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-fork option
25 Create more granular verbosity jw 2003/05/15
27 DOCUMENTATION --------------------------------------------------------
29 Keep list of open issues and todos on the web site
30 Perhaps redo manual as SGML
32 LOGGING --------------------------------------------------------------
33 Make dry run list all updates 2002/04/03
35 Improve error messages
36 Better statistics: Rasmus 2002/03/08
37 Perhaps flush stdout like syslog
38 Log deamon sessions that just list modules
39 Log child death on signal
40 Keep stderr and stdout properly separated (Debian #23626)
41 Log errors with function that reports process of origin
42 verbose output David Stein 2001/12/20
43 Add reason for transfer to file logging
44 debugging of daemon 2002/04/08
47 DEVELOPMENT --------------------------------------------------------
48 Handling duplicate names
49 Use generic zlib 2002/02/25
54 Add machines to build farm
56 PERFORMANCE ----------------------------------------------------------
57 File list structure in memory
58 Traverse just one directory at a time
59 Allow skipping MD4 file_sum 2002/04/08
62 TESTING --------------------------------------------------------------
64 Cross-test versions 2001/08/22
67 Create mutator program for testing
68 Create configure option to enable dangerous tests
69 If tests are skipped, say why.
70 Test daemon feature to disallow particular options.
71 Create pipe program for testing
72 Create test makefile target for some tests
74 RELATED PROJECTS -----------------------------------------------------
76 http://rsync.samba.org/rsync-and-debian/
78 rsyncsplit as alternative to real integration with gzip?
79 reverse rsync over HTTP Range
83 BUGS ---------------------------------------------------------------
85 Fix progress indicator to not corrupt log
87 Progress indicator can produce corrupt output when transferring directories:
90 main/binary-arm/admin/
92 main/binary-arm/comm/8.56kB/s 0:00:52
93 main/binary-arm/devel/
95 main/binary-arm/editors/
96 main/binary-arm/electronics/s 0:00:53
97 main/binary-arm/games/
98 main/binary-arm/graphics/
99 main/binary-arm/hamradio/
100 main/binary-arm/interpreters/
101 main/binary-arm/libs/6.61kB/s 0:00:54
102 main/binary-arm/mail/
103 main/binary-arm/math/
104 main/binary-arm/misc/
111 I don't think we handle this properly on systems that don't have the
112 call. Are there any such?
117 Do not rely on having a group called "nobody"
119 http://www.linuxbase.org/spec/refspecs/LSB_1.1.0/gLSB/usernames.html
121 On Debian it's "nogroup"
126 Incorrect timestamps (Debian #100295)
128 A bit hard to believe, but apparently it happens.
135 Don't detach, because this messes up --srvany.
137 http://sources.redhat.com/ml/cygwin/2001-08/msg00234.html
143 FEATURES ------------------------------------------------------------
145 server-imposed bandwidth limits
152 There are already some patches to do this.
154 BitKeeper uses a server whose login shell is set to bkd. That's
155 probably a reasonable approach.
160 Use chroot only if supported
162 If the platform doesn't support it, then don't even try.
164 If running as non-root, then don't fail, just give a warning.
165 (There was a thread about this a while ago?)
167 http://lists.samba.org/pipermail/rsync/2001-August/thread.html
168 http://lists.samba.org/pipermail/rsync/2001-September/thread.html
173 Allow supplementary groups in rsyncd.conf 2002/04/09
175 Perhaps allow supplementary groups to be specified in rsyncd.conf;
176 then make the first one the primary gid and all the rest be
182 Handling IPv6 on old machines
184 The KAME IPv6 patch is nice in theory but has proved a bit of a
185 nightmare in practice. The basic idea of their patch is that rsync
186 is rewritten to use the new getaddrinfo()/getnameinfo() interface,
187 rather than gethostbyname()/gethostbyaddr() as in rsync 2.4.6.
188 Systems that don't have the new interface are handled by providing
189 our own implementation in lib/, which is selectively linked in.
191 The problem with this is that it is really hard to get right on
192 platforms that have a half-working implementation, so redefining
193 these functions clashes with system headers, and leaving them out
194 breaks. This affects at least OSF/1, RedHat 5, and Cobalt, which
195 are moderately improtant.
197 Perhaps the simplest solution would be to have two different files
198 implementing the same interface, and choose either the new or the
199 old API. This is probably necessary for systems that e.g. have
200 IPv6, but gethostbyaddr() can't handle it. The Linux manpage claims
201 this is currently the case.
203 In fact, our internal sockets interface (things like
204 open_socket_out(), etc) is much narrower than the getaddrinfo()
205 interface, and so probably simpler to get right. In addition, the
206 old code is known to work well on old machines.
208 We could drop the rather large lib/getaddrinfo files.
215 Implement suggestions from http://www.kame.net/newsletter/19980604/
216 and ftp://ftp.iij.ad.jp/pub/RFC/rfc2553.txt
218 If a host has multiple addresses, then listen try to connect to all
219 in order until we get through. (getaddrinfo may return multiple
220 addresses.) This is kind of implemented already.
222 Possibly also when starting as a server we may need to listen on
223 multiple passive addresses. This might be a bit harder, because we
224 may need to select on all of them. Hm.
226 Define a syntax for IPv6 literal addresses. Since they include
227 colons, they tend to break most naming systems, including ours.
228 Based on the HTTP IPv6 syntax, I think we should use
230 rsync://[::1]/foo/bar [::1]::bar
232 which should just take a small change to the parser code.
237 Add ACL support 2001/12/02
239 Transfer ACLs. Need to think of a standard representation.
240 Probably better not to even try to convert between NT and POSIX.
241 Possibly can share some code with Samba.
246 Lazy directory creation
248 With the current common --include '*/' --exclude '*' pattern, people
249 can end up with many empty directories. We might avoid this by
250 lazily creating such directories.
255 proxy authentication 2002/01/23
257 Allow RSYNC_PROXY to be http://user:pass@proxy.foo:3128/, and do
258 HTTP Basic Proxy-Authentication.
260 Multiple schemes are possible, up to and including the insanity that
261 is NTLM, but Basic probably covers most cases.
268 Add --with-socks, and then perhaps a command-line option to put them
269 on or off. This might be more reliable than LD_PRELOAD hacks.
276 rsync to a FAT partition on a Unix machine doesn't work very well at
277 the moment. I think we get errors about invalid filenames and
278 perhaps also trying to do atomic renames.
280 I guess the code to do this is currently #ifdef'd on Windows;
281 perhaps we ought to intelligently fall back to it on Unix too.
286 Allow forcing arbitrary permissions 2002/03/12
288 On 12 Mar 2002, Dave Dykstra <dwd@bell-labs.com> wrote:
289 > If we would add an option to do that functionality, I
290 > would vote for one that was more general which could mask
291 > off any set of permission bits and possibly add any set of
292 > bits. Perhaps a chmod-like syntax if it could be
293 > implemented simply.
295 I think that would be good too. For example, people uploading files
296 to a web server might like to say
298 rsync -avzP --chmod a+rX ./ sourcefrog.net:/home/www/sourcefrog/
300 Ideally the patch would implement as many of the gnu chmod semantics
301 as possible. I think the mode parser should be a separate function
302 that passes back something like (mask,set) description to the rest
303 of the program. For bonus points there would be a test case for the
306 Possibly also --chown
313 --diff david.e.sewell 2002/03/15
315 Allow people to specify the diff command. (Might want to use wdiff,
318 Just diff the temporary file with the destination file, and delete
319 the tmp file rather than moving it into place.
321 Interaction with --partial.
323 Security interactions with daemon mode?
328 Add daemon --no-fork option
330 Very useful for debugging. Also good when running under a
331 daemon-monitoring process that tries to restart the service when the
337 Create more granular verbosity jw 2003/05/15
339 Control output with the --report option.
341 The option takes as a single argument (no whitespace) a
342 comma delimited lists of keywords.
344 This would separate debugging from "logging" as well as
345 fine grained selection of statistical reporting and what
348 http://lists.samba.org/archive/rsync/2003-May/006059.html
352 DOCUMENTATION --------------------------------------------------------
359 Keep list of open issues and todos on the web site
364 Perhaps redo manual as SGML
366 The man page is getting rather large, and there is more information
367 that ought to be added.
369 TexInfo source is probably a dying format.
371 Linuxdoc looks like the most likely contender. I know DocBook is
372 favoured by some people, but it's so bloody verbose, even with emacs
377 LOGGING --------------------------------------------------------------
379 Make dry run list all updates 2002/04/03
383 Mark Santcroos points out that -n fails to list files which have
384 only metadata changes, though it probably should.
386 There may be a Debian bug about this as well.
393 At exit, show how much memory was used for the file list, etc.
395 Also we do a wierd exponential-growth allocation in flist.c. I'm
396 not sure this makes sense with modern mallocs. At any rate it will
397 make us allocate a huge amount of memory for large file lists.
402 Improve error messages
404 If we hang or get SIGINT, then explain where we were up to. Perhaps
405 have a static buffer that contains the current function name, or
406 some kind of description of what we were trying to do. This is a
407 little easier on people than needing to run strace/truss.
409 "The dungeon collapses! You are killed." Rather than "unexpected
410 eof" give a message that is more detailed if possible and also more
413 If we get an error writing to a socket, then we should perhaps
414 continue trying to read to see if an error message comes across
415 explaining why the socket is closed. I'm not sure if this would
416 work, but it would certainly make our messages more helpful.
418 What happens if a directory is missing -x attributes. Do we lose
419 our load? (Debian #28416) Probably fixed now, but a test case would
422 When running as a daemon, some errors should both be returned to the
423 user and logged. This will make interacting with a daemon less
429 Better statistics: Rasmus 2002/03/08
432 hey, how about an rsync option that just gives you the
433 summary without the list of files? And perhaps gives
434 more information like the number of new files, number
435 of changed, deleted, etc. ?
438 nice idea there is --stats but at the moment it's very
439 tridge-oriented rather than user-friendly it would be
440 nice to improve it that would also work well with
446 Perhaps flush stdout like syslog
448 Perhaps flush stdout after each filename, so that people trying to
449 monitor progress in a log file can do so more easily. See
450 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=48108
455 Log deamon sessions that just list modules
457 At the connections that just get a list of modules are not logged,
463 Log child death on signal
465 If a child of the rsync daemon dies with a signal, we should notice
466 that when we reap it and log a message.
471 Keep stderr and stdout properly separated (Debian #23626)
476 Log errors with function that reports process of origin
478 Use a separate function for reporting errors; prefix it with
479 "rsync:" or "rsync(remote)", or perhaps even "rsync(local
485 verbose output David Stein 2001/12/20
487 Indicate whether files are new, updated, or deleted
489 At end of transfer, show how many files were or were not transferred
495 Add reason for transfer to file logging
497 Explain *why* every file is transferred or not (e.g. "local mtime
498 123123 newer than 1283198")
503 debugging of daemon 2002/04/08
505 Add an rsyncd.conf parameter to turn on debugging on the server.
512 Change to using gettext(). Probably need to ship this for platforms
515 Solicit translations.
517 Does anyone care? Before we bother modifying the code, we ought to
518 get the manual translated first, because that's possibly more useful
519 and at any rate demonstrates desire.
523 DEVELOPMENT --------------------------------------------------------
525 Handling duplicate names
527 We need to be careful of duplicate names getting into the file list.
528 See clean_flist(). This could happen if multiple arguments include
531 I think duplicates are only a problem if they're both flowing
532 through the pipeline at the same time. For example we might have
533 updated the first occurrence after reading the checksums for the
534 second. So possibly we just need to make sure that we don't have
535 both in the pipeline at the same time.
537 Possibly if we did one directory at a time that would be sufficient.
539 Alternatively we could pre-process the arguments to make sure no
540 duplicates will ever be inserted. There could be some bad cases
541 when we're collapsing symlinks.
543 We could have a hash table.
545 The root of the problem is that we do not want more than one file
546 list entry referring to the same file. At first glance there are
547 several ways this could happen: symlinks, hardlinks, and repeated
548 names on the command line.
550 If names are repeated on the command line, they may be present in
551 different forms, perhaps by traversing directory paths in different
552 ways, traversing paths including symlinks. Also we need to allow
553 for expansion of globs by rsync.
555 At the moment, clean_flist() requires having the entire file list in
556 memory. Duplicate names are detected just by a string comparison.
558 We don't need to worry about hard links causing duplicates because
559 files are never updated in place. Similarly for symlinks.
561 I think even if we're using a different symlink mode we don't need
564 Unless we're really clever this will introduce a protocol
565 incompatibility, so we need to be able to accept the old format as
571 Use generic zlib 2002/02/25
573 Perhaps don't use our own zlib.
577 - will automatically be up to date with bugfixes in zlib
579 - can leave it out for small rsync on e.g. recovery disks
581 - can use a shared library
583 - avoids people breaking rsync by trying to do this themselves and
586 Should we ship zlib for systems that don't have it, or require
587 people to install it separately?
589 Apparently this will make us incompatible with versions of rsync
590 that use the patched version of rsync. Probably the simplest way to
591 do this is to just disable gzip (with a warning) when talking to old
599 Rather than storing the file list in memory, store it in a TDB.
601 This *might* make memory usage lower while building the file list.
603 Hashtable lookup will mean files are not transmitted in order,
606 This would neatly eliminate one of the major post-fork shared data
614 Build rsync with SPLINT to try to find security holes. Add
615 annotations as necessary. Keep track of the number of warnings
616 found initially, and see how many of them are real bugs, or real
617 security bugs. Knowing the percentage of likely hits would be
618 really interesting for other projects.
625 jra recommends Valgrind:
627 http://devel-home.kde.org/~sewardj/
632 Create release script
638 Build tar file; upload
640 Send announcement to mailing list and c.o.l.a.
642 Make freshmeat announcement
649 Add machines to build farm
651 Cygwin (on different versions of Win32?)
653 HP-UX variants (via HP?)
661 PERFORMANCE ----------------------------------------------------------
663 File list structure in memory
665 Rather than one big array, perhaps have a tree in memory mirroring
668 This might make sorting much faster! (I'm not sure it's a big CPU
671 It might also reduce memory use in storing repeated directory names
672 -- again I'm not sure this is a problem.
677 Traverse just one directory at a time
679 Traverse just one directory at a time. Tridge says it's possible.
681 At the moment rsync reads the whole file list into memory at the
682 start, which makes us use a lot of memory and also not pipeline
683 network access as much as we could.
688 Allow skipping MD4 file_sum 2002/04/08
690 If we're doing a local transfer, or using -W, then perhaps don't
691 send the file checksum. If we're doing a local transfer, then
692 calculating MD4 checksums uses 90% of CPU and is unlikely to be
695 Indeed for transfers over zlib or ssh we can also rely on the
696 transport to have quite strong protection against corruption.
698 Perhaps we should have an option to disable this,
699 analogous to --whole-file, although it would default to
700 disabled. The file checksum takes up a definite space in
701 the protocol -- we can either set it to 0, or perhaps just
709 Perhaps borrow an assembler MD4 from someone?
711 Make sure we call MD4 with properly-sized blocks whenever possible
712 to avoid copying into the residue region?
716 TESTING --------------------------------------------------------------
720 Something that just keeps running rsync continuously over a data set
721 likely to generate problems.
726 Cross-test versions 2001/08/22
728 Part of the regression suite should be making sure that we
729 don't break backwards compatibility: old clients vs new
730 servers and so on. Ideally we would test both up and down
731 from the current release to all old versions.
733 Run current rsync versions against significant past releases.
735 We might need to omit broken old versions, or versions in which
736 particular functionality is broken
738 It might be sufficient to test downloads from well-known public
739 rsync servers running different versions of rsync. This will give
740 some testing and also be the most common case for having different
741 versions and not being able to upgrade.
743 The new --protocol option may help in this.
748 Test on kernel source
750 Download all versions of kernel; unpack, sync between them. Also
751 sync between uncompressed tarballs. Compare directories after
754 Use local mode; ssh; daemon; --whole-file and --no-whole-file.
756 Use awk to pull out the 'speedup' number for each transfer. Make
764 Sparse and non-sparse
769 Create mutator program for testing
771 Insert bytes, delete bytes, swap blocks, ...
776 Create configure option to enable dangerous tests
781 If tests are skipped, say why.
786 Test daemon feature to disallow particular options.
791 Create pipe program for testing
793 Create pipe program that makes slow/jerky connections for
794 testing Versions of read() and write() that corrupt the
795 stream, or abruptly fail
800 Create test makefile target for some tests
802 Separate makefile target to run rough tests -- or perhaps
803 just run them every time?
807 RELATED PROJECTS -----------------------------------------------------
811 Write a small emulation of interactive ftp as a Pythonn program
812 that calls rsync. Commands such as "cd", "ls", "ls *.c" etc map
813 fairly directly into rsync commands: it just needs to remember the
814 current host, directory and so on. We can probably even do
815 completion of remote filenames.
820 http://rsync.samba.org/rsync-and-debian/
828 Exhaustive, tortuous testing
835 rsyncsplit as alternative to real integration with gzip?
840 reverse rsync over HTTP Range
842 Goswin Brederlow suggested this on Debian; I think tridge and I
843 talked about it previous in relation to rproxy.