Re: [PATCH 08/16] t/t5516-fetch-push: use test_config()

From: Johannes Sixt <j6t@kdbg.org>
Date: Jun 22, 2013, 09:57 UTC
Message-ID: <51C57527.8060806@kdbg.org>
In-Reply-To: <7vobazjdrf.fsf@alter.siamese.dyndns.org>

Am 22.06.2013 00:32, schrieb Junio C Hamano:

Show 22 quoted lines

> Ramkumar Ramachandra <artagnon@gmail.com> writes:
> 
>> Replace the 'git config' calls in tests with test_config for greater
>> robustness.
> 
> That may be a good thing in principle, but I _think_
> 
> 	mk_empty testrepo &&
>         (
>         	cd testrepo &&
>                 do whatever to its config &&
>                 run test
> 	)
> 
> sequence is used so that we do not even have to worry about what
> leftover configuration values are in the testrepo/.git/config; so
> does it really matter?
> 
> If this conversion had something more than "s/git config/test_config/"
> replacement, that would indicate that you uncovered a bug in the
> existing test and found a good fix, but that does not seem to be the
> case for this particular patch.

And just let me add that the added benefit of test_config to remove the configuration change after the test case finished would not work because the test_when_finished registration that happens behind the scenes would be forgotten when the sub-shell exits.

Show 9 quoted lines

>> @@ -142,8 +142,8 @@ test_expect_success 'fetch with wildcard' '
>>  	mk_empty testrepo &&
>>  	(
>>  		cd testrepo &&
>> -		git config remote.up.url .. &&
>> -		git config remote.up.fetch "refs/heads/*:refs/remotes/origin/*" &&
>> +		test_config remote.up.url .. &&
>> +		test_config remote.up.fetch "refs/heads/*:refs/remotes/origin/*" &&
>>  		git fetch up &&

...

-- Hannes