10:29 sarbes: Done and done.
18:46 anarsoul: sarbes: looks like your constant LOD MR breaks something. I'd suggest running dEQP locally ;)
18:48 sarbes: Yeah, seen that 🙈
18:51 anarsoul: in particular, cubemap is broken
18:52 anarsoul: prior to your change, constant_lod_bias for cubemap was always -1
18:52 sarbes: Maybe I've got the -1 bias wrong.
18:52 anarsoul: but now it's set only if (instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE && instr->src[i].src_type == nir_tex_src_bias)
18:53 sarbes: DOH.
18:55 anarsoul: there are also 3 crashes
21:40 sarbes: I'm kicking the can when it comes to running deqp tests. I've never managed to compile it.
21:43 sarbes: I'll start anew, but with a larger SD card this time. But it will take some time until I figure things out.