Re: [PATCH] gpg-interface: lazily initialize and read the configuration

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 9, 2023, 16:38 UTC
Message-ID: <xmqqwn4qokug.fsf@gitster.g>
In-Reply-To: <Y+Tr1g+HTn45rsTq@coredump.intra.peff.net>

Jeff King <peff@peff.net> writes:

> This all looks fairly sensible to me. I think we'd really want to see a
> "rev-list --format" test, too. One, because that's the immediate goal of
> this change.

Heh. I primarily wanted to see how much damage to the code does it take to implement such a lazy-loading scheme ;-) Which turned out to be "not much".

Maybe when I have time next time, but no promises.

> But two, because I think we are only guessing that loading
> the config is sufficient here. We've had bug with other subsystems where
> they expected to be initialized but plumbing callers didn't (e.g., the
> lazy init of notes-refs, etc).

Yup.

> I _think_ we're probably good here. Just looking at "git log" (where we
> know --format, etc, works), it doesn't seem to do anything beyond
> initializing the config.

That was my recollection from back when gpg-interface was split out of "git tag".

Thanks for sanity checking.