fix(rt-v1): restore jellyfish-deploy-notification as no-op stub#153
Conversation
Surfaced by encodium/internal_api run 29114404344 (hotfix-SHOP-5216 workflow_dispatch → startup_failure). GH's workflow_dispatch runs startup validation over the entire reusable-workflow tree even for jobs gated `if:` off. A stale `uses: encodium/.github/.github/workflows/ jellyfish-deploy-notification.yaml@main` on a hotfix branch's deploy-eks.yaml — cut BEFORE the sweep merged today — resolves to this now-missing file, and startup validation fails before any job runs. Any long-lived branch across the org has the same footgun. This restores the workflow as a NO-OP stub: same input/secret signature as the deleted original so callers resolve cleanly, but the job body does nothing except log a deprecation ::notice::. No API call, no side effects. Long-lived branches with stale references stop hitting startup_failure immediately. As they merge/rebase, they naturally pick up the sweep from #152. Once branches have cycled (~1 month of active churn), we can delete this stub for real. The stub is a bandage, not a rollback of the retirement. All 22 participant sweep PRs stay merged; per-repo `jellyfish-notify` jobs stay gone.
PR SummaryLow Risk Overview The stub keeps the same Per-repo Jellyfish job removals from the org sweep are unchanged; this is a compatibility bridge until branch churn drops stale references (~1 month), after which the stub can be deleted again. Reviewed by Cursor Bugbot for commit 50509d2. Bugbot is set up for automated code reviews on this repo. Configure here. |
Symptom
encodium/internal_api run 29114404344 — hotfix-SHOP-5216
workflow_dispatch→startup_failure.Root cause:
hotfix-SHOP-5216was cut from main BEFORE today's sweep merged. Itsdeploy-eks.yaml(line 27) still hasuses: encodium/.github/.github/workflows/jellyfish-deploy-notification.yaml@main. GH runs startup validation over the entire reusable-workflow tree even for jobs that would beif:-gated off. Since I deleted the shared workflow in #152 (a3b0afc), tree resolution fails before any job runs.Every long-lived branch across the org that predates the sweep (feature branches, hotfix-*, dependabot PRs, in-flight work) has the same footgun.
Fix (this PR)
Restore
jellyfish-deploy-notification.yamlas a no-op stub:workflow_callinput + secret signature as the deleted original → callers resolve cleanly::notice::pointing at the retirement historyLong-lived branches stop hitting
startup_failureimmediately.Not this PR — the retirement itself stands
jellyfish-notifyjob removals stay in placeDelete-for-real timeline
Once org-wide branch churn has cycled through (rough guideline: ~1 month), delete this stub in a follow-up. Anything actively worked on will have naturally rebased/merged and dropped the stale reference.