Re: [PATCHv2 1/5] revision.h: turn rev_info.early_output back into an unsigned int

From: Jeff King <peff@peff.net>
Date: Jun 12, 2017, 21:59 UTC
Message-ID: <20170612215902.62cnfxdivv76337p@sigill.intra.peff.net>
In-Reply-To: <xmqq60g1ndtz.fsf@gitster.mtv.corp.google.com>

On Mon, Jun 12, 2017 at 01:36:08PM -0700, Junio C Hamano wrote:

Show 11 quoted lines

> > I'm not sure if I was just being thick or if that point (and the fact
> > that --early-output has basically been a noop since 2011!) should be
> > made more explicit.
> >
> > Given that nobody noticed, I kind of wonder if we should consider
> > ripping the feature out entirely.
> 
> Yes, we may want to think about deprecating it, especially given
> that it is not advertised anywhere.
> 
> In any case, the patch looks correct ;-)

Yeah, it's definitely orthogonal to Gábor's patches.

I also wondered who might be using the feature. I assumed it was written with gitk in mind. Digging in the list archive that seems to be the case, and there was even an RFC patch for gitk to use it:

  http://public-inbox.org/git/18221.2285.259487.655684@cargo.ozlabs.ibm.com/

but AFAICT it was never merged.

It looks like QGit had a patch around that time, too, but left the line using "--early-output" commented out until Git merged the patches. And then it never got uncommented. ;)

Tig doesn't seem to use it at all. I don't know about other GUIs. I'd kind of doubt, given the obscurity of the feature (both gitk and qgit authors were involved in the discussion way back in 2007).

The nice thing about deprecating it is that I think callers need to be prepared to handle the case already that it does nothing. So if we just ripped out the code and treated it as a silent noop, everything would just work.

-Peff