From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 23 Sep 2015 20:29:23 GMT
Subject: Re: [PATCH v6 5/8] branch: drop non-commit error reporting
Message-ID: <xmqqr3lox4kc.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <vpq6130ucbh.fsf@grenoble-inp.fr>

Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:

> More precisely: if we find such a branch ref and we're used with an
> option that requires us to lookup the commit, then we report it as an
> error.
> ...
> So I agree with Junio that the commit message is not sufficient: there
> is a behavioral change. I'm OK with it, but the commit message shouldn't
> claim that there isn't.
>
> Porting to ref-filter drops the commit before we get an opportunity to
> complain, so we stop complaining because it's not worth the trouble.

I share the same conclusion.  It may be an unfortunate fallout but
giving the diagnosis was not really the job for this codepath in the
first place.

If it were, then I would have said we should fix it to keep the
behaviour, even if the fix is involved, and that is why I would not
necessarily agree with "not worth the trouble".

> BTW, this looks like an fsck bug:
>
> $ git fsck --strict
> Checking object directories: 100% (256/256), done.
> error: refs/heads/broken: not a commit
> $ echo $?
> 0

Interesting.  Perhaps leave it as a MicroProject for GSoC next year?
;-)