Re: [PATCH] completion: support 'git config --local'

From: Jeff King <peff@peff.net>
Date: Feb 12, 2013, 21:11 UTC
Message-ID: <20130212211140.GA29358@sigill.intra.peff.net>
In-Reply-To: <7vzjz9h1w0.fsf@alter.siamese.dyndns.org>

On Tue, Feb 12, 2013 at 09:34:39AM -0800, Junio C Hamano wrote:

Show 11 quoted lines

> I see the second hunk is new.  Comments?
> [...]
> > @@ -1676,7 +1676,7 @@ _git_config ()
> >  	case "$cur" in
> >  	--*)
> >  		__gitcomp "
> > -			--global --system --file=
> > +			--system --global --local --file=
> >  			--list --replace-all
> >  			--get --get-all --get-regexp
> >  			--add --unset --unset-all

It makes sense to me. It just means that "--local" itself gets completed (while the other hunk is about using the presence of "--local" impacting other completion). It's an orthogonal issue, but I don't mind them in the same patch.

Show 5 quoted lines

> How would this interract with the writing side of "git config"?
> "git config --local foo.bar value" and "git config foo.bar value"
> are the same, no?
> 
> Is it "yes they are the same but it does not hurt?"

It doesn't affect writing at all. The change is in __git_config_get_set_variables, which is used only here:

  --get|--get-all|--unset|--unset-all)
        __gitcomp_nl "$(__git_config_get_set_variables)"

So it is purely about completing existing variables, and it's right to limit itself to a particular file if we know that is what has been given.

I'm not sure I understand the original poster's point about "git config -l --local". "-l" does not take a limiter, does it?

-Peff