Re: Bug in dar version 2.7.8 (?)

Thomas <dar.support.ml-202308-Tv/[email protected]> Mon, 4 Sep 2023 20:34:53 +0200
Newsgroups gmane.comp.sysutils.backup.dar.support
Message-ID <[email protected]>
Hi Andrea!

On Mon, Sep 04, 2023 at 05:29:36PM +0200, Andrea Vai wrote:
>   today I find this output every time my scheduled differential backup
> runs, after I took a Normal backup this morning (as I usually do the
> first days of each month):
> 
> ************************************************************************************************
> ---- exception type = [BUG] ----------
> 	File slice_layout.cpp line 48 : it seems to be a bug here
> -----------------------------------
> ************************************************************************************************

This is a known problem. Denis is locking into this but at the moment we
are stuck. For more details see:
<https://sourceforge.net/p/dar/mailman/message/37888283/> and following.

> The command I run is:
> dar -c /run/media/andrea/BAK_ANDVAI/root_diff_Ore -R / -B /home/andrea/myProgs/Backup/esclusiDiffRoot.dcf -P home/andrea/Backup/backuplog_root_Ore.txt -D -A /home/andrea/Backup/CAT_root -w -v 2>&1

I would guess your reference catalogue /home/andrea/Backup/CAT_root was
done with onfly-isolation. This seems broken with a couple of
g++-compiler combinations. The file slice_layout.cpp gets optimized in a
way that breaks the onfly-isolation function. As I mentioned earlier we
are stuck and looking for a compiler expert to solve the issue or at
least to clearify if this is a dar issue or a g++ issue.

> Dar version:
> $ dar --version
> 
>  dar version 2.7.8, Copyright (C) 2002-2022 Denis Corbin
>  compiled the Dec 28 2022 with GNUC version 12.2.1 20221121 (Red Hat 12.2.1-4)
> ************************************************************************************************

Yes with g++ 12.x.x I definitly have the same problem.

> I am not sure whether the version I used to take the normal backup is
> the same or not. If it's useful I can find it out.
> Is it really a bug, can I do anything to help or am I missing something?

You have now a couple of options/workarounds:
* You create a new catalogue from your last full backup - see option
  --aux in the dar manual. Use this newly created catalogue instead of
  /home/andrea/Backup/CAT_root in your following diff backups.
* You compile DAR yourself with my workaround patch appended here:
  <https://sourceforge.net/p/dar/mailman/message/37890758/>
* Use a DAR compiled with g++ 10.x or below.

Hope this helps to get your backup running again.

Tom