Rdiff-backup

Latest version: v2.4.0

Safety actively analyzes 638741 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 12 of 19

0.11.1

*Warning* Various features have been removed from this version, so this is not a safe upgrade.
Also this version has less error checking, and, if it crashes, this version may be more prone to leave the destination directory in an inconsistent state.
I plan to look at these issues in the next version.
Also, this version is quite different from previous ones, so you cannot run version 0.11.1 on one end of a connection and any previous version on the other side.

The following features have been removed:

....
--mirror-only option: If you just want to mirror something, use
rsync. (Or you could use rdiff-backup and then just delete the
rdiff-backup-data directory, and then update the root mtime.)

--change-source-perms option: This feature was pretty complicated
to implement, and if something happened to rdiff-backup during a
transfer, the old permissions could not be restored.

All "resume" related functionality, like --checkpoint-interval:
This was complicated to implement, and didn't seem to work all
that well.

Directory statistics file: Although the session statistics file is
still generated, the directory statistics file no longer is,
because the new code structure makes it less inconvenient.

The various --exclude and --include options no longer work when
restoring. This may be added later if there is demand.

--windows-mode and filename quoting doesn't work. There have been
several requests for this in the past, so it will probably be
re-added in the next version.
....

Extensive refactoring.
A lot of rdiff-backup's code was structured as if it were still in one file, so it didn't make enough use of Python's module system.

Now rdiff-backup writes metadata (uid, gid, mtime, etc.) to a compressed text file in the rdiff-backup-data directory.
Here are some ramifications:

....
A user does not need root access on the destination side to record
file ownership information.

Some files may be recognized as not having changed based on this
metadata, so it may not be necessary to traverse the whole mirror
directory. This can reduce file access on the destination side.

Even when the --no-hard-links option is given when backing up,
link relationships can be restored properly. However, if this
option is given, mirror files will not be linked together.

Special file types like device and sockets which cannot be created
on the remote side for some reason can still be backed up and
restored properly.
....

Fixed bug with the --{include|exclude}-globbing-filelist options (reported by Claus Herwig).

Added --list-changed-since option to list the files changed since the given date, and added Bud Bruegger's patch to that.
The format and information this option provides will probably change in the near future.

Restoring is now pipelined for better high latency performance, and unchanged files in the target directory will not be recopied.

0.11.0

If get a socket error from trying to create a socket whose name is too long, just skip file instead of exiting with error (bug report by Ivo De Decker).

Added --exclude-special-files switch, which excludes fifos, symlinks, sockets, and device files.

--windows-mode is now short for --windows-time-format --chars-to-quote A-Z: --no-hard-links --exclude-special-files.
Thanks to Paul-Erik Törrönen for some helpful windows info.

Multiple --include and --exclude statements can now be given in a single file.
See the documentation on --{include|exclude}-globbing-filelist.
Thanks to Henrik Lewander for pointing out that command line length could otherwise be a problem.

Fixed bug in filelist processing that ignored leading or trailing whitespace in filelists.
Now filenames with, for instance, trailing spaces can be used in filelists.
Filelists which took advantage of this bug for formatting may have to be edited.

Applied major/minor patch contributed by David S.
rdiff-backup should now correctly copy device files on platforms such as NetBSD.

It is now possible to restore from a read-only filesystem (before rdiff-backup would fail when trying to open log file).
Thanks to Gregor Zattler for bug report.

Fixed bug that prevented certain restores when the source directory was specified with a trailing backslash.

Added a bit more logging so it should be apparent which file was being processed when an error occurs (thanks to Gerd Knops for suggestion).

Fixed bug when using --chars-to-quote and directory deleted that has quoted characters in it.

0.10.1

rdiff-backup should now correctly handle files larger than 2GB.
Thanks to Russ Allbery for telling me how to do this.

0.10.0

Fixed bug, probably introduced in 0.9.3, which prevented restores from a local source to a remote destination.
Reported by Phillip Eby.

Fixed another bug reported by Phillip Eby, where restores would fail if rdiff-backup had only been run once and no increments were available.

A few man page additions regarding restoring, statistics, and --test-server (thanks to Gregor Zattler, Christopher Schanzle, and Tobias Polzin for suggestions).

Fixed comparison bug where rdiff-backup would unnecessarily report a directory as changed when its source size differed from its mirror size.
Thanks to Tim Allen for report.

0.9.5

Fixed --verbosity option (now both -v and --verbosity work).
Thanks to Chris Dumont for report.

**** IMPORTANT **** Fixed serious permissions bug found by Robert Weber.
Previous versions in the 0.9.x branch would throw away high bit permissions (like the setuid and setuid bits).
This would be especially bad when running with the --change-source-perms operation.
Anyone running 0.9.0 - 0.9.4 should upgrade immediately.

Complain about --change-source-perms when running as root, as this option should not be necessary then.

Fixed bug with --windows-mode.
Thanks to Chris Grindstaff for report.

0.9.4

Man page now correctly included in rpm.

To prevent confusion, rdiff-backup script does not have exec permissions until it is installed (thanks Jason Piterak).

Sockets are now replicated.
Why not?
(Suggestion by Mickey Everts)

Bad resuming information (because, say, it is left over from a previous version) should no longer cause exit, except when --resume is specified.

Better error handling in certain cases when errors occur in file reads (thanks to John Goerzen for report).

Page 12 of 19

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.