From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 14 May 2008 18:28:48 GMT
Subject: Re: [PATCH 1/3] diff options: Introduce --ignore-submodules
Message-ID: <7vbq38oi3j.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <alpine.DEB.1.00.0805141803240.30431@racer>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> @@ -3355,6 +3357,9 @@ void diff_addremove(struct diff_options *options,
>  	char concatpath[PATH_MAX];
>  	struct diff_filespec *one, *two;
>  
> +	if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
> +		return;
> +
>  	/* This may look odd, but it is a preparation for
>  	 * feeding "there are unchanged files which should
>  	 * not produce diffs, but when you are doing copy

So both removal and addition of a submodule is an uninteresting event.

> @@ -3399,6 +3404,10 @@ void diff_change(struct diff_options *options,
>  	char concatpath[PATH_MAX];
>  	struct diff_filespec *one, *two;
>  
> +	if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
> +			&& S_ISGITLINK(new_mode))
> +		return;

And a submodule changing from revision A to B is uninteresting.

Is it interesting if something that used to be a blob turns into a
submodule, or vice versa?  The code says it is, but I think it would be
more convenient to treat it as a removal of blob and addition of a
submodule.