Re: Git Test Coverage Report (Friday, Nov 2)

From: Michał Górny <mgorny@gentoo.org>
Date: Nov 3, 2018, 16:22 UTC
Message-ID: <1541262152.1028.20.camel@gentoo.org>
In-Reply-To: <xmqqr2g2mqaq.fsf@gitster-ct.c.googlers.com>

On Sat, 2018-11-03 at 19:03 +0900, Junio C Hamano wrote:

Show 20 quoted lines

> Michał Górny <mgorny@gentoo.org> writes:
> 
> > As for how involved... we'd just have to use a key that has split
> > signing subkey.  Would it be fine to add the subkey to the existing key?
> >  It would imply updating keyids/fingerprints everywhere.
> 
> Yes, that "everywhere" is exactly what I meant by "how involved",
> and your suggestion answers "very much involved".
> 
> If we can easily add _another_ key with a subkey that is not the
> primary one we use for other tests, without touching the existing
> key and the existing tests that use it (including the one I touched
> below--- we'd want to see a sig with a key that is not split is
> shown with the same %GF and %GP), while adding a handful of new
> tests that create signed objects under the new & split key and 
> view them with %GF and %GP, then the end result would be that we
> managed to add a new test case where %GF/%GP are different without
> making very much involved changes.  I guess that was what I was
> getting at.
>

I've just did a little research and came to the following results:

1. modifying the 'C. O. Mitter' key would require changes to 4 tests,

2. modifying the 'Eris Discordia' key would require changes to 2 tests
   (both in 7510).

Do you think 2. would be an acceptable option? I think changing 2 tests would be preferable to proliferating a third key for one test case. Also, given that both failing tests are specifically format string tests, one of them would serve additional purpose of testing %GP!=%GF.

-- 
Best regards,
Michał Górny