[igt-dev] [PATCH i-g-t 1/6] tests/kms_vblank: Make sure that we have a mode set

[Arkadiusz Hiler]

On Mon, Apr 06, 2020 at 12:11:35PM +0300, Imre Deak wrote:
> On Fri, Apr 03, 2020 at 04:03:01PM +0300, Arkadiusz Hiler wrote:
> > Subtest 'invalid' was depending on a "leftover" mode set by something
> > external (usually fbcon) in order to work.
> > 
> > Let's make the test more robust by setting the mode explicitly.
> > 
> > Cc: Imre Deak <imre.deak at intel.com>
> > Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler at intel.com>
> > ---
> >  tests/kms_vblank.c | 14 ++++++++++++++
> >  1 file changed, 14 insertions(+)
> > 
> > diff --git a/tests/kms_vblank.c b/tests/kms_vblank.c
> > index a895ab80..edf8581e 100644
> > --- a/tests/kms_vblank.c
> > +++ b/tests/kms_vblank.c
> > @@ -472,9 +472,22 @@ static void invalid_subtest(data_t *data, int fd)
> >  {
> >  	union drm_wait_vblank vbl;
> >  	unsigned long valid_flags;
> > +	enum pipe pipe;
> > +	igt_output_t* output;
> > +	bool found = false;
> >  
> >  	igt_display_require_output_on_pipe(&data->display, 0);
> 
> Is the above needed still?

nope, removed 

> > +	/* we need to have a mode set, otherwise we will get EINVAL no matter what */
> > +	for_each_pipe_with_single_output(&data->display, pipe, output) {
> > +		data->pipe = pipe;
> > +		found = true;
> > +		break;
> > +	}
> 
> So far pipe A was assumed. I guess it's better to make it more flexible,
> but then you have to add kmstest_get_vbl_flag() to the request flags
> too.

Yeah, I have blanked on that. I have added "base_flags" that will have
this + DRM_VBLANK_RELATIVE set and use this as the initial value.

You can already see it on Trybot as I am experimenting with a fix for 0
outputs becoming a fail instead of a skip:

https://patchwork.freedesktop.org/patch/360280/?series=75545&rev=1

-- 
Cheers,
Arek