[webkit-dev] New EWS status bubbles in Bugzilla

[Carlos Garcia Campos]

El lun, 29-09-2014 a las 23:51 -0700, Alexey Proskuryakov escribió:
> 29 сент. 2014 г., в 23:12, Carlos Garcia Campos <carlosgc at webkit.org> написал(а):
> 
> >> Please try it out, and let me know if something breaks, or is not as
> >> good as it could be!
> > 
> > Looks great! thanks. I've noticed that now we use red also for patches
> > that don't apply. I liked that there was a different color for that
> > case, since it's common when a patch depends on another one, and the
> > patch itself is not necessarily bad.
> 
> I agree that this is a non-obvious case. I chose red because EWS can't
> say anything good about the patch (or anything at all), so it needs
> extra careful attention from a reviewer.
> 
> Other options that I considered were:
> 
> - Keep it purple, as it was. I restricted purple to indicating
> internal errors, so it should pretty much never happen. Using the same
> color for the entirely different case of a patch that does not apply
> seems wrong.

Agree.

> - Make it white. That seems to make sense logically, but it has the
> same problem of using a single color for multiple purposes as red
> does. Also, I expect it to be surprising (I posted my patch several
> hours ago, why is it still white?)

Agree.

> - Add a new color. I'm not sure if this case is common enough for the
> color to become generally recognizable.

Well, I guess it's a matter of getting used to it :-)

> - Instead of per-queue bubbles, show text like "Patch does not apply
> to trunk". Not workable, because queues start at different times, and
> only some of them could fail to apply the patch.

Right.

> Perhaps most importantly, this case is really no different from
> failing to build due to a dependency on another patch, or to failing
> some tests for the same reason. So I don't think that we can represent
> dependency on another patch with color.

Fair enough, I haven't thought about those cases, I agree it's indeed
the same case and we will see red bubbles. Ideally, the EWS might apply
the patches following the dependency chain, but I guess that's not easy
to do.

> Thinking about this now, we could replace bubbles with text when no
> queue has results (i.e. at least one queue has failed to apply, and
> those that didn't still haven't started processing - we can reasonably
> expect that they will fail, too, and even if they apply cleanly, we
> can just revert to showing bubbles then). I can't think of any
> situation where this would be misleading or difficult to comprehend.
> Would this resolve your concern? Would you be willing to post a patch
> implementing this?

Well, I was just surprised to see the bubbles red because I was used to
seeing them purple, but I agree the case it's not different to when a
patch fails to build because of a dependency. So, I'll get used to the
red :-)

> - Alexey
> 
>