Re: diff: --quiet does not imply --exit-code if --diff-filter is present

From: Jeff King <peff@peff.net>
Date: May 31, 2011, 16:25 UTC
Message-ID: <20110531162546.GA11321@sigill.intra.peff.net>
In-Reply-To: <7vvcwqkh4a.fsf@alter.siamese.dyndns.org>

On Tue, May 31, 2011 at 08:46:29AM -0700, Junio C Hamano wrote:

Show 10 quoted lines

> Jeff King <peff@peff.net> writes:
> 
> > Commit 2cfe8a6 (diff --quiet: disable optimization when
> > --diff-filter=X is used, 2011-03-16) fixes this in some
> > cases by disabling the optimization when a filter is
> > present. However, it only tweaked run_diff_files, missing
> > the similar case in diff_tree. Thus the fix worked only for
> > diffing the working tree and index, but not between trees.
> 
> Thanks; a natural question is if we need the same for diff-index, then.

No, it calls straight into unpack_trees. So I think the question is "should unpack_trees respect the QUICK optimization". I suspect it didn't happen simply because unpack_trees is so complex, and there are probably corner cases with merging.

Show 11 quoted lines

> We probably want to have a helper in diff.c that does
> 
> 	int diff_can_quit_early(struct diff_options *opt)
>         {
>         	return (DIFF_OPT_TST(opt, QUICK) &&
>                 	!opt->filter &&
>                         DIFF_OPT_TST(opt, HAS_CHANGES));
> 	}
> 
> It is possible for us to later add new diffcore transformations that need
> a similar "do not stop feeding early, as results may be filtered".

Yeah, that is a good refactoring. It's more readable, and it would have prevented this bug. :)

-Peff