Re: [PATCH 1/4] usage.c: add a die_message() routine

From: Junio C Hamano <gitster@pobox.com>
Date: Dec 6, 2021, 19:46 UTC
Message-ID: <xmqqa6hdfreu.fsf@gitster.g>
In-Reply-To: <xmqqee6pfrkd.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> writes:

Show 15 quoted lines

> It also makes readers wonder why this is not
>
> 	exit(die_message(err, params));
>
> which I take it a sign that this new API is overly loose to allow a
> simple single thing to be done in multiple ways.  Perhaps as the
> series progresses, the picture might improve, but if that is the
> case, perhaps the presentation order needs to be rethought.
> E.g. start without the _builtin that implies override-ability,
> convert the existing code that can benefit from calling die_message(),
> and then finally introduce _builtin that is merely an implementation
> detail, or something like that, perhaps?
>
> In any case, the first step in this four patch series is not enough
> to evaluate if this step makes sense, so let's keep reading.

OK, it is the other way around. This step with _builtin is incomplete introduction of the API, even though it was "we add the new API function without having any meaningful callers, because combining them together into a single patch is too much to chew in a single sitting". A small part of 2/4 that adds the override-able die_message_routine should be part of this step to make this step understandable.

THanks.