Re: [PATCH 8/9 v13] difftool: teach difftool to handle directory diffs

From: David Aguilar <davvid@gmail.com>
Date: Apr 20, 2012, 07:34 UTC
Message-ID: <CAJDDKr7JtauR8sR3YC+wj60sx9DEgf87iaDwue2Cz6FzQX_Z+Q@mail.gmail.com>
In-Reply-To: <xmqqy5prv9ol.fsf@junio.mtv.corp.google.com>

On Thu, Apr 19, 2012 at 11:49 AM, Junio C Hamano <jch@google.com> wrote:

Show 17 quoted lines

> Tim Henigan <tim.henigan@gmail.com> writes:
>
>> I have a revised patch that prunes out all of the above and warns the
>> user when it does so [1].
>
> Thanks.
>
> As long as it works when the user uses "two temporary trees" mode
> without -M/-C, and it keeps working as well as before the change when
> the user uses "one invocation per matched path" mode with -M/-C, I do
> not care too deeply about how it is implemented in the script.
>
>> However, it also prunes them when difftool is called in serial diff
>> mode (i.e. non --dir-diff).
>
> I do not use difftool myself, but I would imagine that it is a grave
> regression, no?

An alternative would be teaching difftool to gracefully handle the R lines in the --raw output.

Simply creating the files as they existed on each side is a reasonable start. It seems better form to handle all possible output from --raw anyways. It's better than pretending it doesn't exist, I think.

Doing "smart" things with the rename information is hairy so it's certainly worth leaving that for another day.

-- 
David