Re: Policy about not-yet-regression tests?

[Graham Percival]

On Fri, Apr 08, 2011 at 06:06:09PM +0200, David Kastrup wrote:
> Han-Wen Nienhuys <address@hidden> writes:
> 
> > On Fri, Apr 8, 2011 at 11:01 AM, David Kastrup <address@hidden> wrote:
> >> do we just want regression tests for things that _do_ work currently,
> >> or is it ok to add tests for things that _should_ work but don't, as
> >> long as the test does not bomb out?
> >
> > I think adding things that don't work is not a good idea; it will
> > confuse people, and add noise for developers.

Agreed.  We have enough confusion about the regtests as it is.

> > We have the bug tracker
> > to keep track of feature requests.
>
> Removing a bug is not the same as adding a feature.

Agreed -- but if you've identified a bug, please add it to the
tracker.

> > Almost all bugfixes should be accompanied by a regtest; if they are
> > not Graham's police should start hunting down offenders.

We have so many communication/organizational problems to fix that
there's no way this is going to happen.  If you see a patch
(helpfully highlighted with the 48-hour countdown) that doesn't
have a regtest, and you think it should have one, then of course
don't hesitate to speak up!  But I don't think we should have a
blanket rule that every patch should have a regtest.

> I am talking about bugs without a fix.  Sometimes they disappear as a
> sideeffect of fixing a different bug, or restructuring code.

I think the issue tracker is the best place for that.  It's true
that sometimes issues get fixed magically, and occasionally we
look through the regtest to identify those.

Cheers,
- Graham