Re: [PATCH] Add an optional argument for --color options

From: Jeff King <peff@peff.net>
Date: Feb 15, 2010, 05:21 UTC
Message-ID: <20100215052139.GH3336@coredump.intra.peff.net>
In-Reply-To: <ca433831002140658r30aa539fy5480cae8298d6d6c@mail.gmail.com>

On Sun, Feb 14, 2010 at 09:58:58AM -0500, Mark Lodato wrote:

Show 11 quoted lines

> > Hmm...the only one I see that silently ignores is "--submodule=bogus".
> > But it seems that "git log -Bfoobar" fails but does not print a useful
> > message.  Probably both should be fixed, and your option should follow
> > the same convention as those.
> 
> Just wondering, why does diff use a separate option parsing mechanism
> than the rest of the code?  Would it be worthwhile to switch to
> parse_opt?  This may make the code cleaner, and it would definitely
> make the command-line interface more consistent with the rest of the
> suite.  From a user's point of view, the biggest win would be "-h"
> printing all of the options, like all the non-diff commands do.

It's historical. The diff option parser predates parse-options by quite a bit, and was never converted. Pierre made some attempts at converting it and the revision parser some time back, and we ended up with the more iterative approach (you can step through each argument with parse-options, and then alternatively feed it to the revision and diff options parser).

I don't remember if there were any technical limitations, though (e.g., places where the revision parser does not conform to parse-options standards or needs some special treatment). You'd have to search the list archives to see what actually happened.

-Peff