Re: [PATCH v4 7/7] gpg-interface t: extend the existing GPG tests with GPGSM

From: Junio C Hamano <gitster@pobox.com>
Date: Jul 18, 2018, 17:06 UTC
Message-ID: <xmqqbmb4lbhh.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <20180718123617.55acfd3b@md1pvb1c.ad001.siemens.net>

Henning Schild <henning.schild@siemens.com> writes:

Show 16 quoted lines

>> > +	unset GIT_COMMITTER_EMAIL &&
>> > +	git config user.email hasnokey@nowhere.com &&
>> > +	git config user.signingkey "" &&
>> > +	test_must_fail git push --signed dst noop ff +noff &&  
>> 
>> This is OK for a test that is known to be always at the end, but
>> also forbids others to further update this script to add more tests
>> at the end, as the standard setting of environment is blown away
>> (the config is probably OK, but test_config to arrange them to be
>> cleaned up would have been nicer), which is not very nice.  I think
>> it should be easy to fix it when it becomes necessary, but at the
>> same time if it is easy to fix, then probably we shouldn't introduce
>> a breakage in the first place, so I am on the fence.
>
> Switched to test_config, this is all coming from copying the previous
> tests, which i left as is.

I was specifically talking about the unsetting of GIT_COMMITTER_EMAIL, not the configuration variables.

An easy way to try one step without the environment without affecting later tests would be to run the above in a subshell, perhaps like

	test_config user.email ... &&
	test_config user.signingkey "" &&
	(
		safe_unset GIT_COMMITTER_EMAIL &&
		test_must_fail git push --signed ...
	) &&