From: Jeff King <peff@peff.net>
Date: Fri, 05 Feb 2010 20:13:25 GMT
Subject: Re: [PATCH 3/3] make git push --dry-run --porcelain exit with status 0 even if updates will be rejected
Message-ID: <20100205201325.GA25697@coredump.intra.peff.net>
In-Reply-To: <20100205200524.GA24027@cthulhu>

On Fri, Feb 05, 2010 at 03:05:24PM -0500, Larry D'Anna wrote:

> > So shouldn't it just be:
> > 
> >   if (args->porcelain && args->dry_run)
> >           return 0;
> > 
> > after we check for transport errors but before the loop that you are
> > modifying.
> 
> I don't know what the deal is with REF_STATUS_EXPECTING_REPORT, so I
> didn't want to modify the behavior in the case that ref->status was
> that.  What does expecting report mean?

It means we told the other side we wanted to push that ref, and we
expect it to give us a status report. Most refs are in that state for a
short period, and then moved to their final state in
builtin-send-pack.c:receive_status. But if we never get a status for
that ref for some reason, then that could be the final state.

But more to the point, I don't think this bit of code should _have_ to
care what it means. If there is a per-ref error with "push --dry-run
--porcelain", it will be shown on that ref's output line. So I think
your proposal should simply be "if dry-run and porcelain, don't bother
looking at per-ref errors at all". You don't care what the per-ref
error is; they are all in the same class from the perspective of this
change.

-Peff