Re: More strict testing requirements?

On Thu, Dec 6, 2018 at 7:12 PM Martin Flöser <[email protected]> wrote:
>
> Am 2018-12-06 13:39, schrieb David Edmundson:
> > kwin-testWindowSelection is still failing repeatedly since Sunday.
> >
> > If we're following this we should block all non-fixing commits.
>
> The test started to fail with
> https://build.kde.org/job/Plasma/job/kwin/job/kf5-qt5%20SUSEQt5.11/225/
>
> According to the build data it's one of:
>
>      Privatize variables in InputDeviceHandler
>      Rework InputDeviceHandler focus tracking
>      Add Wayland touch drag and drop support
>
> @Roman: can you please look into it?

I looked already into it, but I haven't yet figured out why the test
wasn't failing before the offending commit (which is "Rework
InputDeviceHandler focus tracking").

We call Platform::startInteractiveWindowSelection(..) beginning at
https://phabricator.kde.org/source/kwin/browse/master/autotests/integration/window_selection_test.cpp$276
3 times, which should always induce a touch cancel signal, so the
correct signal count value would be 3 indeed.

> And yes I think we should embargo pushes till we are clean again.

We can do it now here as a test case. But most patches in review queue
are probably independent of input/touch and I simply overlooked the
failing test on push. A sensible review platform would have noticed me
about the failing test and in best case prevented me from pushing. I
hope this will be available soon to us. But yea, until then we can try
out the embargo method.

>
> Cheers
> Martin