[igt-dev] [PATCH i-g-t] igt/kms_addfb_basic: require display

[Daniel Vetter]

On Thu, Jun 28, 2018 at 02:06:49PM +0200, Maarten Lankhorst wrote:
> Op 28-06-18 om 13:54 schreef Daniel Vetter:
> > On Thu, Jun 28, 2018 at 12:19:57PM +0200, Maarten Lankhorst wrote:
> >> Op 28-06-18 om 11:56 schreef Jani Nikula:
> >>> Running the tests with i915.disable_display=1 leads to IGT errors. Skip
> >>> tests that need display.
> >>>
> >>> References: http://patchwork.freedesktop.org/patch/msgid/20180608124057.6889-1-jani.nikula@intel.com
> >>> Cc: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> >>> Cc: Daniel Vetter <daniel at ffwll.ch>
> >>> Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> >>> Signed-off-by: Jani Nikula <jani.nikula at intel.com>
> >>>
> >>> ---
> >>>
> >>> What's the best paradigm for this? There's loads of random and cargo
> >>> culted igt_requires for this stuff, with various checks on pipes > 0
> >>> etc.
> >>> ---
> >> igt_display_require_output() is the easiest check to find if an output
> >> is connected, should be approximately same as display enabled. :)
> > Unfortunately not quite. I think we do want to run these tests here on
> > headless systems which have a display block, but no outputs connected.
> > Just for better test coverage and all that.
> >
> > I think what we want is a new
> >
> > igt_require_display(int fd)
> > {
> > 	igt_display_t display;
> > 	igt_display_init(&display, fd);
> > 	igt_require(display->n_pipes > 0, "drm driver without display
> > 	hardware\n");
> > }
> >
> > gtkdoc plus wiring it up needed too ofc.
> >
> > We could also patch up igt_display_init() to check for this (and probably
> > should, not sure about that), but a very simply igt_require helper that
> > only takes the fd seems useful to me for all the low-level tests like
> > kms_addfb_basic here which don't want/need a full igt_display_t
> > initialized.
> But igt_display is not without side effects. It sets the atomic cap, which would be surprising in a require statement.

Hm, then maybe we need to open-coude the drmgetresources call + checking
for numCrtcs. Not that many more lines really.

> i915 without pipes should probably unset DRIVER_MODESET.

Too late by that point, and we're not the only driver doing this.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch