From: Jonathan Nieder <jrnieder@gmail.com>
Date: Tue, 12 Feb 2013 19:56:20 GMT
Subject: Re: [PATCH/FYI v4 13/12] fixup! t/t3511: add some tests of 'cherry-pick -s' functionality
Message-ID: <20130212195620.GB12240@google.com>
In-Reply-To: <1360664260-11803-14-git-send-email-drafnel@gmail.com>

Brandon Casey wrote:

> I'm not sure we should apply this though.  I'm leaning towards saying that
> the 'cherry-pick -s' behavior with respect to a commit with an empty message
> body should be undefined.  If we want it to be undefined then we probably
> shouldn't introduce a test which would have the effect of defining it.

Maybe it would make sense to just check that cherry-pick doesn't
segfault in this case?

That is, compute the output but don't compare it to expected output, as
in:

	test_expect_success 'adding signoff to empty message does something sane' '
		git reset --hard HEAD^ &&
		git cherry-pick --allow-empty-message -s empty-branch &&
		git show --pretty=format:%B -s empty-branch >actual &&

		# sign-off is included *somewhere*
		grep "^Signed-off-by:.*>\$" actual
	'

Alternatively, if there are only a few sane behaviors, a test can check
for all of them and pass as long as git follows one.  I haven't thought
carefully enough about this example to suggest doing that.

Thanks,
Jonathan