ci: add cuda type-check job so CUDA-only borrowck errors fail fast
Some checks failed
build-prerelease / Resolve version stamps (push) Successful in 30s
CI / Format (push) Successful in 37s
CI / CUDA type-check (push) Failing after 3m8s
CI / Clippy (push) Successful in 2m27s
build-prerelease / Build neuron-blackwell (push) Successful in 5m46s
build-prerelease / Build cortex binary (push) Successful in 5m0s
build-prerelease / Build neuron-ampere (push) Successful in 7m39s
CI / Test (push) Successful in 5m37s
CI / Build cortex SRPM (push) Has been skipped
CI / Build neuron SRPM (push) Has been skipped
CI / Publish cortex to COPR (push) Has been skipped
CI / Publish neuron to COPR (push) Has been skipped
CI / Bump version in source (push) Has been skipped
build-prerelease / Package cortex RPM (push) Successful in 1m33s
build-prerelease / Build neuron-ada (push) Successful in 5m12s
build-prerelease / Package helexa-neuron-ada RPM (push) Successful in 3m0s
build-prerelease / Package helexa-neuron-ampere RPM (push) Successful in 3m8s
build-prerelease / Package helexa-neuron-blackwell RPM (push) Successful in 3m43s
build-prerelease / Publish to rpm.lair.cafe (unstable) (push) Successful in 1m9s

Run 244 caught a use-of-moved-value in a `#[cfg(feature = "cuda")]`
block that the default-feature workspace clippy/test gate had no
chance of seeing. The error appeared only when the RPM build
workflow compiled with `--features cuda` — 30+ minutes after push.

Add a `cuda-check` job to ci.yml that runs `cargo check -p neuron
--features cuda --all-targets` on the rpm runner (where nvcc /
cudarc build deps live; the generic `rust` runner doesn't have
them). Borrow-check only — we never run tests here, the runner
has no GPU. Same retry pattern as clippy/test.

Both SRPM jobs (`srpm-cortex`, `srpm-neuron`) now gate on
`cuda-check` so a CUDA build break can't reach the release pipeline.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
2026-05-31 09:49:51 +03:00
parent 6927286cab
commit 1859777332

View File

@@ -92,10 +92,40 @@ jobs:
exit 1
- run: sccache --show-stats
# Type-check the CUDA-only code path. Borrow-check-only — we
# never run the tests here (the runner has no GPU). This catches
# the category of bug where a refactor compiles fine under the
# default feature set (which is what the `clippy` and `test` jobs
# exercise) but fails inside a `#[cfg(feature = "cuda")]` block.
# Lives on `runs-on: rpm` because that's where nvcc / cudarc's
# build prerequisites are installed; the generic `rust` runner
# doesn't have them.
cuda-check:
name: CUDA type-check
runs-on: rpm
steps:
- uses: actions/checkout@v4
- name: cargo check --features cuda (with retry)
run: |
for attempt in 1 2 3; do
echo "::group::cuda-check attempt ${attempt}"
if cargo check -p neuron --features cuda --all-targets; then
echo "::endgroup::"
exit 0
fi
echo "::endgroup::"
echo "cuda-check failed on attempt ${attempt}"
if [ "${attempt}" -lt 3 ]; then
sleep 5
fi
done
echo "cuda-check failed after 3 attempts"
exit 1
srpm-cortex:
name: Build cortex SRPM
runs-on: rpm
needs: [fmt, clippy, test]
needs: [fmt, clippy, test, cuda-check]
if: startsWith(github.ref, 'refs/tags/v')
steps:
- uses: actions/checkout@v4
@@ -155,7 +185,7 @@ jobs:
srpm-neuron:
name: Build neuron SRPM
runs-on: rpm
needs: [fmt, clippy, test]
needs: [fmt, clippy, test, cuda-check]
if: startsWith(github.ref, 'refs/tags/v')
steps:
- uses: actions/checkout@v4