Re: [PATCH v2 3/6] tests: use `git submodule add` and fix expected diffs

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 6, 2023, 23:57 UTC
Message-ID: <xmqqlek977se.fsf@gitster.g>
In-Reply-To: <kl6lsfehihdv.fsf@chooglen-macbookpro.roam.corp.google.com>

Glen Choo <chooglen@google.com> writes:

Show 10 quoted lines

> E.g. we could have ignored .gitmodules during the diff like so
>
>   test_expect_success 'added submodule' '
>           git submodule add ./sm1 &&
>           gitmodules_hash1=$(git rev-parse --short $(git hash-object .gitmodules)) &&
>   -       git diff-index -p --submodule=log HEAD >actual &&
>   +       git diff-index -p --submodule=log HEAD -- :!.gitmodules >actual &&
>
> and then we wouldn't have to adjust the diff. That would be my preferred
> approach, since it keeps the irrelevant details out of the test.

Yes, it certainly does make the diff simpler to read, but the reasoning cuts both ways.

The end-goal of the topic is to propose that updating .gitmodules is an integral part of "adding a submodule" and to make it impossible to just add gitlink without a command line option using "git add", so from that point of view, readers would expect that the diff after "adding a submodule" to have both new gitlink and also a modified (or if this were the first submodule, added) ".gitmodules" file.