mirror of
https://github.com/openobserve/openobserve.git
synced 2026-11-04 16:09:05 +00:00
## In plain words The ENT e2e gate has been posting **false "🚫 Needs Review" failures** on real PRs. None of these were actual test failures — three separate bugs. (An earlier fix on #13284 was lost when that PR was recreated as #13318 without the workflow change, so this re-applies them cleanly on their own branch.) ## The three fixes | # | Bug | Symptom | Fix | |---|-----|---------|-----| | 1 | Always built against ENT **`main`** | Cross-repo feature PRs (e.g. Workflows v1) failed to **compile** — `cannot find function add_workflow` — because the enterprise code lives on the matching ENT branch, not main | Resolve a **same-named ENT branch** and build against it if it exists, else `main` (mirrors ENT's own `playwright.yml`) | | 2 | Correlation grabbed **stale runs** | Gate "completed" in ~3s reporting a **hours-old** run's failure for the same PR | Stamp dispatch time; only accept a run **created at/after** it, newest first | | 3 | `cancelled` treated as failure | A newer push superseded a run (engine `cancel-in-progress`) → false "specs failed" comment | Treat `cancelled`/`skipped`/empty as **neutral** — defer to the newer run, don't comment or fail | Only a genuine `failure`/`timed_out` now posts a comment or fails the gate. ## Notes - Pure CI, no AI. `ent-e2e-gate.yml` only. - Not a required check, so this doesn't block anything — it just stops the false-red noise. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>