On Fri, Aug 11, 2023 at 11:01:14AM -0400, Jeff King wrote:
> > Hmph, doesn't this potentially cause us to emit the two reports
> > alternating, if we are unlucky enough to see a commit with 0
> > generation first (which will silently set gz to ZERO_EXISTS), then
> > another commit with non-zero generation (which will complain we saw
> > non-zero for the current one and earlier we saw zero elsewhere, and
> > then set gz to NUM_EXISTS), and then another commit with 0
> > generation (which will complain the other way, and set gz back again
> > to ZERO_EXISTS)?
> >
> > I am tempted to say this gz business should be done with two bits
> > (seen zero bit and seen non-zero bit), and immediately after we see
> > both kinds, we should report once and stop making further reports,
> > but ...
>
> Yeah, I think you are right. It might be OK, in the sense that we would
> show a different commit each time as we flip-flopped, but it's not clear
> to me how valuable that is.
>
> If the actual commit ids are not valuable, then we could just set bits
> and then at the end of the loop produce one warning:
>
> if (seen_zero && seen_non_zero) {
> graph_report("oops, we saw both types");
> }
Thanks, both. I think that this is a very reasonable suggestion and an
improvement on the existing reporting. I made this change and sent the
revised series as a v2 lower down in the thread.
Thanks,
Taylor
