From: Michael S. Tsirkin <mst@redhat.com>
Date: Thu, 07 Apr 2016 17:30:34 GMT
Subject: Re: [PATCH 1/4] builtin/interpret-trailers.c: allow -t
Message-ID: <20160407202938-mutt-send-email-mst@redhat.com>
In-Reply-To: <xmqqa8l51gae.fsf@gitster.mtv.corp.google.com>

On Thu, Apr 07, 2016 at 10:26:33AM -0700, Junio C Hamano wrote:
> Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:
> 
> >> I am in principle OK with the later step that teaches a single
> >> letter option to end-user facing "git am" that would be turned into
> >> "--trailer" when it calls out to "interpret-trailers" (I haven't
> >> checked if 't' is a sensible choice for that single letter option,
> >> though).
> >
> > If 'am' has -t == --trailer, I think it makes sense to have the same
> > shortcut in interpret-trailers for consistency.
> 
> It is the other way around.  "git am" may be OK with "-t" (or it may
> not--I do not know yet), but other commands that are currently
> unaware of "interpret-trailers" (cherry-pick, revert, etc.) may have
> better uses for a short-and-sweet 't'.
> 
> In the ideal future, "interpret-trailers" should not have to exist
> in the end-users' vocabulary, as all the front-line end-user facing
> programs would be aware of it.  But we are not there.
> 
> Letting it reserve a short-and-sweet 't' that allows it to dictate
> that its callers must have the same 't' is tail wagging the dog that
> I want to avoid.

It's mostly a short-cut I took by copying calls to applypatch.
Are there examples of other commands doing such transformations
on the fly?

-- 
MST