Description
Downstream issue gfx-rs/wgpu#9514
On metal-26 runners we use various components from the MetalToolchain component to validate output shaders. In specific the first command we run is xcrun -sdk macosx metal '-mmacosx-version-min=10.11' '-std=macos-metal1.0' '-Werror=constant-conversion' -x metal tests/out/msl/spv-empty-global-name.metal -o /dev/null.
Platforms affected
Runner images affected
Image version and build link
Current runner version: '2.334.0'
The workflow is available here https://github.com/gfx-rs/wgpu/blob/trunk/.github/workflows/shaders.yml#L70-L92
Latest failed build available here https://github.com/gfx-rs/wgpu/actions/runs/25444343504/job/74644143404
Is it regression?
Yes. Previously (not sure exactly when) the job always succeeds, now it only sometimes succeeds.
Expected behavior
The job functioning always.
Actual behavior
Most of the time this succeeds and everything works. Seemingly at random we get the following error:
Updating crates.io index
Compiling hlsl-snapshots v29.0.0 (/Users/runner/work/wgpu/wgpu/naga/hlsl-snapshots)
Compiling naga-xtask v0.1.0 (/Users/runner/work/wgpu/wgpu/naga/xtask)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 5.17s
Running `/Users/runner/work/wgpu/wgpu/target/debug/naga-xtask validate msl`
Error: naga_xtask::validate] `xcrun -sdk macosx metal '-mmacosx-version-min=10.11' '-std=macos-metal1.0' '-Werror=constant-conversion' -x metal tests/out/msl/spv-empty-global-name.metal -o /dev/null` failed to run; exit code: Some(1)
stdout:
stderr:
error: error: cannot execute tool 'metal' due to missing Metal Toolchain; use: xcodebuild -downloadComponent MetalToolchain
Repro steps
Make a job that relies on xcrun -sdk macosx metal and run over time until you get this error. There's no decernable "this happens then we get a failure", it is random as far as I can tell, maybe about which runner it get allocated.
Description
Downstream issue gfx-rs/wgpu#9514
On
metal-26runners we use various components from the MetalToolchain component to validate output shaders. In specific the first command we run isxcrun -sdk macosx metal '-mmacosx-version-min=10.11' '-std=macos-metal1.0' '-Werror=constant-conversion' -x metal tests/out/msl/spv-empty-global-name.metal -o /dev/null.Platforms affected
Runner images affected
Image version and build link
Current runner version: '2.334.0'
The workflow is available here https://github.com/gfx-rs/wgpu/blob/trunk/.github/workflows/shaders.yml#L70-L92
Latest failed build available here https://github.com/gfx-rs/wgpu/actions/runs/25444343504/job/74644143404
Is it regression?
Yes. Previously (not sure exactly when) the job always succeeds, now it only sometimes succeeds.
Expected behavior
The job functioning always.
Actual behavior
Most of the time this succeeds and everything works. Seemingly at random we get the following error:
Repro steps
Make a job that relies on
xcrun -sdk macosx metaland run over time until you get this error. There's no decernable "this happens then we get a failure", it is random as far as I can tell, maybe about which runner it get allocated.