On Tue, Mar 23, 2021 at 01:25:13PM -0400, Lyude wrote: > From: Lyude Paul <lyude@xxxxxxxxxx> > > Currently we just assume that every cursor size up to data->cursor_max_w/h will > be supported by the driver, and check for support of nonsquare cursors by > checking if we're running on u815 and if so, which variant of intel hardware > we're running on. This isn't really ideal as we're about to enable 32x32 cursor > size tests for nouveau, and Intel hardware doesn't support cursor sizes that > small. > > So, fix this by removing has_nonsquare_cursors() and replacing it with a more > generic require_cursor_size() function which checks whether or not the driver > we're using supports a given cursor size by attempting a test-only atomic > commit. > > v3: > * Also probe for cursor support on systems without atomic support Gave this a go on my i865, and after removing a bogus max_w==max_h assert it all passed, and your non-square probing worked fine. I'll need to send a patch to nuke that bogus assert, and I also had an idea to test the max sized cursor as well (in case it's not square and thus not being tested already)... with the s/u815/i915/ Petri pointed out Reviewed-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> -- Ville Syrjälä Intel _______________________________________________ Nouveau mailing list Nouveau@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/nouveau
