[igt-dev] [PATCH i-g-t 0/2] tests/kms_plane_*: Reset display before execution tests

[Petri Latvala]

On Tue, Aug 17, 2021 at 06:27:12PM +0300, Juha-Pekka Heikkila wrote:
> On 17.8.2021 16.52, Petri Latvala wrote:
> > On Tue, Aug 17, 2021 at 04:45:21PM +0300, Mika Kahola wrote:
> > > Reset display before executing tests. This way display state is
> > > in known state even if the previous test has failed.
> > 
> > I was going to suggest having reset in the common display
> > initialization but it is there already. When do you see this kind of a
> > thing happening for these tests? When running multiple subtests at
> > once?
> > 
> 
> Look like there's bug somewhere with commits hence this attempt fix. Bug can
> be seen if kms_cursor_crc is run after failed kms_plane_alpha_blend, alpha
> will not get reset for kms_cursor_crc and it will fail with crc mismatch
> (plane alpha different on cursor vs primary).
> 
> To me it look like if first igt_commit* after igt_display_reset() is using
> legacy api not all committed values go in but if use atomic api things will
> work. I've reliably reproduced this but didn't have time to look into why
> plane alpha will not get reset.
> 
> Most kms tests will not care about this because they do crc comparisons
> planeN vs planeN

Ah yeah, that kind of makes sense, thanks.

Series is
Acked-by: Petri Latvala <petri.latvala at intel.com>