Repository navigation
ci: adopt CCCL check_result aggregator for Check job status (fixes #900) - #912
Merged
leofang merged 1 commit intoJul 5, 2026
Merged
Conversation
…IDIA#900) Ports the gating-check fix from NVIDIA/cuda-python#2210 (root-cause analysis: NVIDIA/cuda-python#2208) to this repo. Same bug bit us in this repo on 2026-07-03 when a dependabot actions-monthly bump auto-merged with a silent Windows failure and had to be reverted. The previous `cancelled || failure` predicate on the aggregator's `test-*` / `build-docs` / `coverage-report` deps missed the case where an upstream `build-*` job fails and the downstream job's `result` is set to `'skipped'` via default needs-failure propagation. `'skipped'` is not `'cancelled'` or `'failure'`, so the gate exited 0 and merge was allowed. Adopt CCCL's `check_result` pattern: require an explicit `expected="success"` per dependency; anything else (including `skipped` from a failed upstream) fails the gate. `[no-ci]` short-circuits at the top. `should-skip` is now itself in `needs:` and checked, so a crash in the gating job can't masquerade as green. Scope note: this repo has no `[doc-only]` PR mode wired up, so no dynamic `expected` handling is needed. If `[doc-only]` is added later, mirror the pattern from NVIDIA/numba-cuda-mlir#142. Verified the aggregator script with a local simulation across 6 scenarios: normal-green, build-windows-fail-cascade, build-linux-64 fail-cascade, [no-ci], pre-commit-fail, should-skip-itself-fails. All behave correctly. Reference: CCCL gate, https://github.com/NVIDIA/cccl/blob/8c0e6cb1b6412d7bd0070f9ac3f55fa80231961a/.github/workflows/ci-workflow-pull-request.yml#L463-L526 Refs: NVIDIA#900, NVIDIA/cuda-python#2208, NVIDIA/cuda-python#2210, NVIDIA/numba-cuda-mlir#142
Member
Author
|
/ok to test 3e87d3e |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Ports the gating-check fix from NVIDIA/cuda-python#2210 to this repo. Closes #900.
Root cause and design discussion live in NVIDIA/cuda-python#2208 — not repeating here. Bit us in this repo on 2026-07-03 when
5f530b96auto-merged with a silent Windows failure and had to be reverted in441882b2.Change
One file,
.github/workflows/ci.yaml,checks:job only:cancelled || failurepredicate with CCCL'scheck_resultpattern: per-depexpected="success",[no-ci]short-circuit at the top.should-skiptoneeds:and check its result explicitly — a crash in the gating job can no longer masquerade as a green CI run.[doc-only]machinery — this repo doesn't have that PR mode. If added later, mirror ci: adopt CCCL check_result aggregator and add doc-only PR mode numba-cuda-mlir#142.Verification
Local simulation of the new aggregator across 6 scenarios (normal-green, build-windows-fail-cascade, build-linux-64-fail-cascade,
[no-ci], pre-commit-fail, should-skip-itself-fails) — all behave as expected.Reference