Re: [Bug-tar] Extract error after failed open()

[Jan-Benedict Glaw]

On Thu, 2006-11-30 11:51:29 +0200, Sergey Poznyakoff <address@hidden> wrote:
> > GNU tar signals the open() error and prints a message like "skipping
> > to next header", but with my (real 16 GB) copy, "skipping to next
> > header" looked like "skipping to end of archive."  Finally, only half
> > of the data was copied.
> > 
> > Easy solution:  Call skip_member() in the error path as it is done in
> > a different case.
> 
> This will certainly fix this particular case, but the cause of the
> misbehavior will remain unknown.  I have tried to reprouce the bug using
> several crafted archives but was unable to do so.  Could you supply some
> more info on how to reproduce it?

The key point is that the open call needs to fail. Have you been able
to reproduce at least that (strace should easily show that.)

>From my understanding, even with the missing skip_member() call, tar
should be able to find the next file header at some time, but it
missed to do so in my case.

The whole filesystem I copied is about 16GB large (it was streamed
from one directory to another:

        tar cpf - . | ( cd /somewhere && tar xpf -;)

) so this isn't easy to hand out.  Nonetheless, the call to
skip_member() seems to be obviously missing, even in case you fail to
create an archive where tar fails to extract basically all of the rest
of the archive :)  The final point is that I don't have the target
filesystem (read: target HDD which was about to be restored) around
and there's no filesystem on that laptop that I could reformat with
VFAT, large enough to keep up with that amount of data...

MfG, JBG

-- 
      Jan-Benedict Glaw      address@hidden              +49-172-7608481
  Signature of:                        Lauf nicht vor Deinem Glück davon:
  the second  :                             Es könnte hinter Dir stehen!

signature.asc
Description: Digital signature