Re: [PATCH] Add gitattributes file making whitespace checking pickier

From: Daniel Barkalow <barkalow@iabervon.org>
Date: Feb 10, 2008, 22:34 UTC
Message-ID: <alpine.LNX.1.00.0802101724170.13593@iabervon.org>
In-Reply-To: <7v8x1sfprd.fsf@gitster.siamese.dyndns.org>

On Sun, 10 Feb 2008, Junio C Hamano wrote:

Show 16 quoted lines

> Daniel Barkalow <barkalow@iabervon.org> writes:
> 
> > On Sun, 10 Feb 2008, Junio C Hamano wrote:
> >
> >> I've tried "unexpand | expand" to Documentation/*.txt and
> >> compared the formatted documentation before and after the
> >> change, and as we suspected everything seems to match.
> >> 
> >> So I am considering applying this patch.  We may want to tighten
> >> it later but as the initial set of rules this should do.
> >
> > I think it's worthwhile to note that some of the files under t/ have 
> > whitespace that doesn't conform to the rest of the project.
> 
> Yes.  I tried to be careful with the set of attribute files I
> sent out not to corrupt them.  Was I unsuccessful?

Your patch wouldn't corrupt them, but also didn't explicitly protect them. A later change to regularize git shell scripts would look reasonable but risk messing with them, for example.

Show 5 quoted lines

> > On the other hand, I think this patch is an odd combination of stuff; 
> > aren't the contrib exceptions not exceptional, since only *.[ch] and *.txt 
> > gets declared as "whitespace"?
> 
> True, how about this instead?

That makes things more clear, although I'd still like a t/.gitattributes that meant that, regardless of the project's policies in general, t/ files may care about whitespace, so we don't have to worry about that in future changes to the top level .gitattributes. And I'm actually curious as to whether t/ should have some protection against crlf conversion. (I haven't tried checking git out with different crlf-conversion settings and seeing whether the tests still work as intended)

	-Daniel
*This .sig left intentionally blank*