On Tue, Jul 29, 2008 at 04:42:53PM -0700, Junio C Hamano <gitster@pobox.com> wrote:
> True, but with the old code that might blindly have executed
> git-merge-index the test would also have failed, and you would want to
> tell two cases apart, wouldn't you?

Hmm, I wonder what is the right approach to test it. Should I exit with
a different error code with the strategy is invalid?