From: Jeff King <peff@peff.net>
Date: Wed, 18 Jun 2008 03:30:10 GMT
Subject: Re: [PATCH 1/2] parse_options: Add flag to prevent errors for further processing
Message-ID: <20080618033010.GA19657@sigill.intra.peff.net>
In-Reply-To: <7v1w2v2zsh.fsf@gitster.siamese.dyndns.org>

On Tue, Jun 17, 2008 at 08:21:50PM -0700, Junio C Hamano wrote:

> Shawn Bohrer <shawn.bohrer@gmail.com> writes:
> 
> > This adds the PARSE_OPT_NO_ERROR_ON_UNKNOWN flag which prevents
> > parse_options() from erroring out when it finds an unknown option,
> > and leaves the original command and unknown options in argv.
> 
> I have to say that this conceptually is broken.  How would you tell
> without knowing what "--flag" is if the thing in argv[] after that is a
> parameter to that option or the end of the options?

Agreed. I was just about to write the same thing. As it happens, I think
in the case of git-shortlog that there is not likely to be such a
parameter. The only three I see looking over setup_revisions are "-n"
(which is masked by shortlog anyway), "--default", and "-U" (which one
would never need with shortlog).

However I am still opposed to the concept, since its presence as a
parseopt flag implies that it isn't fundamentally broken.

I think the only right way to accomplish this is to convert the revision
and diff parameters into a parseopt-understandable format.

-Peff