The following tests skip when run with device override, but fail when run on the real hardware: - piglit.spec.ext_shader_samples_identical.glsl-es-3_10.compiler.all-functions_vert - piglit.spec.oes_shader_io_blocks.compiler.layout-location-aliasing_vert This breaks Intel CI automation designed to bisect failures. Until this bug is resolved, the tests will be disabled in CI.
This is actually not a piglit issue - it's an issue with how the intel driver does feature enablement. A bunch of features depend on various kernel features, and so the driver does a bunch of "exec" tests on startup, to see what all it can turn on. When you do a device override, it still runs those exec tests, which will generally fail. I think the solution is to assume all the exec tests pass if there's a device override in place.
There is the possiblility that there is somethign with piglit's fast skip actually. We fixed a bug in the python shader_runner layer a bit ago in 892f6e44723eda8d, and I noticed that the fast skip isn't exactly correct, (It always assumes that the requirements are > or >= and never < or <=), so it wouldnt' suprise me if there are more limitaitons beyond that. However, Illia's explination of what's happening makes sense in conjunction with the way fast skipping works (it relies on wflinfo) so if the driver doesn't expose the same feature set with devid overrides it would explain why the tests are being incorrectly marked skip.
-- GitLab Migration Automatic Message -- This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity. You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/piglit/issues/4.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.