From: Jens Lehmann <jens.lehmann@web.de>
Date: Thu, 08 Aug 2013 17:11:04 GMT
Subject: Re: [PATCH v4 5/5] rm: delete .gitmodules entry of submodules removed from the work tree
Message-ID: <5203D128.7030705@web.de>
In-Reply-To: <20130807182809.GA15123@paksenarrion.iveqy.com>

Am 07.08.2013 20:28, schrieb Fredrik Gustafsson:
> On Tue, Aug 06, 2013 at 02:11:56PM -0700, Junio C Hamano wrote:
>> Thanks, will replace the top two commits and queue.  Looks like we
>> are getting ready for 'next'?
> 
> I'm a bit curious about if we should move towards a reentrent libgit
> (which would for example make multithreading easier) or not.

I'm not aware of such an effort in core Git (I always thought that
libgit2 is the project doing what you seem to aim for).

> If so, I suggest that this patch only use die() in builtin/. However I
> know that there's a lot of die() all over libgit today, I'm curious
> about what direction we're heading.

The die() calls are just one part. Global variables are another issue,
we have memory which is implicitly freed on exit ... so unless we
commit ourselves to fix all those issues I see no point in moving the
die() calls into builtin/ in my series.