[Piglit] [PATCH 5/7] arb_enhanced_layouts: Multiple layout qualifiers in a single declaration tests

Andres Gomez agomez at igalia.com
Mon Oct 24 12:58:36 UTC 2016

On Mon, 2016-10-24 at 11:05 +1100, Timothy Arceri wrote:
> On Sat, 2016-10-22 at 23:42 +0300, Andres Gomez wrote:
> > We already had tests for a repeated layout-qualifier-name in a single
> > layout-qualifier. Now, we also add similar tests to check across
> > multiple layout-qualifiers in the same declaration.
> > 
> > From the ARB_enhanced_layouts spec:
> > 
> >     "More than one layout qualifier may appear in a single
> > declaration.
> >      Additionally, the same layout-qualifier-name can occur multiple
> > times
> >      within a layout qualifier or across multiple layout qualifiers
> > in the
> >      same declaration"
> I believe this is just because the spec is written against the GLSL
> 4.30 spec. I'm not so sure that enhanced layouts should add this
> support. Did you test the Nvidia and Amd binary drivers to see what
> they do?

I run the tests with the proprietary drivers in a GNU/Linux box.

NVidia and AMD are a mess with respect to this. Some pass, some not.

My conclusion is that their implementation depends on the specific
layout-qualifier-name tested. Therefore, if with, let's say
max_vertices, they don't allow duplicates with different values, they
will fail in all the cases; with duplicates in a single layout-
qualifier or among several in the same declaration, no matter
if ARB_shading_language_420pack or ARB_enhanced_layouts are supported.

If duplicates with different values are allowed for the layout-
qualifier-name tested, then the ones failing will be the negative
tests. A mess ...

Let's drop this by now. It is always easy to add more tests later.

> Also I believe we should already have negative tests for this so some
> of these tests are likely duplicates of existing tests,
> probably ARB_shading_language_420pack tests.

Oh, right, it is stupid to test for the absence in 2 different places.
Good catch.



More information about the Piglit mailing list