From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 10 Oct 2012 16:52:34 GMT
Subject: Re: [PATCH v2 0/7] Tying loose ends on grep-pcre
Message-ID: <7vlifeclrh.fsf@alter.siamese.dyndns.org>
In-Reply-To: <5075919F.5030306@alum.mit.edu>

Michael Haggerty <mhagger@alum.mit.edu> writes:

>> The last one teaches "git log" family to honor the "grep.*"
>> configuration variables, e.g. "grep.patterntype", so that you can
>> say "git -c grep.patterntype=perl log --grep='(?:pcre)'".
>
> Maybe this has been discussed already, but it seems to me that adding a
> persistent setting that affects how "git log --grep" interprets the
> pattern argument could break some scripts that assume that the "old"
> interpretation is always used.  Shouldn't this at least be documented as
> a backwards incompatibility?

If somebody scripts around "log" with hardcoded query "--grep=..."
strings, they can force a particular variant from such a command
line at the same time.  But as always, responsibility of doing so is
on the person who writes such a script; "log" being a Porcelain, we
value ease-of-use in interactive case more than cast-in-stone
interface stability like we do for plumbing commands.

And that is exactly why the series avoids changing the behaviour for
the "rev-list" plumbing.