Re: [rdiff-backup-users] crash in gzip.py

[Joe Steele]

On 12/13/2015 7:03 AM, Andrea Cozzolino wrote:
> I have quite the same problem and in the *same* environment
> (openSUSE 13.2/rdiff-backup 1.2.8).
>
> I have a broken and, so far, unrecoverable backup store.
>
> Each time I try to backup my home dir, I got a "Negative seek"
> error while "Writing file object to
> /backup/storage/BackUp/tseenfoo/home-andcoz_broken/rdiff-backup-data/increments/.mozilla/firefox/c3ggw723.suse/content-prefs.sqlite.2015-11-23T02:12:37+01:00.snapshot.gz"
...
> Traceback (most recent call last):
...
>    File "/usr/lib64/python2.7/site-packages/rdiff_backup/rpath.py", line
> 97, in copyfileobj
>      outputfp.seek(-1, os.SEEK_CUR)
>    File "/usr/lib64/python2.7/gzip.py", line 423, in seek
>      raise IOError('Negative seek in write mode')
> IOError: Negative seek in write mode

The vanilla rdiff-backup 1.2.8 does not have "outputfp.seek(-1, 
os.SEEK_CUR)" on line 97 of rpath.py.

In looking at the packaging of rdiff-backup by openSUSE, I see that they 
incorporate several patches not included in the upstream source.

Their revision history can be found here:

https://build.opensuse.org/package/revisions/openSUSE:Factory/rdiff-backup

One of their changes (which seems to be the source of your problem) was 
related to the handling of sparse files and was incorporated here:

https://build.opensuse.org/request/show/222465

Apparently, that was causing somebody to have same problem that you are 
having, and another patch was added:

https://build.opensuse.org/request/show/259636

Based on the line numbers in your log, it would appear that you do not 
have that second patch which was intended to fix the problem.

In looking at the openSUSE packages 
(http://software.opensuse.org/download.html?project=Archiving&package=rdiff-backup), 
it appears that openSUSE 13.2 contains rdiff-backup-1.2.8-42.3.  It is 
also evident when looking at the source that revision 42.3 contains the 
second patch referenced above (and which you do not seem to have).  So 
that would imply that you are not using their current revision(?).

--Joe