[Mesa-ci-intel] Regressions on staging/19.1

[Clayton Craft]

On Mon, Oct 14, 2019 at 01:28:15PM +0200, Juan A. Suarez Romero wrote:
>
>Any news on this?
>

Sorry this took me a while, some other things came up that required my
attention and it took a while to pick through the results and figure out what
caused some of the failures.

Basically, I didn't find any regressions that I could blame on mesa. There are a
few unstable/flaky tests in the 19.1 branch, and I don't know why (I wasn't able
to bisect to any particular commit..). Since this is the last release for 19.1
and there are no obvious test regressions, I think it's OK to release from a
CI/regression perspective.

For completeness, I've included my notes/comments about failures I see on the
19.1 staging daily build below.

-Clayton


gen9 - glcts:
        KHR-GL46.enhanced_layouts.varying_location_aliasing_with_mixed_auxiliary_storage

        Test fixed with a newer GL CTS than what is used in the 19.1 staging
        branch CI job


gen9/gen9 atom - piglit:
        spec.intel_shader_atomic_float_minmax.execution.ssbo-atomiccompswap-float-negative-zero
        spec.intel_shader_atomic_float_minmax.execution.ssbo-atomicmax-float-nan
        spec.intel_shader_atomic_float_minmax.execution.ssbo-atomicmin-float-nan

        These seems to be failures associated with piglit, they do not fail on
        newer versions of piglit. I think they can be ignored for the 19.1
        branch.

All platforms - crucible:
        func.multiview.count_6.q0
        func.multiview.count_6.masked_0_2.q0
        func.multiview.count_6.masked_1_3_5.q0
        func.multiview.count_6.masked_3_4.q0

        These can be ignored. New tests added to crucible recently and CI was
        erroneously using newer crucible builds (this has been corrected and
        will be reflected in the next build)

g33/g965 - dEQP:
        dEQP-EGL.functional.sharing.gles2.multithread.random.programs.link.7
        dEQP-EGL.functional.sharing.gles2.multithread.simple.textures.teximage2d_copytexsubimage2d_render
        dEQP-EGL.functional.render.multi_thread.gles2.rgba8888_pbuffer

        These tests appear to be flaky on the 19.1 stable branch, I wasn't able
        to blame any particular commit for it. I think they can be ignored

SNB - piglit:
        piglit.spec.arb_texture_multisample.texelfetch.2-gs-usampler2dms
        piglit.spec.glsl-1_50.execution.built-in-functions.gs-op-bitxor-abs-not-int-int
        piglit.spec.glsl-1_50.execution.built-in-functions.gs-op-bitand-neg-abs-ivec4-int

        These tests appear to be flaky on the 19.1 stable branch, I wasn't able
        to blame any particular commit for it. I think they can be ignored


BSW - gl cts/dEQP/piglit:

        dEQP-GLES3.functional.negative_api.texture.compressedteximage3d_invalid_astc_target
        piglit.spec.!opengl 1_1.read-front clear-front-first samples=16 bsw     m64
        piglit.spec.arb_sample_shading.samplemask 16  
        piglit.spec.khr_texture_compression_astc.miptree-gles srgb-fp 
        piglit.spec.arb_fragment_shader_interlock.compiler.begininvocationinterlock-inside-non-main-function_frag
        piglit.spec.arb_fragment_shader_interlock.compiler.endinvocationinterlock-inside-for_frag
        piglit.spec.arb_fragment_shader_interlock.compiler.endinvocationinterlock-inside-while_frag
        piglit.spec.!opengl 1_1.read-front samples=16 
        piglit.spec.arb_fragment_shader_interlock.compiler.begininvocationinterlock-inside-for_frag
        piglit.spec.arb_fragment_shader_interlock.compiler.begininvocationinterlock-inside-while_frag
        piglit.spec.arb_fragment_shader_interlock.compiler.endinvocationinterlock-before-begininvocationinterlock_frag
        piglit.spec.arb_fragment_shader_interlock.compiler.endinvocationinterlock-inside-non-main-function_frag
        KHR-NoContext.es32.robustness.lose_context_on_reset   
        KHR-NoContext.es32.robustness.readnpixels     
        KHR-NoContext.es32.robustness.no_reset_notification   
        KHR-NoContext.es32.context_flags.debug_flag_set_case  
        KHR-NoContext.es32.context_flags.robust_flag_set_case 
        KHR-NoContext.es32.context_flags.all_flags_set_case

        These tests appear to be flaky on the 19.1 stable branch, I wasn't able
        to blame any particular commit for it. I think they can be ignored. The
        kernel on BSW systems in CI was recently updated to resolve some test
        regressions on master, it's possible the test instability here is due to
        the newer mainline kernel running on them...


gen9 atom - dEQP:
        dEQP-GLES2.functional.shaders.texture_functions.vertex.texturecubelod 
        dEQP-GLES2.functional.shaders.fragdata.invalid_assign_to_1    
        dEQP-GLES2.functional.texture.vertex.cube.filtering.linear_mipmap_linear_linear_mirror
        dEQP-GLES2.functional.texture.vertex.cube.wrap.mirror_mirror  
        dEQP-GLES2.functional.texture.vertex.cube.wrap.clamp_clamp    
        dEQP-GLES2.functional.fbo.api.attachment_query_empty_fbo      
        dEQP-GLES3.functional.shaders.texture_functions.texturelodoffset.usampler3d_vertex
        dEQP-GLES3.functional.pbo.renderbuffer.rgb10_a2_triangles     
        dEQP-GLES2.functional.texture.mipmap.cube.basic.linear_nearest
        dEQP-GLES2.functional.texture.mipmap.cube.projected.linear_nearest    
        dEQP-GLES2.functional.texture.mipmap.cube.bias.linear_nearest 
        dEQP-GLES2.functional.texture.vertex.cube.filtering.linear_mipmap_linear_nearest_clamp
        dEQP-GLES2.functional.texture.vertex.cube.wrap.clamp_repeat   
        dEQP-GLES2.functional.fragment_ops.interaction.basic_shader.61
        dEQP-GLES3.functional.shaders.texture_functions.texturelodoffset.sampler3d_float_vertex
        dEQP-GLES3.functional.shaders.texture_functions.textureprojlodoffset.isampler3d_vertex
        dEQP-GLES3.functional.shaders.fragdata.draw_buffers   
        dEQP-GLES2.functional.texture.vertex.cube.filtering.linear_mipmap_linear_linear_clamp
        dEQP-GLES2.functional.texture.vertex.cube.filtering.linear_mipmap_linear_nearest_mirror
        dEQP-GLES2.functional.texture.vertex.cube.wrap.mirror_clamp   
        dEQP-GLES2.functional.negative_api.state.get_framebuffer_attachment_parameteriv
        dEQP-GLES3.functional.shaders.fragdata.invalid_assign_to_1    
        dEQP-GLES31.functional.tessellation.user_defined_io.negative.es32.per_vertex_incorrect_control_explicit_output_array_size_1

        These seem to be flaky on 19.1.. I wasn't able to blame any particular
        commit in the branch for the flakiness.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/mesa-ci-intel/attachments/20191014/4606d3bd/attachment.sig>