Improved the program that checks for broken large-file locking.

[rsync/rsync.git] / NEWS

NEWS for rsync 2.6.2 (UNRELEASED)
Protocol: 28 (unchanged)
Changes since 2.6.1:

  BUG FIXES:

    - Fixed a major bug in the sorting of the filenames when --relative
      is used for some sources (probably just "/", but don't depend on
      that).  This fix ensures that we ask for the right file-list item
      when requesting changes from the sender.

    - Rsync now checks the return value of the close() function to
      better report disk-full problems on an NFS file system.

    - Restored the old daemon-server behavior of logging error messages
      rather than returning them to the user.  (A better long-term fix
      will be sought in the future.)

    - An obscure uninitialized-variable bug was fixed in the uid/gid
      code.  (This bug probably had no ill effects.)

    - Got rid of the configure check for sys/sysctl.h (it wasn't used
      and was causing a problem on some systems).

  ENHANCEMENTS:

    - Two new diffs were added to the patches dir.

\f
NEWS for rsync 2.6.1 (26 Apr 2004)
Protocol: 28 (changed)
Changes since 2.6.0:

  SECURITY FIXES:

    - Paths sent to an rsync daemon are more thoroughly sanitized when
      chroot is not used.  If you're running a non-read-only rsync
      daemon with chroot disabled, *please upgrade*, ESPECIALLY if the
      user privs you run rsync under is anything above "nobody".

  ENHANCEMENTS:

    - Lower memory use, more optimal transfer of data over the socket,
      and lower CPU usage (see the INTERNAL section for details).

    - The RSYNC_PROXY environment variable can now contain a
      "USER:PASS@" prefix before the "HOST:PORT" information.
      (Bardur Arantsson)

    - The --progress output now mentions how far along in the transfer
      we are, including both a count of files transferred and a
      percentage of the total file-count that we've processed.  It also
      shows better current-rate-of-transfer and remaining-transfer-time
      values.

    - The configure script now accepts --with-rsyncd-conf=PATH to
      override the default value of the /etc/rsyncd.conf file.

    - Added a couple extra diffs in the "patches" dir, removed the ones
      that got applied, and rebuilt the rest.

    - Documentation changes now attempt to describe some often mis-
      understood features more clearly.

  BUG FIXES:

    - When -x (--one-file-system) is combined with -L (--copy-links) or
      --copy-unsafe-links, no symlinked files are skipped, even if the
      referent file is on a different filesystem.

    - The --link-dest code now works properly for a non-root user when
      (1) the UIDs of the source and destination differ and -o was
      specified, or (2) when the group of the source can't be used on
      the destination and -g was specified.

    - Fixed a bug in the handling of -H (hard-links) that might cause
      the expanded PATH/NAME value of the current item to get
      overwritten (due to an expanded-name caching bug).
      
    - We now reset the "new data has been sent" flag at the start of
      each file we send.  This makes sure that an interrupted transfer
      with the --partial option set doesn't keep a shorter temp file
      than the current basis file when no new data has been transfered
      over the wire for that file.

    - Fixed a byte-order problem in --batch-mode on big-endian machines.
      (Jay Fenlason)

    - Fixed configure bug when running "./configure --disable-ipv6".

    - Fixed "make test" bug when build dir is not the source dir.

    - When using --cvs-exclude, the exclude items we get from a
      per-directory's .cvsignore file once again only affect that one
      directory (not all following directories too).  The items are also
      now properly word-split and parsed without any +/- prefix parsing.

    - When specifying the USER@HOST: prefix for a file, the USER part
      can now contain an '@', if needed (i.e. the last '@' is used to
      find the HOST, not the first).

    - Fixed some bugs in the handling of group IDs for non-root users:
      (1) It properly handles a group that the sender didn't have a name
      for (it would previously skip changing the group on any files in
      that group).  (2) If --numeric-ids is used, rsync no longer
      attempts to set groups that the user doesn't have the permission
      to set.

    - Fixed the "refuse options" setting in the rsyncd.conf file.

    - Improved the -x (--one-file-system) flag's handling of any mount-
      point directories we encounter.  It is both more optimal (in that
      it no longer does a useless scan of the contents of the mount-
      point dirs) and also fixes a bug where a remapped mount of the
      original filesystem could get discovered in a subdir we should be
      ignoring.

    - Rsync no longer discards a double-slash at the start of a filename
      when trying to open the file.  It also no longer constructs names
      that start with a double slash (unless the user supplied them).

    - Path-specifying options to a daemon should now work the same with
      or without chroot turned on.  Previously, such a option (such as
      --link-dest) would get its absolute path munged into a relative
      one if chroot was not on, making that setting fairly useless.
      Rsync now transforms the path into one that is based on the
      module's base dir when chroot is not enabled.

    - Fixed compilation problem on Tru64 Unix (having to do with
      sockaddr.sa_len and sockaddr.sin_len).

    - Fixed a compatibility problem interacting with older rsync
      versions that might send us an empty --suffix value without
      telling us that --backup-dir was specified.

    - The "hosts allow" option for a daemon-over-remote-shell process
      now has improved support for IPv6 addresses and a fix for systems
      that have a length field in their socket structs.

    - Fixed the ability to request an empty backup --suffix when sending
      files to an rsync daemon.

  INTERNAL:

    - Most of the I/O is now buffered, which results in a pretty large
      speedup when running under MS Windows.  (Craig Barratt)

    - Optimizations to the name-handling/comparing code have made some
      significant reductions in user-CPU time for large file sets.

    - Some cleanup of the variable types make the code more consistent.

    - Reduced memory requirements of hard link preservation.
      (J.W. Schultz)

    - Implemented a new algorithm for hard-link handling that speeds up
      the code significantly.  (J.W. Schultz and Wayne Davison)

    - The --hard-link option now uses the first existing file in the
      group of linked files as the basis for the transfer.  This
      prevents the sub-optimal transfer of a file's data when a new
      hardlink is added on the sending side and it sorts alphabetically
      earlier in the list than the files that are already present on the
      receiving side.

    - Dropped support for protocol versions less than 20 (2.3.0 released
      15 Mar 1999) and activated warnings for protocols less than 25
      (2.5.0 released 23 Aug 2001). (Wayne Davison and J.W. Schultz,
      severally)

    - More optimal data transmission for --hard-links (protocol 28).

    - More optimal data transmission for --checksum (protocol 28).

    - Less memory is used when --checksum is specified.

    - Less memory is used in the file list (a per-file savings).

    - The generator is now better about not modifying the file list
      during the transfer in order to avoid a copy-on-write memory
      bifurcation (on systems where fork() uses shared memory).
      Previously, rsync's shared memory would slowly become unshared,
      resulting in real memory usage nearly doubling on the receiving
      side by the end of the transfer.  Now, as long as permissions
      are being preserved, the shared memory should remain that way
      for the entire transfer.

    - Changed hardlink info and file_struct + strings to use allocation
      pools.  This reduces memory use for large file-sets and permits
      freeing memory to the OS.  (J.W. Schultz) 

    - The 2 pipes used between the receiver and generator processes
      (which are forked on the same machine) were reduced to 1 pipe and
      the protocol improved so that (1) it is now impossible to have the
      "redo" pipe fill up and hang rsync, and (2) trailing messages from
      the receiver don't get lost on their way through the generator
      over to the sender (which mainly affected hard-link messages and
      verbose --stats output).

    - Improved the internal uid/gid code to be more portable and a
      little more optimized.

    - The device numbers sent when using --devices are now sent as
      separate major/minor values with 32-bit accuracy (protocol 28).
      Previously, the copied devices were sent as a single 32-bit
      number.  This will make inter-operation of 64-bit binaries more
      compatible with their 32-bit brethren (with both ends of the
      connection are using protocol 28).  Note that optimizations in the
      binary protocol for sending the device numbers often results in
      fewer bytes being used than before, even though more precision is
      now available.

    - Some cleanup of the exclude/include structures and its code made
      things clearer (internally), simpler, and more efficient.

    - The reading & writing of the file-list in batch-mode is now
      handled by the same code that sends & receives the list over the
      wire.  This makes it much easier to maintain.  (Note that the
      batch code is still considered to be experimental.)