Archive exact open PR handoff representations#8677
Archive exact open PR handoff representations#8677aragora-automation-fable[bot] wants to merge 12 commits into
Conversation
…xact-open-pr-transition
…xact-open-pr-transition
Aragora Code ReviewAdvisory-only review. No issues found. |
OpenAI independent model reviewReviewer: codex/openai - independent focused dogfood review by a non-author lane, grounded on the exact PR head. Verdict: PASS Blocking findings: none. Files reviewed:
Focused dogfood verdict: The change is narrowly scoped to exact-open-PR terminal handling in the outbox reconciler. It requires narrow GitHub evidence, exact branch match, and full desired-head match before archiving an outbox item, and it preserves a keep decision when an open PR exists at a different head or branch. The tests cover the archive/receipt path and the mismatch keep path, which directly exercises the queue-pressure scenario this PR is meant to resolve. Validation observed: live required checks for lint, typecheck, Generate & Validate, sdk-parity, and TypeScript SDK Type Check are green; aragora-merge-quorum is the only failing required check and is failing because model evidence is missing. I reviewed the current PR diff and existing comments/reviews. I did not run local tests in this lane. dogfood: yes This is not merge authorization. |
Require exact-open-PR terminal archival to stay on PR-publication handoffs, honor local work markers and local-evidence base refs, and only run narrow GitHub proof after the bulk open-PR state is available.\n\nCo-authored-by: codex[bot] <codex[bot]@users.noreply.github.com>
Co-authored-by: codex[bot] <codex[bot]@users.noreply.github.com>
OpenAI independent model reviewReviewer: codex/openai independent focused dogfood review by a non-author lane, grounded on the exact PR head. Verdict: PASS Blocking findings: none. Files reviewed:
Focused dogfood verdict: The change makes exact-open-PR outbox archival proof-gated instead of relying on broad open-PR branch presence alone. It requires publication intent, non-conflicting local evidence, no local-work markers, full desired-head matching, branch/base agreement, non-draft exact PR evidence, and owner-liveness safety before synthesizing a terminal receipt. The tests cover archive success, head/base/branch/draft mismatch keeps, local work/conflicting local evidence keeps, owner possible-unpushed-work/live-owner keeps, idempotency-key routing, and unhealthy GitHub fail-closed behavior. Validation observed: live required checks for lint, typecheck, Generate & Validate, sdk-parity, and TypeScript SDK Type Check are green. The only failing required check is aragora-merge-quorum because current-head model evidence is missing. I reviewed the current PR diff and visible comments/reviews. I did not run local tests in this lane. dogfood: yes This is not merge authorization. |
Keep automation outbox handoffs with contradictory local_evidence from being terminal-archived by landed-on-main, missing-branch, or merged-PR proof paths. Add regression coverage for each archive path. Co-authored-by: Codex <codex@openai.com>
Use the selected state root for exact-open-PR owner liveness checks, reuse the shared head-field aliases, and allow the proof-gated narrow REST exact-PR path when the bulk open-PR cache misses a branch. Co-authored-by: codex[bot] <codex[bot]@users.noreply.github.com>
OpenAI independent focused dogfood reviewReviewer: codex/openai independent focused dogfood review by a non-author lane, grounded on exact PR head. Verdict: PASS Blocking findings: none. Files reviewed:
Focused dogfood verdict: The current-head change adds a stricter exact-open-PR terminal archival path for automation outbox handoffs. It gates archival on publication intent, non-conflicting local evidence, no local-work markers, full desired-head equality, matching branch/base GitHub evidence, non-draft exact open PR evidence, and owner-liveness checks before writing a synthetic receipt. The tests exercise archive success, narrow REST and bulk-cache miss cases, local evidence conflicts, draft, head, branch, and base mismatch keeps, unhealthy GitHub fail-closed behavior, and owner state plumbing. This reduces stale outbox pressure without broad-draining or deleting uncertain work. Validation observed: live required checks for lint, typecheck, Generate and Validate, sdk-parity, and TypeScript SDK Type Check are green; aragora-merge-quorum is failing because current-head model evidence is incomplete. I reviewed the exact current diff, comments and reviews, and failed quorum log. I did not run local tests in this lane. dogfood: yes This is not merge authorization. |
|
Closing as a queue-drain churner, not as a judgment that the idea can never return. Current live state at close:
Under the current queue-drain objective, this meets the FREEZE CHURNERS rule: do not keep patching or recollecting on a PR that has churned across 3+ heads and still fails quorum. The reversible drain action is to close this PR and let any future attempt return as a simpler replacement that removes the fragile abstraction or narrows the scope enough to settle cleanly. |
|
Queue-drain freeze: closing this reopened churner again under the hard anti-treadmill rule. Live state before this close:
Per the queue-drain FREEZE CHURNERS rule, this should not receive another patch or evidence cycle in this loop. The reversible drain action is to close it and preserve the branch for a deliberately smaller replacement if this outbox/handoff transition needs to return. |
Summary
Validation
Live dry-run proof