Re: [PATCH v4 2/5] t1517: test commands that are designed to be run outside repository

From: Patrick Steinhardt <ps@pks.im>
Date: May 15, 2024, 14:25 UTC
Message-ID: <ZkTFvVihZAY184Gm@tanuki>
In-Reply-To: <xmqq4jaz2may.fsf@gitster.g>

On Wed, May 15, 2024 at 07:15:49AM -0700, Junio C Hamano wrote:

Show 21 quoted lines

> Patrick Steinhardt <ps@pks.im> writes:
> 
> > On Tue, May 14, 2024 at 08:08:19AM -0700, Junio C Hamano wrote:
> >> Patrick Steinhardt <ps@pks.im> writes:
> >> 
> >> > Is there a specific reason why this needs a whole patch suite, as
> >> > opposed to adding the tests to the respective test suites of the
> >> > commands?
> >> 
> >> Yes, testing out-of-repository operations needs certain trick and
> >> people forget to write such tests using the GIT_CEILING_DIRECTORIES
> >> mechanism.  Having one place where we have an enumeration of
> >> commands that are designed to be usable outside repository is a
> >> handy way to make sure that we have enough test coverage.  It would
> >> make it easy to control how GIT_DEFAULT_HASH environment is set
> >> during these tests to have them in all one place.
> >
> > We already have the "nogit" command that neatly encapsulates all of this
> > logic, so the trickery is contained in a single spot in practice.
> 
> Heh, you asked for "a" specific reason, and I listed three.

Hm. I guess I got thrown off a bit as this was written in a single paragraph, so my mind didn't process it as different reasons :)

> The tests that are spread across many scripts make it harder to see if
> we have enough coverage for the out-of-repository operations,

Put this way I in fact agree with you.

> and the use of "nongit" helper does not change the equation, does it?

No, it doesn't really.

Thanks!

Patrick