Re: [PATCH] tests: prefer host Git to verify chainlint self-checks

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 14, 2023, 16:16 UTC
Message-ID: <xmqqbkaspxn6.fsf@gitster.g>
In-Reply-To: <CAPig+cQ2-PB24n0xfcoSy_1UT-VbEZUXXJ9QbA8FBA8Vfyd6Ng@mail.gmail.com>

Eric Sunshine <sunshine@sunshineco.com> writes:

Show 9 quoted lines

>>   - Fix the ".expect" files so that we can avoid all of these games.
>>
>> I actually like the last option most. I'll have a go at it and send this
>> third version out in a bit.
>
> I sent a reply[1] in the other thread explaining why I'm still leaning
> toward `sed` to smooth over these minor differences rather than
> churning the "expect" files, especially since the minor differences
> are not significant to what is actually being tested.

If it is just one time bulk conversion under t/chainlint/ to match what the chainlint.pl script produces, with the possibility of similar bulk updates in the future when the script gets updated, I tend to agree with Patrick that getting rid of the fuzzy comparison will be the best way forward.

Especially if the fuzzy comparison is done only to hide differences between what the old chainlint.sed used to produce and what the current version produces, that is. If for some reason the script started to create subtly different output for other reasons (e.g., it may produce different whitespaces on a particular platform, or with a specific version of perl interpreter), we'd better be aware of it, instead of blindly ignoring the differences without inspecting them and verifying that they are benign.

By going through the single conversion pain, it will force us to think twice before breaking its output stability while updating chainlint.pl, which would also be a good thing.

THanks.