New GL Conformance tests fail on all intel platforms:
Note that constant sequence *passes* for es2
Git revision is from master, 93161be. However the tests fail as far back as the 10.6 branch point.
The tesst are nor run in the previous Intel internal port of the Khronos CTS suit.
piglit.es2-cts.shaders.negative.initialize and piglit.es3-cts.shaders.negative.initialize are fixed by the following patch on the mesa-dev list for review:
piglit.es3-cts.shaders.negative.constant_sequence is going to require a bit more effort, I think. :(
with your patches:
on top off: Mesa git@aee28a0aa
pass on HSW and BDW.
These should all be fixed by the following series that was pushed to master a couple hours ago. These should all eventually get picked to 11.0 and 10.6 stable branches.
92635a8 glsl: In later GLSL versions, sequence operator is cannot be a constant expression
05e4601 glsl: Add method to determine whether an expression contains the sequence operator
bb329f2 glsl: Restrict initializers for global variables to constant expression in ES
3524d6d glsl: Only set ir_variable::constant_value for const-decorated variables
5bc68f0 glsl: Use constant_initializer instead of constant_value to determine whether to keep an unused uniform
313372c glsl/linker: Use constant_initializer instead of constant_value to initialize uniforms
8acce5d ff_fragment_shader: Use binding to set the sampler unit
43b07eb glsl: Allow built-in functions as constant expressions in OpenGL ES 1.00
I verified the fix.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct.