Re: rescue command segv's always

[Ryan Charles Underwood]

> > No, in fact with this patch exactly the same behavior is shown,
> > dereferencing NULL in the same library function.  I did verify that the
> > library and the parted binary were rebuilt from scratch with the updated
> > codes.
> 
> This is weird, I can't reproduce the bug with the patch...

Just to check, I threw away the whole build tree and started from scratch
(with patch).  Same thing happened.  Then rebuilding static worked.

> > This time I built with CFLAGS=-g, --enable-debug and --enable-all-static
> > in the hopes that Valgrind would give me the line number of the error in
> > the shared library.  [Un]fortunately, this time it worked, but with many
> > (!!) memory errors noted.
> 
> Are you sure they really are memory errors ? Valgrind uses to report lots
> of ioctl errors that indeed are not really errors when I use it to debug 
> Parted :p

In this case, not heap errors, but references into uninitialized memory.  I
know the spurious errors you refer to though.

> > Unfortunately the scan does not find my ext3 partition.  It was simply 
> > deleted with fdisk, not overwritten or anything, I don't see what the 
> > matter is.  I have used parted rescue before with success.
> 
> I reproduced your partition scheme, with a supplementary partition 10
> from 41 to 49g, formatted everybody with ext3, deleted it and rescued it
> without any problem, but well this test doesn't mean anything because
> it could be different from the exact position of your partition 10

Very strange.  What exactly could I do to manually scan for an ext3
partition?  I don't care much about the false positives, I just wonder
if the FS header has disappeared or been scribbled somehow.  That would
be a terrible loss if so.

Thanks,
Ryan