The check has been blocking the staging→main auto-promote PR (#2361) since 2026-04-30T07:17Z with: fatal: origin/main...<head>: no merge base Root cause: the workflow does `git fetch origin <base> --depth=1` which overwrites checkout@v4's full-history clone with a shallow tip — destroying the ancestry the subsequent `git diff origin/main...HEAD` (three-dot, merge-base form) needs. This deadlocks every staging→main promote PR until manually fixed. The auto-promote runs were succeeding at the gate-check phase but the subsequent PR-merge step waited 30 min for the failing check and timed out, skipping the publish + redeploy dispatch tail. Fleet recovery for any production-only fix went through staging fine but never reached main. Fix: drop --depth=1 so the explicit fetch preserves full history. The leading comment is updated to call out this trap so a future maintainer doesn't re-add the flag thinking it's a perf win. No test added: this is a workflow-config one-liner that the existing PR check itself exercises end-to-end (the real signal is PR #2361 going green after this lands). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| scripts | ||
| workflows | ||
| CODEOWNERS | ||
| dependabot.yml | ||