08:22zzag: emersion: https://gitlab.freedesktop.org/mesa/libdrm/-/merge_requests/463 removed the license text, was that intentional?
08:23zzag: ah
08:23zzag: I'm blind
08:23zzag: SPDX-License-Identifier: MIT
09:09emersion: yeah
09:10emersion: the kernel is replacing license texts with SPDX comments
11:59anonymix007[m]: I'm trying to port ANV to Windows. Some apps such as vkcube and vkgears already work, even FurMark GL works via Zink, but the "native" VK FurMark demo has issues (toroid is entirely black). I also ran some CTS tests and saw a lot of different failures (see https://paste.debian.net/hidden/9cd568df for 'dEQP-VK.geometry.*' results). Are there some self tests in ANV which would pinpoint me to the actual problem? Or how would one
11:59anonymix007[m]: generally debug this sort of issues?
12:56zmike: probably try to create some kind of packet dumper for cmdbufs and compare against the windows proprietary driver
13:48anonymix007[m]: Will running it with `INTEL_DEBUG=bat` and comparing it to ANV in Linux do?
13:51anonymix007[m]: I'm not trying to make ANV compatible with the proprietary KMD, but rather trying to make it work in Windows VMs via DRM native contexts, so I don't think comparison with the proprietary driver makes much sense here.
14:08Hazematman: anonymix007[m]: I think the best thing is trying to build intuition about what could be failing in a test and how that translate to what the GPU is actually doing. There is a guide about debugging here giving tips on what to try and tweak to be understand what could be failing https://docs.mesa3d.org/graphics-debugging/debugging-misrenderings-crashes.html
14:17zmike: mareko: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/42750
15:05anonymix007[m]: INTEL_DEBUG=bat produced identical dumps for dEQP-VK.geometry.basic.side_effect_with_condition. I assume this means that the shaders aren't miscompiled, right?
16:55zmike: mareko: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/42760
16:55zmike: also I have finally stopped using your amd address
17:02karolherbst: Anybody want to review my nir changes in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/42646 ? It's adding a helper to check if a nir shader is completely linked (as in all referenced functions starting from the entry point have an implementation)