From: Omar Polo <op@omarpolo.com>
Subject: Re: got patch stats about conflicts and rejects
To: Stefan Sperling <stsp@stsp.name>
Cc: gameoftrees@openbsd.org
Date: Fri, 12 May 2023 15:11:14 +0200

On 2023/05/11 09:53:09 +0200, Stefan Sperling <stsp@stsp.name> wrote:
> On Thu, May 11, 2023 at 08:59:13AM +0200, Omar Polo wrote:
> > On 2023/05/11 07:34:26 +0200, Stefan Sperling <stsp@stsp.name> wrote:
> > > This left me wondering where the problem was. Other commands treat
> > > conflicted files as an expected outcome rather than an error.
> > 
> > Admittedly I haven't noticed this difference.  I considered them a
> > failure because the presence of a conflict means that we couldn't do
> > what we were asked to.
> 
> Not always a failure. It will depend on the specific conflict whether
> there is a real problem and only a human can judge this.

You're right!  In fact, even a successful diff application (in the
search/replace sense) might not be, in fact, correct.

> Applying patches that are based on out of date files can easily result
> in conflicts that are trivial to resolve manually.

also true.

Do you think we should change something else in got patch?  I haven't
thought too hard about our return codes and what has to be considered
an "error"; I reasoned strictly from the point of view of the patch
operation without taking into consideration the "meaning", because
that's something got patch can't do, but maybe there are better ways
to communicate what it did.