Fix WSL support, update troubleshooting guide

[Wumpf]

Related

• Fixes rerun 0.21.0 WGPU Error on WSL #8564
• Some steps towards Make graphics adapter creation list all adapters one by one and inform why they weren't picked #8475
• Dependent on Extend WgpuSetup, egui_kittest now prefers software rasterizers for testing emilk/egui#5506
• Dependent on Update egui emilk/egui_plot#67

What

What's broken

Since Rerun 0.19 we picking in 2D & 3D view was broken & accompanied by log spam.

The reason we failed 0.19+ is unknown but likely either that something caused us picking the "webgl-like" compatibility tier or a change in wgpu no longer allows us to texture copy depth buffers.

Rerun 0.21 then broke startup entirely. This happens due to two distinct problems

• we added a check that foresees the picking issue, aborts application start and tells the user to update their driver (which.. as far as I tried doesn't help)
• wgpu did a validation related thing that that accidentally makes us fail startup because it makes the assumption that compute shader support == glsl support for buffers with runtime size (very very reasonable, but also wrong it turns out 😭 WSL2 GL device fails creation gfx-rs/wgpu#6841)

This Fix

I decided to move forward more aggressively in a manner that leaves all these intertwined issues behind and sets us up better for simpler & more advanced rendering in the future:

• the troubleshooting guide now explains how to update the mesa drivers
• we configure wgpu now such that it allows picking those non-compliant Vulkan drivers

-> As a result we can draw with Vulkan rather than a (typically) outdated GL driver which is in line with us moving slowly towards WebGPU support as the minspec.
(See code comments for remaining weirdness ;-))

Turns out another benefit with this is that multi-gpu setups behave like on the host now: Previously, MESA_D3D12_DEFAULT_ADAPTER_NAME had to be set to pick which adapter is used. Now all adapters are passed through WSL correctly, meaning that our regular adapter selection logic (which is to be further improved) works fine, i.e. it prefers discrete over integrated GPU unless WGPU_POWER_PREF=low is set.

Tested on WSL with

• Ubuntu 22.04
• Ubuntu 24.04

previous tests indicate that Ubuntu 20 should behave similar to 22.

Future work

Currently, there's a little bit of not super straight forward logging left from wgpu, warning about the wild GPU setup, but I actually feel better leaving this in for now - after all the WSL rendering stack is still rather adventurous and we should keep encouraging using the host system directly.

[2025-01-07T20:49:52Z INFO  re_sdk_comms::server] Hosting a SDK server over TCP at 0.0.0.0:9876. Connect with the Rerun logging SDK.
[2025-01-07T20:49:52Z WARN  wgpu_hal::vulkan::instance] InstanceFlags::VALIDATION requested, but unable to find layer: VK_LAYER_KHRONOS_validation
[2025-01-07T20:49:53Z WARN  wgpu_hal::gles::egl] Re-initializing Gles context due to Wayland window
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
[2025-01-07T20:49:54Z WARN  wgpu_hal::vulkan::adapter] Adapter is not Vulkan compliant: Microsoft Direct3D12 (NVIDIA GeForce RTX 4070)
[2025-01-07T20:49:54Z WARN  wgpu_hal::vulkan::adapter] Adapter is not Vulkan compliant: Microsoft Direct3D12 (AMD Radeon(TM) Graphics)
[2025-01-07T20:49:54Z WARN  wgpu_hal::gles::adapter] Max vertex attribute stride unknown. Assuming it is 2048
[2025-01-07T20:49:54Z WARN  wgpu_hal::vulkan::adapter] Adapter is not Vulkan compliant: Microsoft Direct3D12 (NVIDIA GeForce RTX 4070)
[2025-01-07T20:49:54Z WARN  wgpu_hal::vulkan::adapter] Adapter is not Vulkan compliant: Microsoft Direct3D12 (AMD Radeon(TM) Graphics)
[2025-01-07T20:49:54Z WARN  wgpu_hal::gles::adapter] Max vertex attribute stride unknown. Assuming it is 2048
[2025-01-07T20:49:54Z INFO  egui_wgpu] There were 4 available wgpu adapters: {backend: Vulkan, device_type: DiscreteGpu, name: "Microsoft Direct3D12 (NVIDIA GeForce RTX 4070)", driver: "Dozen", driver_info: "Mesa 24.3.3 - kisak-mesa PPA", vendor: NVIDIA (0x10DE), device: 0x2786}, {backend: Vulkan, device_type: IntegratedGpu, name: "Microsoft Direct3D12 (AMD Radeon(TM) Graphics)", driver: "Dozen", driver_info: "Mesa 24.3.3 - kisak-mesa PPA", vendor: AMD (0x1002), device: 0x164E}, {backend: Vulkan, device_type: Cpu, name: "llvmpipe (LLVM 15.0.7, 256 bits)", driver: "llvmpipe", driver_info: "Mesa 24.3.3 - kisak-mesa PPA (LLVM 15.0.7)", vendor: Mesa (0x10005)}, {backend: Gl, device_type: Other, name: "D3D12 (AMD Radeon(TM) Graphics)", driver_info: "4.2 (Core Profile) Mesa 23.2.1-1ubuntu3.1~22.04.3"}
[2025-01-07T20:49:54Z WARN  wgpu_core::instance] Missing downlevel flags: DownlevelFlags(FULL_DRAW_INDEX_UINT32 | SURFACE_VIEW_FORMATS)
   The underlying API or device in use does not support enough features to be a fully compliant implementation of WebGPU. A subset of the features can still be used. If you are running this program on native and not in a browser and wish to limit the features you use to the supported subset, call Adapter::downlevel_properties or Device::downlevel_properties to get a listing of the features the current platform supports.
[2025-01-07T20:49:54Z WARN  wgpu_core::instance] DownlevelCapabilities {
       flags: DownlevelFlags(
           COMPUTE_SHADERS | FRAGMENT_WRITABLE_STORAGE | INDIRECT_EXECUTION | BASE_VERTEX | READ_ONLY_DEPTH_STENCIL | NON_POWER_OF_TWO_MIPMAPPED_TEXTURES | CUBE_ARRAY_TEXTURES | COMPARISON_SAMPLERS | INDEPENDENT_BLEND | VERTEX_STORAGE | ANISOTROPIC_FILTERING | FRAGMENT_STORAGE | MULTISAMPLED_SHADING | DEPTH_TEXTURE_AND_BUFFER_COPIES | WEBGPU_TEXTURE_FORMAT_SUPPORT | BUFFER_BINDINGS_NOT_16_BYTE_ALIGNED | UNRESTRICTED_INDEX_BUFFER | DEPTH_BIAS_CLAMP | VIEW_FORMATS | UNRESTRICTED_EXTERNAL_TEXTURE_COPIES | NONBLOCKING_QUERY_RESOLVE | VERTEX_AND_INSTANCE_INDEX_RESPECTS_RESPECTIVE_FIRST_VALUE_IN_INDIRECT_DRAW,
       ),
       limits: DownlevelLimits,
       shader_model: Sm5,
   }

(what does bug me is the duplicated messages, may hint at some issues in either wgpu or our init dance... but that's a story for another time)

[github-actions]

Web viewer built successfully. If applicable, you should also test it:

• I have tested the web viewer

Result	Commit	Link	Manifest
✅	0d274bb	https://rerun.io/viewer/pr/8610	+nightly +main

^{Note: This comment is updated whenever you push a commit.}

[github-actions]

Latest documentation preview deployed successfully.

Result	Commit	Link
✅	0d274bb	https://landing-gh51uqs66-rerun.vercel.app/docs

^{Note: This comment is updated whenever you push a commit.}

[emilk]

What happens if we do?

[Wumpf]

hum. excellent question. I'll find out!

[Wumpf]

seems to be undefined behavior, couldn't find anything on it.
Clairifed comment though a bit: This is not the number of indices but the value for the indices, i.e. the vertex count is limited to 16.8mio if this flag is not present on the adapter