[Mesa-dev] [PATCH] st/mesa: expose OES_geometry_shader and OES_texture_cube_map_array

Ian Romanick idr at freedesktop.org
Thu Sep 1 17:58:21 UTC 2016


I sent a piglit patch for this on 8/24.  I missed that Tapani sent a
R-b, so I'll push it now.

On 09/01/2016 07:49 AM, Ilia Mirkin wrote:
> On Thu, Sep 1, 2016 at 2:23 AM, Michel Dänzer <michel at daenzer.net> wrote:
>>
>> This change broke these piglit tests with radeonsi:
>>
>> spec at glsl-es-3.10@built-in constants
>> spec at oes_geometry_shader@built-in constants
>>
>> All the shaders fail compilation like this:
>>
>> Failed to compile shader gl_MaxAtomicCounterBindings: 0:6(8): error: #version 150 layout qualifier `points' used
>> 0:7(8): error: #version 150 layout qualifier `points' used
>> 0:7(31): error: #version 150 max_vertices qualifier specified
> 
> Actually the test was already broken (wrt this feature). It forgets to
> put in a OES_geometry_shader enable. Similarly it sticks
> OES_tessellation_shader enables into each shader stage, which is
> apparently illegal.
> 
>> Also, FWIW, the change enabled the tests
>>
>> spec at oes_geometry_shader@sso_validation at user-defined-gs-input-(not-)in-block
>>
>> but they fail:
>>
>> Mesa: User error: GL_INVALID_OPERATION in glValidateProgramPipeline failed to validate the pipeline
>> Probe color at (0,0)
>>   Expected: 0.000000 1.000000 0.000000 1.000000
>>   Observed: 0.000000 0.000000 0.000000 1.000000
>> Test failure on line 70
>> PIGLIT: {"result": "fail" }
> 
> I believe that's a disagreement between piglit and mesa, but I'm not
> sure. Ian would probably know better.
> 
>   -ilia
> 



More information about the mesa-dev mailing list