OpenCL objects are just pointers in the end, using random pointer value
is likely to make an application crash.
Also, all others tests use NULL as an invalid object.
See also Francisco Jerez comments:
http://lists.freedesktop.org/archives/mesa-dev/2014-November/070520.html
From NVIDIA's side, I don't think it would be that big of a problem to just
keep running our tests with PIGLIT_FORCE_WINDOW=1 or -fbo. However, my
reasoning is that Piglit should default to the 'correct' behavior according to
the OpenGL spec. While the NVIDIA driver is the only one I know of
Most tests do indeed use an FBO, either directly or via the fbo
winsys, and all will be well. However some don't (glean, probably
others) because they were written in the stone ages. Those are the
ones that have problems.
Perhaps we should just suck up the slower-ness and use that as an
incentive
On Sat, 2015-08-15 at 10:40 +1000, Timothy Arceri wrote:
On Thu, 2015-08-13 at 11:41 +0300, Tapani Pälli wrote:
These test usage of readonly and writeonly memory qualifiers for
ssbo block members.
Looks ok to me. Reviewed-by: Timothy Arceri t_arc...@yahoo.com.au
On second thought can you
On Thu, 2015-08-13 at 10:57 +0300, Tapani Pälli wrote:
Test checks that parser is able to deal with memory qualifiers given
to shader storage block.
Signed-off-by: Tapani Pälli tapani.pa...@intel.com
---
.../preprocessor/memory-qualifiers.vert | 12
1
On Thu, 2015-08-13 at 11:41 +0300, Tapani Pälli wrote:
These test usage of readonly and writeonly memory qualifiers for
ssbo block members.
Looks ok to me. Reviewed-by: Timothy Arceri t_arc...@yahoo.com.au
Signed-off-by: Tapani Pälli tapani.pa...@intel.com
---