|
1 | 1 | # Turn the outcome of the expected-failures strict array-API jobs into |
2 | 2 | # check runs on the PR. This runs as a separate workflow_run workflow because |
3 | 3 | # pull_request runs for fork PRs get a read-only GITHUB_TOKEN and cannot |
4 | | -# create check runs themselves. Each check uses a stable name (the matrix |
5 | | -# entry name); the outcome is carried by the conclusion and output instead. |
| 4 | +# create check runs themselves. Each check uses a stable name ("strict-status |
| 5 | +# / " plus the matrix entry name); the outcome is carried by the conclusion |
| 6 | +# and output instead. |
6 | 7 | # These checks are informational only (an expected test failure is reported |
7 | 8 | # as neutral, so the PR rollup stays green) and are not intended to be |
8 | 9 | # required checks while the failures are expected. |
@@ -126,13 +127,16 @@ jobs: |
126 | 127 | 'complete its tests step. This is an infrastructure error in the ' + |
127 | 128 | 'job itself, not a test failure; see the run log for details.'; |
128 | 129 | } |
129 | | - // Stable check name (just the matrix entry name); the outcome |
130 | | - // lives in the conclusion and output so the checks UI does not |
131 | | - // accumulate differently-named checks per outcome. |
| 130 | + // Stable check name; the outcome lives in the conclusion and |
| 131 | + // output so the checks UI does not accumulate differently-named |
| 132 | + // checks per outcome. The prefix keeps the name distinct from |
| 133 | + // the CI job's own check (e.g. "ubuntu-py313-strict") -- the PR |
| 134 | + // merge box shows only one check per name, so an identically |
| 135 | + // named check run would be hidden behind the CI job entry. |
132 | 136 | await github.rest.checks.create({ |
133 | 137 | owner: context.repo.owner, |
134 | 138 | repo: context.repo.repo, |
135 | | - name, |
| 139 | + name: `strict-status / ${name}`, |
136 | 140 | head_sha: run.head_sha, |
137 | 141 | status: 'completed', |
138 | 142 | conclusion, |
|
0 commit comments