From: Jakub Narebski <jnareb@gmail.com>
Date: Mon, 07 Jan 2008 00:31:56 GMT
Subject: Re: [PATCH] Test "git log --diff-filter"
Message-ID: <200801070131.57722.jnareb@gmail.com>
In-Reply-To: <7vmyrj7kq5.fsf@gitster.siamese.dyndns.org>

Junio C Hamano wrote:
> Jakub Narebski <jnareb@gmail.com> writes:
> 
> > My test checks all --diff-filter filters relevant to git-diff-tree,
> > i.e. ADMRCBT, and not only AMD.
> 
> Ah, I see.  Thanks --- that could have been stated in the log
> message.  Maybe we would want to add them to existing test
> script, instead of adding a whole new one?

The test as it stands now checks if --diff-filter select appropriate
revisions, even without patch output. I think it is enough, as I don't
see how we could screw up to filter AMD correctly, and not all others...
...perhaps with exception of pair breaking, and how they are filtered
using --diff-filter=M and --diff-filter=B; but this impression might
be caused by the fact that pair breaking is the only one which doesn't
use symbol ('B') in raw diff format output.

> > P.S. By the way, it is IMHO a bit strange that --pretty=oneline uses
> > newline as a terminator (it means that there is a newline at the end of
> > "git log --pretty=oneline), while --pretty="format:%s" uses newline as
> > a separator...
> 
> Yeah, I tend to agree, although I learned to live with it long
> time ago.

IMHO that is design bug. Perhaps it should be changed? This way, at least
conceptually oneline, short, medium, full, fuller, email formats might be
considered simply pre-defined format:<sth> formats.

Am I mistaken in thinking that the rest of git always use terminators,
and not separators for records output?
-- 
Jakub Narebski
Poland