feat: require validation cwd and HEAD proof in reviewer reports (Closes #398)

Rebased onto current prgs/master; merged with #396 validation failure history by keeping both validator rules and workflow handoff fields. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
2026-07-07 13:09:58 -04:00
parent 1a1e679246
commit 1320a55a7e
6 changed files with 441 additions and 0 deletions
@@ -213,6 +213,12 @@ def test_validation_failure_history_verifier_exported():
    assert callable(assess_validation_failure_history_report)


+def test_validation_cwd_proof_verifier_exported():
+    from review_proofs import assess_validation_cwd_proof_report
+
+    assert callable(assess_validation_cwd_proof_report)
+
+
 def test_prior_blocker_skip_verifier_exported():
    from review_proofs import assess_prior_blocker_skip_proof

@@ -0,0 +1,135 @@
+"""Tests for validation cwd/HEAD proof verifier (#398)."""
+import sys
+import unittest
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from final_report_validator import assess_final_report_validator  # noqa: E402
+from reviewer_validation_cwd_proof import assess_validation_cwd_proof_report  # noqa: E402
+
+ROOT = "/Users/jasonwalker/Development/Gitea-Tools"
+WORKTREE = f"{ROOT}/branches/review-feat-issue-398"
+HEAD = "f5953549aad5e822f14f52d3ea3c6d7990109384"
+
+
+def _proof_backed_report() -> str:
+    return "\n".join([
+        f"Candidate head SHA: {HEAD}",
+        f"pwd: {WORKTREE}",
+        f"git rev-parse HEAD: {HEAD}",
+        "git status --short --branch: ## feat/issue-398...prgs/master",
+        f"Validation command: cd {WORKTREE} && venv/bin/python -m pytest tests/ -q",
+        "Result: 1497 passed, 6 skipped",
+    ])
+
+
+class TestValidationCwdProof(unittest.TestCase):
+    def test_no_validation_claim_passes(self):
+        result = assess_validation_cwd_proof_report("Review decision: approve")
+        self.assertTrue(result["proven"])
+
+    def test_missing_cwd_proof_fails(self):
+        result = assess_validation_cwd_proof_report(
+            f"Validation command: pytest tests/\nCandidate head SHA: {HEAD}",
+            validation_session={"validation_ran": True, "expected_head_sha": HEAD},
+        )
+        self.assertFalse(result["proven"])
+        self.assertTrue(result["block"])
+
+    def test_main_checkout_cwd_blocks(self):
+        result = assess_validation_cwd_proof_report(
+            "\n".join([
+                f"Candidate head SHA: {HEAD}",
+                f"pwd: {ROOT}",
+                f"git rev-parse HEAD: {HEAD}",
+                "git status --short --branch: ## master",
+                "Validation command: pytest tests/ -q",
+            ]),
+            validation_session={"validation_ran": True, "expected_head_sha": HEAD},
+            project_root=ROOT,
+        )
+        self.assertFalse(result["proven"])
+        self.assertTrue(result["violations"])
+
+    def test_wrong_head_blocks(self):
+        wrong = "a" * 40
+        result = assess_validation_cwd_proof_report(
+            "\n".join([
+                f"Candidate head SHA: {HEAD}",
+                f"pwd: {WORKTREE}",
+                f"git rev-parse HEAD: {wrong}",
+                "git status --short --branch: clean",
+                f"Validation command: cd {WORKTREE} && pytest -q",
+            ]),
+            validation_session={"validation_ran": True, "expected_head_sha": HEAD},
+            project_root=ROOT,
+        )
+        self.assertFalse(result["proven"])
+        self.assertTrue(result["violations"])
+
+    def test_fully_proof_backed_passes(self):
+        result = assess_validation_cwd_proof_report(
+            _proof_backed_report(),
+            validation_session={"validation_ran": True, "expected_head_sha": HEAD},
+            project_root=ROOT,
+        )
+        self.assertTrue(result["proven"], result["reasons"])
+
+    def test_baseline_without_cwd_fails(self):
+        result = assess_validation_cwd_proof_report(
+            "\n".join([
+                _proof_backed_report(),
+                "Baseline validation command: pytest tests/ -q",
+            ]),
+            validation_session={
+                "validation_ran": True,
+                "expected_head_sha": HEAD,
+                "baseline_validation_ran": True,
+            },
+            project_root=ROOT,
+        )
+        self.assertFalse(result["proven"])
+        self.assertTrue(
+            any("baseline" in r.lower() for r in result["reasons"])
+        )
+
+    def test_baseline_with_full_proof_passes(self):
+        result = assess_validation_cwd_proof_report(
+            "\n".join([
+                _proof_backed_report(),
+                f"Baseline worktree: {ROOT}/branches/baseline-master-pr376",
+                f"Baseline target SHA: {HEAD}",
+                f"Baseline validation command: cd {ROOT}/branches/baseline-master-pr376 && pytest -q",
+            ]),
+            validation_session={
+                "validation_ran": True,
+                "expected_head_sha": HEAD,
+                "baseline_validation_ran": True,
+            },
+            project_root=ROOT,
+        )
+        self.assertTrue(result["proven"], result["reasons"])
+
+    def test_final_report_validator_integration(self):
+        result = assess_final_report_validator(
+            "Validation command: pytest tests/ -q",
+            "review_pr",
+            validation_session={"validation_ran": True},
+        )
+        self.assertTrue(result["blocked"] or result["downgraded"])
+        self.assertTrue(
+            any(
+                f.get("rule_id") == "reviewer.validation_cwd_proof"
+                for f in result.get("findings") or []
+            )
+        )
+
+    def test_exported_from_review_proofs(self):
+        from review_proofs import assess_validation_cwd_proof_report as exported
+
+        self.assertTrue(callable(exported))
+
+
+if __name__ == "__main__":
+    unittest.main()