[aw-failures] [aw-fix] P1: PR Code Quality Reviewer — safe-outputs review 422 "can not request changes on your own PR" on *-mine [Content truncated due to length]

[github-actions]

Fix the PR-review safe-output path to downgrade REQUEST_CHANGES→COMMENT (and drop unresolvable line comments) when the PR is authored by the workflow's own identity — it hard-fails every review on maintainer/bot *-miner PRs.

Problem

The safe_outputs job fails outright when the reviewer targets a PR whose author is the same identity the workflow runs as. GitHub pulls.createReview returns 422 "Review Can not request changes on your own pull request", and the body-only retry fails identically. Result: 4/4 safe outputs failed, review never posted.

Affected workflow / runs

• Workflow: PR Code Quality Reviewer (.github/workflows/pr-code-quality-reviewer.lock.yml)
• Failing runs (both on automated miner branches, PR author == workflow identity):
  • §29946005059 — linter-miner/stringbytesroundtrip..., 2026-07-22 18:17Z (PR [linter-miner] feat(linters): add stringbytesroundtrip linter #47375)
  • §29910168345 — eslint-miner/require-fs-io-try-catch..., 2026-07-22 09:59Z
• Every copilot/*-branch run in the same window succeeded → the failure tracks author-identity, not infra.

Evidence

Two 422 signatures in one finalization of §29946005059:

1. create_pull_request_review_comment → 422 "Line could not be resolved and Line could not be resolved" (unresolvable diff lines) → falls back to body-only review.
2. submit_pull_request_review (event REQUEST_CHANGES) → 422 "Review Can not request changes on your own pull request" (non-retryable). The body-only retry keeps the same event and fails identically.

audit-diff (§29946005059 vs successful §29947043700): no firewall anomalies, no tooling/MCP drift, comparable metrics (10 vs 12 API calls). Confirms a pure API-policy failure, not a code/infra regression.

Probable root cause

The safe-outputs PR-review finalizer unconditionally submits event=REQUEST_CHANGES with line-anchored comments without (a) checking whether the PR author equals the reviewing identity, or (b) validating comment lines are resolvable in the diff. GitHub rejects self-authored REQUEST_CHANGES with a non-retryable 422; the body-only fallback reuses the same event, so it also fails.

Proposed remediation

1. Detect self-authored PRs (PR author login == workflow token identity) and downgrade REQUEST_CHANGES/APPROVE → COMMENT before submission.
2. On "Line could not be resolved", drop the unresolvable comments and submit the remaining valid ones — and ensure the body-only fallback also switches to event=COMMENT so it can succeed on self-authored PRs.
3. If nothing remains postable after downgrade, emit a clean noop instead of hard-failing the job.

Success criteria / verification

• Reviewer run on a *-miner/self-authored PR completes with a posted COMMENT review (or clean noop); safe_outputs job green.
• No 422 "can not request changes on your own pull request" in Process Safe Outputs logs.
• Backfill re-run against PR [linter-miner] feat(linters): add stringbytesroundtrip linter #47375 posts a body-only COMMENT review successfully.

---

Parent: #44298 (PR Code Quality Reviewer reliability). Analyzed runs: 29946005059, 29910168345, 29947043700.
Related to #44298

Generated by 🔍 [aw] Failure Investigator (6h) · age00 131.7 AIC · ⌖ 37.4 AIC · ⊞ 3.8K · ◷

• expires on Jul 29, 2026, 11:22 AM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I've identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · gpt54 2.76 AIC · ⌖ 5.39 AIC · ⊞ 7.2K · ◷