ci/concurrency-cancel
PR pipelines cancel superseded runs
| Severity | info |
| Tiers | all |
| Stacks | any |
| Scope | root |
| Kind | expr |
| Source | library v0.1.0 |
| Selected by | fleetlint-recommended@v0.1.0 |
Requirement
GitHub and Gitea workflows triggered by pull requests declare concurrency with cancel-in-progress (top level, or on every job). GitLab merge request pipelines mark their jobs interruptible: true (per job or under default:) and do not set workflow: auto_cancel: on_new_commit: none. Forgejo workflows are not checked.
Why
Every push to a PR otherwise queues another full run behind the obsolete one.
Applies when
workflows().exists(w, "pull_request" in triggers(w) && ci_kind(w) != "forgejo")Check
For each item in:
workflows().filter(w, "pull_request" in triggers(w) && ci_kind(w) != "forgejo")type(yaml(item)) == map && (ci_kind(item) == "gitlab"
? !("workflow" in yaml(item) && type(yaml(item).workflow) == map
&& "auto_cancel" in yaml(item).workflow && type(yaml(item).workflow.auto_cancel) == map
&& "on_new_commit" in yaml(item).workflow.auto_cancel && yaml(item).workflow.auto_cancel.on_new_commit == "none")
&& (jobs(item).all(j, j.interruptible)
|| workflows().exists(g, ci_kind(g) == "gitlab" && type(yaml(g)) == map && "default" in yaml(g)
&& type(yaml(g)["default"]) == map && "interruptible" in yaml(g)["default"]
&& yaml(g)["default"].interruptible == true))
: ("concurrency" in yaml(item) && type(yaml(item).concurrency) == map
&& "cancel-in-progress" in yaml(item).concurrency && yaml(item).concurrency["cancel-in-progress"] != false)
|| ("jobs" in yaml(item) && type(yaml(item).jobs) == map && yaml(item).jobs.size() > 0
&& yaml(item).jobs.all(k, type(yaml(item).jobs[k]) == map && "concurrency" in yaml(item).jobs[k]
&& type(yaml(item).jobs[k].concurrency) == map && "cancel-in-progress" in yaml(item).jobs[k].concurrency
&& yaml(item).jobs[k].concurrency["cancel-in-progress"] != false)))Fix
Add a concurrency block keyed on the workflow and ref with cancel-in-progress: true (GitLab: default: interruptible: true).
Agent instruction: Add concurrency: { group: ${{ github.workflow }}-${{ github.ref }}, cancel-in-progress: true } at the top level of the workflow; for GitLab add interruptible: true under default: and leave deploy jobs non-interruptible if they must finish.
See: https://fleetlint.org/baseline/#continuous-integration
Generated by fleetlint docs from catalog v0.1.0.