ci: trust same-head OpenCode review OID

[seonghobae]

Summary

• Treat an OpenCode review attached to the exact PR head commit OID as current-head approval.
• Remove the committedDate/submittedAt ordering check that caused false negatives when Git commit metadata had a later timestamp.
• Keep old-head reviews blocked by OID mismatch and keep conflict PRs on the conflict repair guide path.

Evidence

• Live PRs such as UI(접근성): 모바일 내비게이션 버튼에 focus 스타일 추가 #451 and build(deps-dev): bump typescript-eslint from 8.60.1 to 8.62.0 #446 had review.commit.oid == headRefOid, but the scheduler blocked them as lacking current-head approval because committedDate was later than submittedAt.
• Dry-run after this change reports 11 update-branch candidates: UI(접근성): 모바일 내비게이션 버튼에 focus 스타일 추가 #451, 🎨 Palette: 컴팩트 네비게이션 키보드 접근성 개선 #448, build(deps-dev): bump typescript-eslint from 8.60.1 to 8.62.0 #446, build(deps-dev): bump vite from 8.0.16 to 8.1.0 #444, build(deps-dev): bump @vitejs/plugin-react from 6.0.2 to 6.0.3 #443, build(deps-dev): bump @tauri-apps/cli from 2.11.2 to 2.11.3 #442, build(deps): bump @base-ui/react from 1.5.0 to 1.6.0 #441, build(deps-dev): bump @types/node from 25.9.3 to 26.0.0 #440, build(deps-dev): bump eslint-plugin-jsdoc from 63.0.5 to 63.0.7 #439, 🧹 Split oversized test function to improve readability and granularity #400, and 🧹 [code health improvement] Extract nested functions from scorecard_sarif_upload_normalization_violations #387.

Validation

• python3 scripts/ci/pr_review_merge_scheduler.py --self-test
• python3 -m py_compile scripts/ci/pr_review_merge_scheduler.py
• python3 scripts/ci/pr_review_merge_scheduler.py --repo ContextualWisdomLab/bandscope --base-branch develop --project-flow git-flow --dry-run --no-trigger-reviews --enable-auto-merge --update-branches --max-prs 100
• git diff --check

[Copilot]

Pull request overview

Adjusts the CI PR review/merge scheduler so an OpenCode “APPROVED” review attached to the exact current PR head commit OID is treated as current-head approval, eliminating false negatives caused by relying on Git timestamps.

Changes:

• Simplifies current-head review validation to an OID equality check (review.commit.oid == headRefOid).
• Removes the committedDate/submittedAt ordering-based staleness logic and the related auto-merge disable branch.
• Updates self_test() expectations to reflect the new approval semantics.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[opencode-agent]

OpenCode Review Overview

• Head SHA: 99c9b98bfef9f8c25ed0faa10032f86f5f077b24
• Workflow run: 28189055131
• Workflow attempt: 1
• Gate result: APPROVE (approval step)

Pull request overview

Reviewed workflow files and dependencies; no issues detected.

Findings

No blocking findings from OpenCode's independent review.

Verification

• Review source: independent OpenCode review of the current checkout, focused changed hunks, and current-head GitHub Check evidence.
• Structural exploration: completed before approval; if structural exploration, changed-file inspection, or evidence completeness is missing, OpenCode must not approve.
• Result: APPROVE
• Reason: No source-backed blockers found and structural exploration completed.

Gate evidence

• Head SHA: 99c9b98bfef9f8c25ed0faa10032f86f5f077b24
• Workflow run: 28189055131
• Workflow attempt: 1

[opencode-agent]

OpenCode reviewed the current-head evidence but found unresolved human review threads before approval.

• Problem: OpenCode reached an APPROVE control result, but the approval step found unresolved, non-outdated human review thread evidence on the current pull request.
• Root cause: Human review feedback can arrive after bounded model evidence is prepared, so the approval step must re-query GitHub immediately before publishing an approval.
• Fix: Address or resolve the listed human review thread(s), then re-run OpenCode on the current head.
• Regression test: Keep the approval gate querying reviewThreads(first: 100) after model output and before create_pull_review APPROVE.

Review thread evidence

Latest unresolved human review thread evidence

scripts/ci/pr_review_merge_scheduler.py line 329

• Latest human comment: @copilot-pull-request-reviewer at 2026-06-25T17:16:54Z

• Comment URL: #458 (comment)

• Comment excerpt: This PR removes the timestamp-based staleness helpers/messages (e.g. stale_current_head_review_reason, review_submitted_datetime(review), and the submitted_at > head_time / "does not postdate" text). The supply-chain policy test services/analysis-engine/tests/test_supply_chain_policy.py currently asserts those exact strings exist in this file (around lines 5015–5023), so CI will fail unless that test is updated to match the new OID-only contract. Also, after removing the timestamp comparison logic, the GraphQL commits(last: 1) { ... committedDate ... } fields appear unused by the sch

• Result: REQUEST_CHANGES

• Reason: unresolved human review thread(s) were present before approval.

• Head SHA: 99c9b98bfef9f8c25ed0faa10032f86f5f077b24

• Workflow run: 28187945340

• Workflow attempt: 1

[opencode-agent]

Pull request overview

Reviewed workflow files and dependencies; no issues detected.

Findings

No blocking findings from OpenCode's independent review.

Verification

• Review source: independent OpenCode review of the current checkout, focused changed hunks, and current-head GitHub Check evidence.
• Structural exploration: completed before approval; if structural exploration, changed-file inspection, or evidence completeness is missing, OpenCode must not approve.
• Result: APPROVE
• Reason: No source-backed blockers found and structural exploration completed.

Gate evidence

• Head SHA: 99c9b98bfef9f8c25ed0faa10032f86f5f077b24
• Workflow run: 28189055131
• Workflow attempt: 1