Re: [PATCH 4/4] t7401: add a WARNING and a NEEDSWORK

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 6, 2020, 21:11 UTC
Message-ID: <xmqqa6z7jx7q.fsf@gitster.c.googlers.com>
In-Reply-To: <20200806112702.GA15343@konoha>

Shourya Shukla <shouryashukla.oo@gmail.com> writes:

Show 19 quoted lines

> On 05/08 02:36, Junio C Hamano wrote:
>> Shourya Shukla <shouryashukla.oo@gmail.com> writes:
>> 
>> > Add a WARNING regarding the usage of 'git add' instead of 'git
>> > submodule add' to add submodules to the superproject.
>> 
>> Is that a warning worthy thing?  As far as I know, using "git add"
>> to register a gitlink is perfectly fine and a supported way to start
>> a new submodule.  It may have to be followed by other steps like
>> "git config -f .gitmodules" (e.g. when operations that needs to use
>> the contents recorded in the .gitmodules file are to be tested), but
>> writing tests using lower-level ingredients for finer grained tests
>> is not all that unusual, is it?  I dunno.
>
> The thing is that 'git submodule {init,deinit}' fail when there is no
> .gitmodules. I can initiliase the .gitmodules separately using 'git
> config -f .gitmodules' but I think it will be better to use 'git
> submodule add' throughout the script rather than worry about it all the
> time.

On the other hand, we do want to make sure that the workflow using lower-level tools continues to work, so that is a balancing act.

> But again, if the warning seems unnecessary, then I can obviously use
> 'git config' to initilaise the submodules and change this commit. What
> do you reckon?

If you need "git submodule init" etc. to work in this test, yes, you can change the test to either use "git submodule add" instead, or "git config -f .gitmodules" in addition. If you don't, there is no need to change anything, no?