes as a single argument (no whitespace) a
343 comma delimited lists of keywords.
345 This would separate debugging from "logging" as well as
346 fine grained selection of statistical reporting and what
349 http://lists.samba.org/archive/rsync/2003-May/006059.html
353 DOCUMENTATION --------------------------------------------------------
360 Keep list of open issues and todos on the web site
365 Update web site from CVS
370 Perhaps redo manual as SGML
372 The man page is getting rather large, and there is more information
373 that ought to be added.
375 TexInfo source is probably a dying format.
377 Linuxdoc looks like the most likely contender. I know DocBook is
378 favoured by some people, but it's so bloody verbose, even with emacs
383 LOGGING --------------------------------------------------------------
385 Make dry run list all updates 2002/04/03
389 Mark Santcroos points out that -n fails to list files which have
390 only metadata changes, though it probably should.
392 There may be a Debian bug about this as well.
399 At exit, show how much memory was used for the file list, etc.
401 Also we do a wierd exponential-growth allocation in flist.c. I'm
402 not sure this makes sense with modern mallocs. At any rate it will
403 make us allocate a huge amount of memory for large file lists.
408 Improve error messages
410 If we hang or get SIGINT, then explain where we were up to. Perhaps
411 have a static buffer that contains the current function name, or
412 some kind of description of what we were trying to do. This is a
413 little easier on people than needing to run strace/truss.
415 "The dungeon collapses! You are killed." Rather than "unexpected
416 eof" give a message that is more detailed if possible and also more
419 If we get an error writing to a socket, then we should perhaps
420 continue trying to read to see if an error message comes across
421 explaining why the socket is closed. I'm not sure if this would
422 work, but it would certainly make our messages more helpful.
424 What happens if a directory is missing -x attributes. Do we lose
425 our load? (Debian #28416) Probably fixed now, but a test case would
428 When running as a daemon, some errors should both be returned to the
429 user and logged. This will make interacting with a daemon less
435 Better statistics: Rasmus 2002/03/08
438 hey, how about an rsync option that just gives you the
439 summary without the list of files? And perhaps gives
440 more information like the number of new files, number
441 of changed, deleted, etc. ?
444 nice idea there is --stats but at the moment it's very
445 tridge-oriented rather than user-friendly it would be
446 nice to improve it that would also work well with
452 Perhaps flush stdout like syslog
454 Perhaps flush stdout after each filename, so that people trying to
455 monitor progress in a log file can do so more easily. See
456 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=48108
461 Log deamon sessions that just list modules
463 At the connections that just get a list of modules are not logged,
469 Log child death on signal
471 If a child of the rsync daemon dies with a signal, we should notice
472 that when we reap it and log a message.
477 Keep stderr and stdout properly separated (Debian #23626)
482 Log errors with function that reports process of origin
484 Use a separate function for reporting errors; prefix it with
485 "rsync:" or "rsync(remote)", or perhaps even "rsync(local
491 verbose output David Stein 2001/12/20
493 Indicate whether files are new, updated, or deleted
495 At end of transfer, show how many files were or were not transferred
501 Add reason for transfer to file logging
503 Explain *why* every file is transferred or not (e.g. "local mtime
504 123123 newer than 1283198")
509 debugging of daemon 2002/04/08
511 Add an rsyncd.conf parameter to turn on debugging on the server.
518 Change to using gettext(). Probably need to ship this for platforms
521 Solicit translations.
523 Does anyone care? Before we bother modifying the code, we ought to
524 get the manual translated first, because that's possibly more useful
525 and at any rate demonstrates desire.
529 DEVELOPMENT --------------------------------------------------------
531 Handling duplicate names
533 We need to be careful of duplicate names getting into the file list.
534 See clean_flist(). This could happen if multiple arguments include
537 I think duplicates are only a problem if they're both flowing
538 through the pipeline at the same time. For example we might have
539 updated the first occurrence after reading the checksums for the
540 second. So possibly we just need to make sure that we don't have
541 both in the pipeline at the same time.
543 Possibly if we did one directory at a time that would be sufficient.
545 Alternatively we could pre-process the arguments to make sure no
546 duplicates will ever be inserted. There could be some bad cases
547 when we're collapsing symlinks.
549 We could have a hash table.
551 The root of the problem is that we do not want more than one file
552 list entry referring to the same file. At first glance there are
553 several ways this could happen: symlinks, hardlinks, and repeated
554 names on the command line.
556 If names are repeated on the command line, they may be present in
557 different forms, perhaps by traversing directory paths in different
558 ways, traversing paths including symlinks. Also we need to allow
559 for expansion of globs by rsync.
561 At the moment, clean_flist() requires having the entire file list in
562 memory. Duplicate names are detected just by a string comparison.
564 We don't need to worry about hard links causing duplicates because
565 files are never updated in place. Similarly for symlinks.
567 I think even if we're using a different symlink mode we don't need
570 Unless we're really clever this will introduce a protocol
571 incompatibility, so we need to be able to accept the old format as
577 Use generic zlib 2002/02/25
579 Perhaps don't use our own zlib.
583 - will automatically be up to date with bugfixes in zlib
585 - can leave it out for small rsync on e.g. recovery disks
587 - can use a shared library
589 - avoids people breaking rsync by trying to do this themselves and
592 Should we ship zlib for systems that don't have it, or require
593 people to install it separately?
595 Apparently this will make us incompatible with versions of rsync
596 that use the patched version of rsync. Probably the simplest way to
597 do this is to just disable gzip (with a warning) when talking to old
605 Rather than storing the file list in memory, store it in a TDB.
607 This *might* make memory usage lower while building the file list.
609 Hashtable lookup will mean files are not transmitted in order,
612 This would neatly eliminate one of the major post-fork shared data
620 Build rsync with SPLINT to try to find security holes. Add
621 annotations as necessary. Keep track of the number of warnings
622 found initially, and see how many of them are real bugs, or real
623 security bugs. Knowing the percentage of likely hits would be
624 really interesting for other projects.
631 jra recommends Valgrind:
633 http://devel-home.kde.org/~sewardj/
638 Create release script
644 Build tar file; upload
646 Send announcement to mailing list and c.o.l.a.
648 Make freshmeat announcement
655 Add machines to build farm
657 Cygwin (on different versions of Win32?)
659 HP-UX variants (via HP?)
667 PERFORMANCE ----------------------------------------------------------
669 File list structure in memory
671 Rather than one big array, perhaps have a tree in memory mirroring
674 This might make sorting much faster! (I'm not sure it's a big CPU
677 It might also reduce memory use in storing repeated directory names
678 -- again I'm not sure this is a problem.
683 Traverse just one directory at a time
685 Traverse just one directory at a time. Tridge says it's possible.
687 At the moment rsync reads the whole file list into memory at the
688 start, which makes us use a lot of memory and also not pipeline
689 network access as much as we could.
694 Allow skipping MD4 file_sum 2002/04/08
696 If we're doing a local transfer, or using -W, then perhaps don't
697 send the file checksum. If we're doing a local transfer, then
698 calculating MD4 checksums uses 90% of CPU and is unlikely to be
701 Indeed for transfers over zlib or ssh we can also rely on the
702 transport to have quite strong protection against corruption.
704 Perhaps we should have an option to disable this,
705 analogous to --whole-file, although it would default to
706 disabled. The file checksum takes up a definite space in
707 the protocol -- we can either set it to 0, or perhaps just
715 Perhaps borrow an assembler MD4 from someone?
717 Make sure we call MD4 with properly-sized blocks whenever possible
718 to avoid copying into the residue region?
722 TESTING --------------------------------------------------------------
726 Something that just keeps running rsync continuously over a data set
727 likely to generate problems.
732 Cross-test versions 2001/08/22
734 Part of the regression suite should be making sure that we
735 don't break backwards compatibility: old clients vs new
736 servers and so on. Ideally we would test both up and down
737 from the current release to all old versions.
739 Run current rsync versions against significant past releases.
741 We might need to omit broken old versions, or versions in which
742 particular functionality is broken
744 It might be sufficient to test downloads from well-known public
745 rsync servers running different versions of rsync. This will give
746 some testing and also be the most common case for having different
747 versions and not being able to upgrade.
749 The new --protocol option may help in this.
754 Test on kernel source
756 Download all versions of kernel; unpack, sync between them. Also
757 sync between uncompressed tarballs. Compare directories after
760 Use local mode; ssh; daemon; --whole-file and --no-whole-file.
762 Use awk to pull out the 'speedup' number for each transfer. Make
770 Sparse and non-sparse
775 Create mutator program for testing
777 Insert bytes, delete bytes, swap blocks, ...
782 Create configure option to enable dangerous tests
787 If tests are skipped, say why.
792 Test daemon feature to disallow particular options.
797 Create pipe program for testing
799 Create pipe program that makes slow/jerky connections for
800 testing Versions of read() and write() that corrupt the
801 stream, or abruptly fail
806 Create test makefile target for some tests
808 Separate makefile target to run rough tests -- or perhaps
809 just run them every time?
813 RELATED PROJECTS -----------------------------------------------------
817 Write a small emulation of interactive ftp as a Pythonn program
818 that calls rsync. Commands such as "cd", "ls", "ls *.c" etc map
819 fairly directly into rsync commands: it just needs to remember the
820 current host, directory and so on. We can probably even do
821 completion of remote filenames.
826 http://rsync.samba.org/rsync-and-debian/
834 Exhaustive, tortuous testing
841 rsyncsplit as alternative to real integration with gzip?
846 reverse rsync over HTTP Range
848 Goswin Brederlow suggested this on Debian; I think tridge and I
849 talked about it previous in relation to rproxy.