ci: fix Check job status to detect skipped results (fixes #2208) by leofang · Pull Request #2210 · NVIDIA/cuda-python

leofang · 2026-06-13T01:29:05Z

Replaces the cancelled || failure predicate in .github/workflows/ci.yml's checks: job with CCCL's check_result pattern: per-dep expected="success", with expected="skipped" for legitimate [doc-only] skips, and an early short-circuit for [no-ci].

Reference: https://github.com/NVIDIA/cccl/blob/8c0e6cb1b6412d7bd0070f9ac3f55fa80231961a/.github/workflows/ci-workflow-pull-request.yml#L463-L526

End-to-end validation (pre-fix vs post-fix reproducer): #2210 (comment)

…esult FixesNVIDIA#2208. The previous `Check job status` body only treated a dep's `result == 'cancelled' || == 'failure'` as failure, letting `'skipped'` slip through silently. When a `build-*` job fails, the dependent `test-*` job is set to `'skipped'` by default needs-failure propagation, and the aggregator passes -- exactly the case demonstrated by NVIDIA#2209. Adopt CCCL's `check_result` pattern: explicit `expected="success"` per dependency, with `expected="skipped"` for legitimate `[doc-only]` skips, and an early short-circuit for `[no-ci]`. Now any deviation from the expected status (including `'skipped'` from a failed upstream) fails the aggregator. Reference: NVIDIA/cccl ci-workflow-pull-request.yml L463-L526.

copy-pr-bot · 2026-06-13T01:29:09Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

leofang · 2026-06-14T05:07:15Z

/ok to test 170c799

leofang · 2026-06-15T17:54:18Z

/ok to test 739c499

Mirrors the NVIDIA#2209 reproducer on this branch's new aggregator. Expected outcome: - All Build * matrix entries fail at the env-vars build step. - Downstream Test * jobs are skipped (cascaded needs-failure). - Check job status now reports failure -- the symmetric, post-fix counterpart to NVIDIA#2209 (where Check job status reported success despite the same set of failures). DO NOT MERGE while this commit is present. Remove before merge. Refs NVIDIA#2208, NVIDIA#2209.

leofang · 2026-06-16T01:36:56Z

/ok to test 811388c

leofang · 2026-06-16T01:57:38Z

Gating-check fix validated end-to-end

The new aggregator catches the exact failure mode that motivated #2208. Demonstrated by running the #2209 reproducer (intentional exit 7 in ci/tools/env-vars) against this branch:

Pre-fix (PR [DO-NOT-MERGE] Demonstrate CI gating-check bug: build fail -> Check job status green #2209, head deddae4794): https://github.com/NVIDIA/cuda-python/actions/runs/27451928558 — 24 Build failures, all Test * wheels skipped, Check job status = success (the bug).
Post-fix (this branch, head 811388c99e): https://github.com/NVIDIA/cuda-python/actions/runs/27588163601 — same 24 Build failures, same Test * wheels skipped, Check job status = failure (the fix working).

Same reproducer, opposite verdict. The skipped-as-success hole is closed.

The demo commit 811388c99e will be reverted next so this PR returns to a mergeable state.

This reverts commit 811388c.

leofang · 2026-06-16T02:01:32Z

/ok to test 7d31a04

rwgk

Looks good to me and codex gpt-5.5:

The new check_result approach is a clear improvement over the previous cancelled || failure checks: it closes the skipped-as-success hole while keeping the intentional [doc-only] and [no-ci] cases explicit.

The one subtle behavior change I noticed is that fork or non-NVIDIA repo runs may now report a red final aggregator where jobs are skipped by owner-gated conditions. I agree that is not a concern for this PR, and it is preferable to avoid weakening the aggregator in a way that could reintroduce false-green CI.

github-actions · 2026-06-16T07:22:23Z

Doc Preview CI
Preview removed because the pull request was closed or merged.

leofang · 2026-06-17T02:28:28Z

@u7k4rs6 please do not fire up nonsense agents against our repo and add noise. This is not a welcomed behavior.

u7k4rs6 · 2026-06-17T04:51:43Z

@leofang Apologies, that was an automated agent I should not have run against your repo, especially on an already-merged PR. It added noise and got the core point wrong. I've turned it off. Thanks for the patience.

* ci: adopt CCCL check_result aggregator and add [doc-only] PR mode Mirrors the gating-check fix in NVIDIA/cuda-python#2210 (root-cause analysis: NVIDIA/cuda-python#2208) and bundles in cuda-python's `[doc-only]` PR-mode feature, which had not been propagated to this repo. Changes: * should-skip: emit both `skip` (existing `[no-ci]` gate) and a new `doc_only` output, computed from PR title containing `[doc-only]`. Single `gh pr view` call now drives both, matching cuda-python's pattern. * `[doc-only]` gating on non-doc-chain jobs: build-llvm-linux-aarch64, build-llvm-windows, build-linux-aarch64, build-windows, and all 11 test-* jobs now have `&& \!fromJSON(needs.should-skip.outputs.doc-only)` in their `if:`. The linux-64 chain (build-llvm-linux-64, build-linux-64) and build-docs are intentionally NOT `[doc-only]`-gated because build-docs needs the linux-64 wheel. * checks: aggregator rewritten to CCCL's `check_result` pattern: explicit `expected="success"` per dependency, anything else (including `'skipped'` from a failed upstream) fails the gate. `[no-ci]` short-circuits at the top; `[doc-only]` flips `expected` to `"skipped"` for every test-* dep. `should-skip` itself is now in `needs:` and checked, so a crash in the gating job can't masquerade as a green CI run. Verified the aggregator script with a local simulation across 8 scenarios: normal-green, build-fail-cascade, `[no-ci]`, `[doc-only]`-OK, `[doc-only]`-doc-fails, build-windows-fail-cascade, single-test-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/cuda-python#2208, NVIDIA/cuda-python#2210 * Bump the actions-monthly group across 1 directory with 5 updates (#125) Bumps the actions-monthly group with 5 updates in the / directory: | Package | From | To | | --- | --- | --- | | [astral-sh/setup-uv](https://github.com/astral-sh/setup-uv) | `8.1.0` | `8.2.0` | | [github/codeql-action](https://github.com/github/codeql-action) | `4.35.3` | `4.36.2` | | [pypa/cibuildwheel](https://github.com/pypa/cibuildwheel) | `3.4.1` | `4.1.0` | | [prefix-dev/setup-pixi](https://github.com/prefix-dev/setup-pixi) | `0.9.5` | `0.9.6` | | [test-summary/action](https://github.com/test-summary/action) | `2.4` | `2.6` | Updates `astral-sh/setup-uv` from 8.1.0 to 8.2.0 - [Release notes](https://github.com/astral-sh/setup-uv/releases) - [Commits](astral-sh/setup-uv@0880764...fac544c) Updates `github/codeql-action` from 4.35.3 to 4.36.2 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](github/codeql-action@e46ed2c...8aad20d) Updates `pypa/cibuildwheel` from 3.4.1 to 4.1.0 - [Release notes](https://github.com/pypa/cibuildwheel/releases) - [Changelog](https://github.com/pypa/cibuildwheel/blob/main/docs/changelog.md) - [Commits](pypa/cibuildwheel@8d2b08b...2947353) Updates `prefix-dev/setup-pixi` from 0.9.5 to 0.9.6 - [Release notes](https://github.com/prefix-dev/setup-pixi/releases) - [Commits](prefix-dev/setup-pixi@1b2de7f...5185adf) Updates `test-summary/action` from 2.4 to 2.6 - [Release notes](https://github.com/test-summary/action/releases) - [Commits](test-summary/action@v2.4...v2.6) --- updated-dependencies: - dependency-name: astral-sh/setup-uv dependency-version: 8.2.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions-monthly - dependency-name: github/codeql-action dependency-version: 4.36.2 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions-monthly - dependency-name: prefix-dev/setup-pixi dependency-version: 0.9.6 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: actions-monthly - dependency-name: pypa/cibuildwheel dependency-version: 4.0.0 dependency-type: direct:production update-type: version-update:semver-major dependency-group: actions-monthly - dependency-name: test-summary/action dependency-version: '2.6' dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions-monthly ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * ci: opt out of delvewheel repair on Windows (cibuildwheel v4 default) cibuildwheel v4 made `delvewheel repair -w {dest_dir} -v {wheel}` the default for `CIBW_REPAIR_WHEEL_COMMAND_WINDOWS`; v3.x had no default. We intentionally do not need wheel repair on Windows because the wheel is built with `/MT` (static CRT) per #127 -- it is self-contained and has no DLL dependencies for delvewheel to bundle. Without this opt-out, the default delvewheel invocation discovers an unresolvable reference to the project's own internal nanobind module and fails wheel build for every win-64 matrix entry. This regression re-surfaced when the cibuildwheel v3.4.1 -> v4.1.0 bump from #125 was cherry-picked into this branch to validate the new aggregator catches it. With the opt-out, the build succeeds again. Refs: #127 (static-CRT Windows wheel design), #125 / #134 (original cibuildwheel bump + revert). --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Atmn Patel <atmnp@nvidia.com>

…) (#912) ## 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 [`5f530b96`](5f530b96) auto-merged with a silent Windows failure and had to be reverted in [`441882b2`](441882b2). ## Change One file, `.github/workflows/ci.yaml`, `checks:` job only: - Replace the `cancelled || failure` predicate with CCCL's `check_result` pattern: per-dep `expected="success"`, `[no-ci]` short-circuit at the top. - Add `should-skip` to `needs:` and check its result explicitly — a crash in the gating job can no longer masquerade as a green CI run. - No `[doc-only]` machinery — this repo doesn't have that PR mode. If added later, mirror NVIDIA/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 - Tracking issue on this repo: #900 - Root cause: NVIDIA/cuda-python#2208 - Source-of-truth fix: NVIDIA/cuda-python#2210 - Sibling port to numba-cuda-mlir: NVIDIA/numba-cuda-mlir#142 - CCCL gate pattern: https://github.com/NVIDIA/cccl/blob/8c0e6cb1b6412d7bd0070f9ac3f55fa80231961a/.github/workflows/ci-workflow-pull-request.yml#L463-L526

github-actionsBot added the CI/CD CI/CD infrastructure label Jun 13, 2026

This comment has been minimized.
Sign in to view

leofang mentioned this pull request Jun 14, 2026
tests: fix parametrize patterns rejected by pytest 9.1.0 #2212
Merged

Merge branch 'main' into leofang/fix-issue-2208-gating-check
739c499

leofang added bug Something isn't working P0 High priority - Must do! labels Jun 15, 2026

leofang added this to the cuda.core v1.1.0 milestone Jun 15, 2026

leofang added 2 commits June 16, 2026 01:36

Merge branch 'main' into leofang/fix-issue-2208-gating-check
5983cbe

Revert "[demo] Re-inject env-vars exit 7 to validate gating-check fix"
7d31a04
This reverts commit 811388c.

leofang self-assigned this Jun 16, 2026

leofang marked this pull request as ready for review June 16, 2026 02:14

rwgk approved these changes Jun 16, 2026
View reviewed changes

rwgk merged commit 671af69 into NVIDIA:mainJun 16, 2026
203 of 206 checks passed

leofang mentioned this pull request Jun 16, 2026
ci: adopt CCCL check_result aggregator and add doc-only PR mode NVIDIA/numba-cuda-mlir#142
Merged

leofang deleted the leofang/fix-issue-2208-gating-check branch June 16, 2026 12:20

This comment was marked as duplicate.
Sign in to view

This comment was marked as outdated.
Sign in to view

leofang mentioned this pull request Jun 25, 2026
CI: Check job status aggregator passes when build jobs fail (port of NVIDIA/cuda-python#2208) NVIDIA/numba-cuda#900
Closed

leofang mentioned this pull request Jul 4, 2026
ci: adopt CCCL check_result aggregator for Check job status (fixes #900) NVIDIA/numba-cuda#912
Merged

Navigation Menu

Uh oh!

ci: fix Check job status to detect skipped results (fixes #2208) - #2210

ci: fix Check job status to detect skipped results (fixes #2208)#2210
rwgk merged 5 commits into
NVIDIA:mainfrom
leofang:leofang/fix-issue-2208-gating-check

leofang commented Jun 13, 2026•
edited
Loading

Uh oh!

copy-pr-botBot commented Jun 13, 2026

Uh oh!

leofang commented Jun 14, 2026

Uh oh!

This comment has been minimized.

leofang commented Jun 15, 2026

Uh oh!

leofang commented Jun 16, 2026

Uh oh!

leofang commented Jun 16, 2026•
edited
Loading

Uh oh!

leofang commented Jun 16, 2026

Uh oh!

rwgk left a comment

Uh oh!

Uh oh!

github-actionsBot commented Jun 16, 2026

Uh oh!

This comment was marked as duplicate.

Uh oh!

This comment was marked as outdated.

Uh oh!

leofang commented Jun 17, 2026

Uh oh!

u7k4rs6 commented Jun 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

leofang commented Jun 13, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

copy-pr-botBot commented Jun 13, 2026

Uh oh!

leofang commented Jun 14, 2026

Uh oh!

This comment has been minimized.

leofang commented Jun 15, 2026

Uh oh!

leofang commented Jun 16, 2026

Uh oh!

leofang commented Jun 16, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Gating-check fix validated end-to-end

Uh oh!

leofang commented Jun 16, 2026

Uh oh!

rwgk left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actionsBot commented Jun 16, 2026

Uh oh!

This comment was marked as duplicate.

Uh oh!

This comment was marked as outdated.

Uh oh!

leofang commented Jun 17, 2026

Uh oh!

u7k4rs6 commented Jun 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

leofang commented Jun 13, 2026•
edited
Loading

leofang commented Jun 16, 2026•
edited
Loading