From: Jeff King <peff@peff.net>
Date: Mon, 25 Mar 2013 21:26:05 GMT
Subject: Re: [PATCH 5/9] add test for streaming corrupt blobs
Message-ID: <20130325212605.GA19303@sigill.intra.peff.net>
In-Reply-To: <20130325211038.GD1414@google.com>

On Mon, Mar 25, 2013 at 02:10:38PM -0700, Jonathan Nieder wrote:

> > +# convert "1234abcd" to ".git/objects/12/34abcd"
> > +obj_to_file() {
> > +	echo "$(git rev-parse --git-dir)/objects/$(git rev-parse "$1" | sed 's,..,&/,')"
> > +}
> 
> Maybe this would be clearer in multiple lines?
> 
> 	commit=$(git rev-parse --verify "$1") &&
> 	git_dir=$(git rev-parse --git-dir) &&
> 	tail=${commit#??} &&
> 	echo "$git_dir/objects/${commit%$tail}/$tail"

Yeah, it started as:

  echo "$1" | sed 's,..,&/,'

and kind of got out of hand as it grew features. I'd be fine with your
version (though $commit is not right, as it is any object, and in fact
the test uses blobs).

> > +
> > +# Convert byte at offset "$2" of object "$1" into '\0'
> > +corrupt_byte() {
> > +	obj_file=$(obj_to_file "$1") &&
> > +	chmod +w "$obj_file" &&
> > +	printf '\0' | dd of="$obj_file" bs=1 seek="$2"
> 
> Some other tests such as t4205 also rely on "printf" being
> binary-safe.  Phew.

Yeah, I think it should be fine, though the choice of character does not
actually matter, as long as it is different from what is currently at
that position (for the sake of simplicity, I just determined
experimentally that the given object is corrupted with the offset and
character I chose).

-Peff