Re: [PATCH] HOME must be set before calling git-init when creating test repositories

From: Alex Riesen <raa.lkml@gmail.com>
Date: Mar 26, 2011, 10:08 UTC
Message-ID: <AANLkTi=+SZGxLyP8vFPpmK8DZvke6-Tu-crwq5+89qWx@mail.gmail.com>
In-Reply-To: <7vvcz64ygm.fsf@alter.siamese.dyndns.org>

On Fri, Mar 25, 2011 at 23:39, Junio C Hamano <gitster@pobox.com> wrote:

Show 24 quoted lines

> Alex Riesen <raa.lkml@gmail.com> writes:
>
>> On Fri, Mar 25, 2011 at 22:30, Junio C Hamano <gitster@pobox.com> wrote:
>>> Alex Riesen <raa.lkml@gmail.com> writes:
>>>
>>>> On Fri, Mar 25, 2011 at 21:49, Junio C Hamano <gitster@pobox.com> wrote:
>>>>
>>>>> Doesn't this change the location of HOME used during the test as well?
>>>>
>>>> As long as the test only includes test-lib.sh only once - it doesn't.
>>>> Why? Or rather, how?
>>>
>>> I thought you moved HOME=$(pwd) across "cd somewhere-else".  Doesn't it
>>> change what is returned from pwd?
>>>
>>
>> Oh, it does. That's why the second patch (prefixed "[PATCH, fixed]").
>> It makes HOME to be "$(pwd)/somewhere-else", or precisely:
>>
>>   HOME="$(pwd)"/"$test"
>>   export HOME
>
> What happens to people who has non-empty "$root", iow, their $test begins
> with '/'?

It's still under $test then.

> I am not saying that having HOME at t/ directory instead of t/trash-*/
> directory would necessarily break things (I don't know).  I am just
> pointing out that the patch changes behaviour.

It does. I still think we're better off using the test's trash directory for a this. For instance, consider the case when a user's .gitconfig created by one of the tests collides with .gitconfig's of the other tests. Either when running in parallel or just sequentially: the .gitconfig in "t/" is not cleaned up after a test finishes.