12:48 <enunes> rellla: we have some tests that differ between CI and local runs?

15:27 <rellla> enunes: i guess there is some issue with DIRTY flags

15:28 <rellla> what do the *caselist.txt files in the artifacts mean?

15:39 <rellla> i guess the difference is how i run deqp. i don't use deqp-runner...

15:52 <rellla> dEQP-GLES2.functional.clipping.polygon.multiple_0_viewport_corner for example seems to fail right after the viewport was changed in a wide_line test. i guess i'm missing the viewport again to the original values!?

15:53 <rellla> *to set

16:11 <rellla> enunes: https://gitlab.freedesktop.org/rellla/mesa/-/commits/lima-vp_test is a quick hack to it.

16:12 <rellla> though i might need to think about how to do it a bit more pretty ...

16:21 <enunes> rellla: hmm strange, so you managed to reproduce it locally?

16:21 <enunes> I tried your branch and the CI failure didnt reproduce for me

16:32 <rellla> enunes: i ran the pipeline with my fix and that succeeded, see https://gitlab.freedesktop.org/rellla/mesa/-/pipelines/433857

16:33 <rellla> do you have deqp-runner locally?

16:33 <rellla> i need to set it up here first.

16:39 <anarsoul> rellla: yeah, you need to reset view port back to original values if it's not a line draw