From: Elijah Newren <newren@gmail.com>
Date: Wed, 22 Sep 2010 04:41:17 GMT
Subject: Re: [PATCH 09/37] t6020: Modernize style a bit
Message-ID: <AANLkTintJNQKq9Tg0D9g0_Qz9u62mWWR+Pz5OS2f_p2+@mail.gmail.com>
In-Reply-To: <7v7hietuvf.fsf@alter.siamese.dyndns.org>

On Tue, Sep 21, 2010 at 7:44 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Elijah Newren <newren@gmail.com> writes:
>
>>> The old version requested a particular kind of failure. Are you saying
>>> that with modern 'git merge' all non-zero exit codes mean the same kind of
>>> failure?
>>
>> No, I'm saying that I don't think the test meant to try to distinguish
>> between exit codes.
>
> The check for status 1 comes from 72d1216 (New test case: merge with
> directory/file conflicts, 2005-12-03), and if you read git-merge.sh from
> that era, you would notice that at least we do not want to see status 2
> (merge wrapper failed in a bad way) from this test.  git-merge was
> designed to exit with status 1 when (and only when) the merge request was
> reasonable, it did its best effort and left conflict to be resolved in the
> working tree.

Indeed; looks like I didn't check closely enough.