Re: What's cooking in git.git (Mar 2011, #02; Sun, 20)

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 21, 2011, 14:28 UTC
Message-ID: <7vfwqgtunj.fsf@alter.siamese.dyndns.org>
In-Reply-To: <20110321111643.GE16334@sigill.intra.peff.net>

Jeff King <peff@peff.net> writes:

Show 21 quoted lines

>> Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:
>> 
>> > Junio C Hamano <gitster@pobox.com> writes:
>> >
>> >> Unfortunately
>> >> we do not have a good way to let people who have already heard
>> >> about the transition to squelch it yet.
>> >
>> > How about an advice.addWithoutPath config variable to disable it?
>> 
>> I don't think it is worth it; I'd rather see :/ magic implemented earlier.
>> 
>> Another possibility is to error out on "git add -u" without pathspec in a
>> subdirectory as an error in 1.8.0, and then make it full-tree in 1.8.X in
>> a distant future, but that wouldn't be as pleasant.
>
> I thought we cared about giving fans of the current behavior a way to
> keep it (without having to use additional syntax). Script-writers
> already need to use "./" to be portable to old versions of git, so it's
> not hurting them (unless you want to count N years into the future, when
> we can discount such old versions).

So you would rather have "addu.relativetocwd" configuration variable that will warn when unset but the behaviour defaults to true for now, and later flip to false?

I'd be in favor of that (the variable name can be painted in any color). The patch series was an illustration of the logical conclusion of Duy's patch that punted on the migration path, and if it showed it would not be pleasant, it served its purpose ;-).