[Dar-news] dar release 2.7.18 is available

Hi,

source package and windows binary for release 2.7.18 are available at:

https://sourceforge.net/projects/dar/files/dar/2.7.18

As an alternative to Sourceforge you can fetch source and windows
binary packages but also fully featured dar_static binaries for Linux/x86_64
from the _low bandwidth_ ftp and https servers at:

ftp://ftp.edrusb.org/dar.linux.free.fr/Releases
https://dar.edrusb.org/dar.linux.free.fr/Releases

Package signatures are available at:

http://dar.linux.free.fr/sign/v2.7.18

Author's keys are available on public key servers and on dar website:

http://dar.linux.free.fr/doc/authentification.html

Release 2.7.18 brings several bug fixes which are detailed in the extract
of the Changelog below.

Thanks to mannio on dar-support mailing-list for bug report


Cheers,
Denis Corbin

from 2.7.17 to 2.7.18
- fixed bug that broke the ability to repair an archive through ftp/sftp
  (repairing stays possible locally on the machine where resides the 
backup)
- fixed the reading of archive header (in-place field) when the -R path
  given at creation time was valid or common unix path (/some//path for
  example).
- fixed bug met when reading compressed data with multiple threads (-G
  option with argument 2 or larger) in sequential read mode.
- fixed bug avoiding dar to repair archives relying on block compression,
  testing the "repaired" archive lead to reporting corrupted data for all
  compressed files and compressed EA.
- disabling min-digits checks for auxiliary and reference backups when
  they are stored in a remote repository (these checks was avoiding using
  remote auxiliary and reference backups).
- fixing bug in libcurl adaptation layer to libdar (fichier_libcurl) where
  while reading, the end of file was not detected properly in very rare
  context, leading the encryption layer trying (and failing) to decipher
  the clear data located at end of archive.
- fixed bug met when reading a ciphered backup in sequential read mode with
  the help of an external catalog (the few first data and EA were reported
  as corrupted in sequential read mode, not in direct read mode or without
  the help of an external catalog).
- fixing bug in non-regression test routines, which, once fixed, revealed
  the following bugs (about sequential reading mode):
- reading a encrypted and sliced archive in sequential-read mode lead dar
  asking for the an after-last non existing slice at the end of the
  operation.
- comparing an archive having binary delta patch for a file failed in
  sequential read mode with self reporting bug in cat_file.cpp line 1479
- fixing multi-threaded deciphering when reaching the end of the archive
  and if a skip() instruction is given from the master thread to change
  reading position (bug type message at parallel_tronconneuse.cpp line 662
  was reported).
- fixing bug in archive class implementation met when comparing an archive
  with filesystem in sequential read mode with the help of an isolated
  catalogue. In such context, the hard linked inode can only be compared
  at the first hard link encountered, other occurrence are skipped. The
  isolated catalog gave a wrong information as it was read in direct mode
  while fetching the data and EA to compare were read in sequential mode
  without the ability to skip back to re-read a hard linked inode, this
  caused the second time comparison to be reported as failed instead of
  being flagged as skipped.
- fixing bug met when restoring gpg signed archive in sequential read
  mode, where dar/libdar reported failed restoring EA and/or FSA for some
  directories.
- fixing bug met when reading archive summary (-l and -q options) in
  sequential read mode on sliced archive (was asking for the last slice
  due to skip at end of archive while in sequential read mode)