Without this, under various conditions, (including file_name being
a directory), maybe_backup_file would return true without having
done anything, nor having set after_backup_name. Then, if any
error was detected while extracting the member, (such as directory
already exists), the previously backed-up file would be restored,
potentially destroying user data.
This closes Debian bug #508199, with thanks to Eric Lammerts for
reporting it with a concise test case.
Carl Worth [Tue, 4 Aug 2009 19:20:05 +0000 (12:20 -0700)]
Drop undefined B macro from tar man page, (fixes lintian warning)
This undefined macro was causing misformatting of the first line
of the man page:
The first argument to should be...
rather than:
The first argument to tar should be...
Presumably just dropping it misses out on some intending formatting,
(emboldening of the word tar perhaps?), but this is at least better
than the word being missing.
* src/create.c (dump_hard_link): Always look up in the link table
if remove_files_option is set. Patch suggested by Theodore Ts'o
<tytso@mit.edu>.
(check_links): Remove extra newline from the warning message.
* tests/link02.at, tests/link03.at: New testcases.
* tests/Makefile.am (TESTSUITE_AT): Add link02.at and link03.at
* tests/testsuite.at: Include link02.at and link03.at
The idea behind these patches was just fine, but upstream just
accepted a slightly modified version, so I'm reverting mine in
favor of that from upstream.
Carl Worth [Thu, 30 Jul 2009 03:45:58 +0000 (20:45 -0700)]
Preserve hard links with --remove-files
When the --remove-files option is in effect, it is no longer
reliable to use a file's link count to determine if we should
use the hash table for hard links. Instead, we look into the
hash table for every file when under the influence of the
--remove-files option.
Carl Worth [Wed, 29 Jul 2009 23:10:01 +0000 (16:10 -0700)]
Don't close file stream before EOF, closes #525818
To fix bug #235820 src/list.c was modified to avoid printing
a confusing warning message about a lone zero block. However,
the change also caused the input stream to be closed before
reading EOF which could cause a SIGPIPE to be sent to the
writing process.
In this change, the previous fix is made more narrow to exclude
only the printing of the warning message, but without the side
effect that was causing the SIGPIPE.
projects / tar / log