[Development] Proposal: Allow contributors to +1 sanity review.

André Pönitz andre.poenitz at mathematik.tu-chemnitz.de
Tue Aug 13 23:04:14 CEST 2013

On Tue, Aug 13, 2013 at 10:45:21AM -0700, Thiago Macieira wrote:
> On terça-feira, 13 de agosto de 2013 10:26:42, Alan Alpert wrote:
> > > (or because they enjoy outsmarting the system just for the sake
> > > of it, like you apparently do)
> > 
> > We *define* the system. The fact that I have to work around it in
> > order to do my job properly shows that the system is defective and
> > needs to be rectified. Lars mentioned at the contributor summit that
> > we need to make the review process easier for people, and fixing the
> > system to have fewer unnecessary hurdles is an obvious way to do that.
> > 
> > A more effective change which I recommend would be to have the bot not
> > give -1 on the heuristics which are known to give false positives more
> > frequently. It's enough that it provides a warning comment, allowing
> > the contributors to address the issue if legitimate. But the -1 can
> > also be viewed as a hint that there's relevant output, since it's so
> > prominent in the gerrit interface, so the -1 isn't a problem so long
> > as it doesn't slow us down.
> I don't see what's wrong with the current system. Let's keep it.
> The bot usually doesn't make a mistake. When it does, we can override it. 
> That's why we can override it in the first place: because it sometimes makes 
> mistakes.
> I have overridden the bot sometimes when it complained that I used British 
> English in the commit message. I know I shouldn't do that.

To be honest, that's pretty much the only case where I find myself
regularly disagreeing with the bots myself.

It's certainly valuable to enforce "one style" when it comes to
APIs (would be "fun" to have, say "QColor" and "setColour" at the
same time...), but for "free text" as in commit messages that never
gets touched again I am tempted to value cultural diversity higher
than technical streamlining ;-}


More information about the Development mailing list