Re: [PATCH] add status.relativePaths config variable

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 8, 2007, 08:14 UTC
Message-ID: <7v1w9xy55h.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <20071208075538.GA4812@coredump.intra.peff.net>

Jeff King <peff@peff.net> writes:

Show 12 quoted lines

> On Fri, Dec 07, 2007 at 11:34:14PM -0800, Junio C Hamano wrote:
> ...
>> I have been wondering ever since receiving this patch if this is a good
>> interface change.  Was there a problem if instead:
>> 
>> 	- The implementation of wt_status_prepare(&s) stays as before;
>> 
>> 	- run_status(), after calling wt_status_prepare(&s), notices the
>>           configuration variable, and sets s.prefix conditionally;
>
> That would work fine. My reasoning was: the point of wt_status_prepare
> is to initialize the wt_status object.

Yes, just like diffopts and revs. They initialize the object to a plain vanilla defaults, and the caller uses other methods (either direct assignments to members or by calling helper functions such as diff_opt_parse() and setup_revisions()) to fill in specialized values.

And s.prefix is very much special case. That's the reasoning behind my suggestion.

> ..., which is a little silly, since as a side effect we will do a find
> on all untracked files).

Ah, that is probably the side effect of direct rewrite from shell script to C. We should drop that and replace with what Alex did recently to git-commit --no-edit codepath.

> I am fine with either; your call. Documentation patch is below.

Thanks.  Will take it.