Re: [Bug-tar] segfault getting cwd with --listed-incremental

[Joerg Schilling]

Paul Eggert <address@hidden> wrote:

> > Could you explain why you believe this is relevant?
> > My impression is that an archive from a/b a/dot/../b cannot
> > be restored correctly in case that the symlink dot is missing from
> > the archive or even archived only after the named two files.
>
> It was just a toy example.  But I think the idea is that a/b a/dot/../b
> are both like mount points, i.e., they are both directories that are assumed 
> to
> be present during both a dump and a corresponding restore.

OK, but I believe it is best to avoid this at all in incrementals.

> > I am still sure that the decision for star (to require "-C dir .")
> > for a "dump" is a useful decision that helps to prevent archives
> > that cannot be restored.
>
> This sounds like a tradeoff between ease-of-use and idiot-proofing,
> which is a design space with more than one reasonable solution.

I am not convinced that there is a need to have a single incremental
that spawn more than one filesystem. Star did have thousands of successful
real life incremental dump/restore cycles cince May 2005 (after the last
bug was fixed) with man gigabytes of data for each incremental. I believe the 
idea of "make if simple and correct" was a good decision.

Jörg

-- 
 EMail:address@hidden (home) Jörg Schilling D-13353 Berlin
       address@hidden                (uni)  
       address@hidden (work) Blog: http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily