fix(test): align live-draft/head-mismatch test with #1697's contract

[seonghobae]

Description

While driving docs/product-goal-directive.md's reconciliation work on a different branch, a routine full-suite run after merging current main in turned up 2 new failures in tests/test_opencode_live_draft_state_regression.py::test_draft_exemption_fails_closed_when_live_head_moved (both script parametrizations).

Root-caused via an isolated origin/main worktree (confirms this is a pre-existing regression on main, not caused by anything on the branch that surfaced it): #1697 ("fix(opencode): retire stale draft/head dispatches without false failure") deliberately reordered opencode-review.yml's "Request current-head OpenCode review execution" and "Fail closed" steps so live-draft state is checked before live-head match, not after — a currently-draft PR needs no review dispatched no matter which head a stale/racing event or live re-fetch names, so the old ordering's hard exit 1 on any head mismatch was pure red-X noise for an already-benign case.

#1697 added thorough new coverage for this intended behavior in tests/test_opencode_required_verdict_regression.py (test_request_review_step_exempts_a_draft_pr_whose_live_head_has_moved, test_fail_closed_step_exempts_a_draft_pr_whose_live_head_has_moved), but left this sibling file's test_draft_exemption_fails_closed_when_live_head_moved pinned to the old, pre-#1697 contract — directly contradicting the new tests added in the same PR. Since main's own required coverage-evidence check runs the full suite, every PR currently rebasing onto main (including .github#1538, this session's own docs PR) would inherit this failure regardless of its own diff.

Fix

Updated the test (renamed to test_draft_exemption_applies_even_when_live_head_has_moved) to assert the same contract #1697's own new tests establish: draft state exempts dispatch with exit 0 and no error output, even when the live head has also moved. No production code changed — #1697's reordering is the intentional, correct fix; this PR only brings the one stale test file back in sync with it.

Verification

• python3 -m pytest tests/test_opencode_live_draft_state_regression.py -q → 19 passed
• python3 -m pytest tests -q (full suite) → 2563 passed, 1 skipped, 21 subtests, 0 failed
• Reproduced the failure first on an unmodified origin/main git worktree to confirm it predates and is independent of this fix

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist

• Self-reviewed
• Root-caused against a clean origin/main checkout before touching anything
• New and existing unit tests pass locally with my changes
• My changes generate no new warnings

---

🤖 Generated with Claude Code

https://claude.ai/code/session_019YgmaEKnPhrsbSyQBierFH

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

Duplicate of #1705

This is the same fix for the same main-tip regression (test_draft_exemption_fails_closed_when_live_head_moved vs. #1697's intentional check-reorder), independently arrived at in a parallel session at essentially the same timestamp — the diffs are functionally identical (same assertions, even the same renamed test function name).

#1705 already carries this exact delta and has full verification behind it (coverage run -m pytest tests -q: 2577 passed / 1 skipped / 0 failed; coverage report: 100% statements/branches on scripts/ci; interrogate: 100.0% docstrings) plus a PR description documenting the cross-validated root cause (confirmed independently by three other background agents working on #1242, #1245, #1476).

Closing this one as a full successor takeover by #1705, per this session's PR-closure policy (duplicate with an identical, already-verified successor) — not a bare close. No delta from this PR is lost; #1705 contains it. If #1705 doesn't land for any reason, please reopen this one rather than starting a third copy.

---

Generated by Claude Code

[harbourviewcompany-create]

Evidence log — closure audit — 2026-09-02

PR #1703 is closed and unmerged. Its exact head is abb5bbfa8f2a9bc46914ea9a04bd6bb1dc56cfa0.

Historical validation recorded on the PR body: focused tests/test_opencode_live_draft_state_regression.py → 19 passed; full python3 -m pytest tests -q → 2563 passed, 1 skipped, 21 subtests, 0 failed.

The GitHub Actions records for that exact head are historical skipped/cancelled runs across the security and provenance workflows. They are not current protected-main evidence. The PR is closed, so it is not a merge candidate and should not be reopened merely to manufacture current-head gate evidence; the fix it carried was subsequently incorporated into the mainline evolution.

Gate decision: CLOSED / NOT A MERGE CANDIDATE.