On Mon, Oct 21, 2013 at 06:15:06PM +0200, Daniel Vetter wrote: > On Mon, Oct 21, 2013 at 6:01 PM, <ville.syrjala@xxxxxxxxxxxxxxx> wrote: > > From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> > > > > We store cursor_x/y as int16_t internally, but the user provided > > coordinates are int32_t. Clamp the coordinates so that they don't > > overflow the int16_t. Since the cursor is only 64x64 in size, the > > clamping can't cause any visual changes. > > > > Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx> > > > For shits&giggles: You've bothered with a subtest for this in your crc > cursor test? Should be a one-liner and would neatly test whether QA > catches it ... Yeah it's there. Actually I added the test before I knew the bug existed. The test is what found the bug for me. It's not a specific subtest though, but part of all the bigger cursor-offscreen subtests. The subtest split I'm using there isn't very fine grained. Maybe I should change that actually. It would help in pinpointing the failing cases more accurately. -- Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx