-NEWS for rsync 2.6.9 (unreleased)
-Protocol: 29 (unchanged)
-Changes since 2.6.8:
+NEWS for rsync 3.0.0 (UNRELEASED)
+Protocol: 30 (changed)
+Changes since 2.6.9:
BUG FIXES:
- - An rsync daemon that is receiving files with "use chroot = no" no longer
- sanitizes the symlink target strings. This means that each symlink's
- value will now be accepted (and thus returned) with its symlink info
- intact. Also, in order to keep things safe, any option that tells a
- non-chroot daemon to treat some symlinks as their referent (such as
- --copy-links or --keep-dirlinks) now manually checks the symlink chain
- to ensure that the symlinks do not try to escape past the top of the
- module's path. Both these changes make a non-chroot daemon behave the
- same as a chroot daemon with regard to symlinks, and also avoids a
- potential problem where a pre-existing symlink could have escaped the
- module's hierarchy.
-
- - Fixed a overzealous sanitizing bug in the handling of the --*-dest
- options (--link-dest, --copy-dest, and --compare-dest): if the copy's
- destination dir is deeper than the top of the module's path, these
- options now accept a safe number of ../ (parent-dir) references (since
- these options are relative to the destination dir). The old code
- incorrectly chopped off all "../" prefixes for these options, no matter
- how deep the destination directory was in the module's hierarchy.
-
- - Fixed a bug where a deferred info/error/log message could get sent
- directly to the sender instead of being handled by rwrite() in the
- generator. This fixes an "unexpected tag 3" fatal error, and should
- also fix a potential problem where a deferred info/error message from
- the receiver might bypass the log file and get sent only to the client
- process. (These problems could only affect an rsync daemon that was
- receiving files.)
-
- - Make sure that the --link-dest option can still do its job even when -I
- or --size-only is specified.
-
- - The daemon now calls more timezone-using functions prior to doing a
- chroot. This should help some C libraries to generate proper timestamps
- from inside a chrooted daemon (and to not try to access /etc/timezone
- over and over again).
+ - Fixed the output of -ii when combined with one of the --*-dest options:
+ it now itemizes all the items, not just the changed ones.
+
+ - Made the output of all file types consistent when using a --*-dest
+ option. Prior versions used to output too many creation events for
+ matching items.
+
+ - The code that waits for a child pid now handles being interrupted by a
+ signal. This fixes a problem with the pre-xfer exec function not being
+ able to get the exit status from the script.
+
+ - A negated filter rule now sends the negation option when sending the
+ filter rules.
+
+ - Fixed a problem with the --out-format (aka --log-format) option: it
+ would output superfluous directory information for a non-daemon rsync.
+
+ - Fixed a problem with -vv (double --verbose) and --stats when "pushing"
+ files (which includes local copies). Version 2.6.9 would complete the
+ copy, but exit with an error when the receiver output its memory stats.
+
+ - If --password-file is used on a non-daemon transfer, rsync now complains
+ and exits. This should help users figure out that they can't use this
+ option to control a remote shell's password prompt.
+
+ - Make sure that directory permissions of a newly-created destination
+ directory are handled right when --perms is left off.
ENHANCEMENTS:
- - ...
+ - A new incremental-recursion algorithm is now used when rsync is talking
+ to another 3.0.0 version. This starts the transfer going more quickly
+ (before all the files have been found), and requires much less memory.
+ See the --recursive option in the manpage for some restrictions.
- DEVELOPER RELATED:
+ - The default --delete algorithm is now --delete-during when talking to a
+ 3.x rsync. This is a faster scan than using --delete-before (which is
+ the default when talking to older rsync versions), and is compatible with
+ the new incremental recursion mode.
+
+ - Added the --delete-delay option, which is a more efficient way to delete
+ files at the end of the transfer without needing a separate delete pass.
+
+ - Added the --acls (-A) option to preserve Access Control Lists. This is
+ an improved version of the prior patch that was available. (If you need
+ to have backward compatibility with old, patched versions, the new
+ acls.diff patch that will add that.)
+
+ - Added the --xattrs (-X) option to preserver extended attributes. This is
+ an improved version of the prior patch that was available. (If you need
+ to have backward compatibility with old, patched versions, the new
+ xattrs.diff patch that will add that.)
+
+ - Added the --fake-super option that allows a non-super user to preserve
+ all attributes of a file by using a special extended-attribute idiom.
+ There is also an analogous "fake super" option for an rsync daemon.
+
+ - Added the --iconv option, which allows rsync to convert filenames from
+ one character-set to another during the transfer. The default is to make
+ this feature available as long as your system has iconv_open(). If
+ compilation fails, specify --disable-iconv to configure and rebuild. If
+ you want rsync to perform character-set conversions by default, you can
+ specify --enable-iconv=CONVERT_STRING with the default value for the
+ --iconv option that you wish to use. For example, --enable-iconv=. is a
+ good choice. See the rsync man page for an explanation of the --iconv
+ option's settings.
+
+ - You may specify --max-delete=0 to a 3.0.0 client as long as the receiving
+ side is at least version 3.0.0. This means that you can pull from an
+ older rsync with this option, but pushing to an older rsync will generate
+ an error. *Be sure to never specify a 0 value to an older rsync client,
+ or it will be silently ignored.*
- - The acls.diff and xattrs.diff patches have received a bunch of work to
- make them much closer to being acceptable in the main distribution.
+ - The --hard-link option now uses less memory on both the sending and
+ receiving side for all protocol versions. For protocol 30, the use of a
+ hashtable on the sending side allows us to more efficiently convey to the
+ receiver what files are linked together. This reduces the amount of data
+ sent over the socket by a considerable margin (rather than adding more
+ data), and limits the in-memory storage of the device+inode information
+ to just the sending side for the new protocol 30, or to the receiving
+ side when speaking an older protocol (note that older rsync versions kept
+ the device+inode information on both sides).
- - Updated config.guess and config.sub to their 2006-02-23 version.
+ - A lot more --no-OPTION override options are now supported.
+
+ INTERNAL:
+
+ - The file-list sorting algorithm now uses a sort that keeps any same-
+ named items in the same order as they were specified. This allows
+ rsync to always ensure that the first of the duplicates is the one
+ that will be included in the copy. The new sort was also faster than
+ the GNU C version of qsort() (and mergesort() too) in my testing.
+
+ - Rsync now supports the transfer of 64-bit timestamps (time_t values).
+
+ - Fixed a build problem with older (2.x) versions of gcc.
+
+ - Added some isType() functions that make dealing with signed characters
+ easier without forcing variables via casts.
+
+ - Upgraded the included popt version to 1.10.2 and improved its use of
+ string-handling functions.
+
+ - Added missing prototypes for compatibility functions from the lib dir.
+
+ - Configure determines if iconv() has a const arg, allowing us to avoid a
+ compiler warning.
+
+ - Made the sending of some numbers more efficient for protocol 30.
+
+ - Make sure that a daemon process doesn't mind if the client was weird and
+ omitted the --server option.
+
+ - Improved the use of "const" on pointers.
+
+ DEVELOPER RELATED:
- - Updated various files to include the latest FSF address and to have
- consistent opening comments.
+ - Rsync now has a way of handling protocol-version changes during the
+ development of a new protocol version. This exchange of sub-version
+ info does not interfere with the {MIN,MAX}_PROTOCOL_VERSION checking
+ in older versions (which would be the case if every minor change made
+ to the protocol in CVS incremented the main PROTOCOL_VERSION value).