[Piglit] Issue with Piglit pixel ownership assumptions

James Jones jajones at nvidia.com
Wed Aug 12 17:54:02 PDT 2015

The PIGLIT_NO_WINDOW patch seems like the way to go to me, since 
defaulting to unreliable behavior seems bad, but allowing users to fall 
back to it to test on drivers that don't enforce pixel ownership seems 
fine.  I'd prefer if someone outside NVIDIA could review as well, but for:


Reviewed-by: James Jones <jajones at nvidia.com>

I'll commit this soon if no one objects.


On 07/20/2015 02:44 PM, Alexander Goins wrote:
> Hello all,
> During some recent debugging, I seem to have discovered  an issue with
> Piglit that could cause problems with some drivers. With -auto, and
> without -fbo or PIGLIT_FORCE_WINDOW=1, Piglit defaults to rendering to
> the default framebuffer without mapping a window. This means that it
> doesn't take pixel ownership, which could cause a problem with some
> drivers, particularly NVIDIA's with Unified Back Buffer enabled. The
> OpenGL spec requires that pixel ownership be established in order to
> guarantee that the window system doesn't clobber the output. There are
> more details in the patch descriptions.
> It would probably be best to deprecate the functionality of not showing
> a window altogether, making users use FBOs if they want offscreen
> rendering. I made a patch that simply removes that functionality. As a
> softer solution, however, I also made a patch that makes displaying the
> window the default, and replaced PIGLIT_FORCE_WINDOW with
> PIGLIT_NO_WINDOW, a flag that allows tests to run without mapping a
> window. This way, users that have been relying on the existing
> functionality can continue using it.
> Both patches are included.
> Thanks,
> Alex

More information about the Piglit mailing list