From: Jeff King <peff@peff.net>
Date: Fri, 14 Jun 2013 22:19:36 GMT
Subject: Re: [PATCH 2/2] unpack_entry: do not die when we fail to apply a delta
Message-ID: <20130614221936.GD29138@sigill.intra.peff.net>
In-Reply-To: <7vk3lwmk0r.fsf@alter.siamese.dyndns.org>

On Fri, Jun 14, 2013 at 02:59:00PM -0700, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> >  test_expect_success \
> > +    'corruption of delta base reference pointing to wrong object' \
> > +    'create_new_pack --delta-base-offset &&
> > +     git prune-packed &&
> > +     printf "\220\033" | do_corrupt_object $blob_3 2 &&
> 
> Interesting.  You cheated in a different way with a hardcoded
> offset, instead of hardcoded knowledge of where the object name
> is stored in binary in the .idx file ;-)

Yes. We could get it with:

  git show-index <"$pack.idx" |
  cut -d' ' -f1 |
  perl -e '
    @pos = map { chomp; $_ } <>;
    my $ofs = $pos[2] - $pos[0];

    my @bin;
    unshift @bin, $ofs & 127;
    while ($ofs >>= 7) {
      $ofs--;
      unshift @bin, 128 | ($ofs & 127);
    }

    binmode STDOUT;
    print chr for @bin;
  '

if that's not too ugly. Maybe the REF_DELTA one is less ugly, then, as
it would not need to do the packed offset encoding, but just convert
$blob1 from hex into binary.

-Peff