feat(lock): allow exact-owner renewal of expired author issue locks (Closes #760)

An expired author issue lease could not be renewed by the exact session that already owned it. `assess_same_issue_lease_conflict` computed same-owner evidence and then returned on the expired branch before consulting it, and `assess_expired_lock_reclaim` only permits takeover on a dead PID or a missing worktree. Because the recorded PID is the long-lived MCP daemon rather than the authoring task, a lease that expires under a live daemon is the ordinary case for any author task outliving the TTL — and in that case the owner's own lock became permanently unmodifiable through sanctioned tools. Add `issue_lock_renewal`, a pure evidence assessor for that one case, and evaluate its disposition before the expired foreign-takeover return. Renewal requires an exact match of remote, org, repo, issue number, operation type, branch, realpath-normalized worktree, claimant username, and claimant profile; a registered worktree that exists, sits on the locked branch, and is clean; local and remote heads that agree; and, when an owning PR exists, a PR head that agrees too. No competing live lock, competing branch claim, or other owning PR may exist. Any missing or contradictory evidence fails closed. PID liveness is never authorization: it is recorded as evidence and is neither necessary nor sufficient (AC16). Only an expired lease is ever a candidate, so a live foreign lease stays non-recoverable (AC12) and dead-PID takeover keeps its existing #601 conditions (AC11). Renewal is never caller-declarable — the waiver is server-computed and `gitea_lock_issue` gains no parameter (AC14). A sanctioned renewal records prior PID, prior expiry, replacement PID, new expiry, claimant, and its supporting proof under `lease_renewal`, and reuses the #772 compare-and-swap so two sessions observing the same expired lease cannot both win. Absolute wall-clock expiry is preserved. Sliding heartbeat renewal, fencing tokens, and the shared cross-role lifecycle remain #790's scope and are deliberately not implemented here; #790 stays sequenced behind this change. Tests: 40 new cases covering the positive path, every near-match and foreign-owner refusal, the non-candidate cases, the gate-ordering regression, the renewal record, downstream mutation ownership, and AC14/AC17. Two #772 test doubles now forward the new keyword. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_01Ti8deB36iWcjHmE9cuxour
2026-07-22 00:39:51 -04:00
parent 3d0c13fa5a
commit 1a97ced133
5 changed files with 1085 additions and 20 deletions
@@ -0,0 +1,447 @@
+"""Exact-owner renewal of an expired author issue lease (#760).
+
+Covers the renewal disposition that lets the exact recorded owner re-acquire
+its own lock after the wall-clock lease expires — including while the recording
+MCP daemon PID is still alive — plus every rejection condition that must keep
+failing closed, and the pre-existing dead-PID and live-foreign dispositions
+that must remain untouched.
+"""
+
+import inspect
+import os
+import subprocess
+import sys
+import tempfile
+import unittest
+from datetime import datetime, timedelta, timezone
+
+sys.path.insert(0, str(__import__("pathlib").Path(__file__).resolve().parent.parent))
+
+import issue_lock_renewal  # noqa: E402
+import issue_lock_store  # noqa: E402
+
+ISSUE = 5150
+BRANCH = f"fix/issue-{ISSUE}-demo"
+WORKTREE = "/scratch/wt-5150"
+HEAD = "c" * 40
+OTHER_SHA = "d" * 40
+IDENTITY = "example-user"
+PROFILE = "example-author"
+REMOTE = "prgs"
+ORG = "ExampleOrg"
+REPO = "ExampleRepo"
+
+
+def dead_pid() -> int:
+    """A PID that has certainly exited (spawned, then reaped)."""
+    proc = subprocess.Popen([sys.executable, "-c", "pass"])
+    proc.wait()
+    return proc.pid
+
+
+def past_ts(hours: int = 1) -> str:
+    return (
+        (datetime.now(timezone.utc) - timedelta(hours=hours))
+        .isoformat()
+        .replace("+00:00", "Z")
+    )
+
+
+def future_ts(hours: int = 4) -> str:
+    return (
+        (datetime.now(timezone.utc) + timedelta(hours=hours))
+        .isoformat()
+        .replace("+00:00", "Z")
+    )
+
+
+def make_lock(*, expires_at: str | None = None, pid: int | None = None, **overrides):
+    """An expired lock owned by a still-alive daemon PID — the #760 condition."""
+    lock = {
+        "issue_number": ISSUE,
+        "branch_name": BRANCH,
+        "worktree_path": WORKTREE,
+        "remote": REMOTE,
+        "org": ORG,
+        "repo": REPO,
+        # os.getpid() is unambiguously alive: the whole point of #760 is that
+        # daemon liveness is not evidence of an active author task.
+        "session_pid": os.getpid() if pid is None else pid,
+        "lock_generation": 3,
+        "work_lease": {
+            "operation_type": issue_lock_store.AUTHOR_ISSUE_WORK_LEASE,
+            "issue_number": ISSUE,
+            "branch": BRANCH,
+            "worktree_path": WORKTREE,
+            "claimant": {"username": IDENTITY, "profile": PROFILE},
+            "created_at": past_ts(5),
+            "expires_at": expires_at or past_ts(),
+        },
+    }
+    lease_overrides = overrides.pop("work_lease", None)
+    if lease_overrides:
+        lock["work_lease"].update(lease_overrides)
+    lock.update(overrides)
+    return lock
+
+
+def assess(lock=None, **overrides):
+    """Run the assessor with all-passing evidence unless overridden."""
+    kwargs = {
+        "issue_number": ISSUE,
+        "branch_name": BRANCH,
+        "worktree_path": WORKTREE,
+        "remote": REMOTE,
+        "org": ORG,
+        "repo": REPO,
+        "identity": IDENTITY,
+        "profile": PROFILE,
+        "current_branch": BRANCH,
+        "porcelain_status": "",
+        "worktree_exists": True,
+        "head_sha": HEAD,
+        "remote_head_sha": HEAD,
+        "pr_head_sha": None,
+        "pr_number": None,
+        "competing_live_locks": [],
+        "candidate_branches": [BRANCH],
+        "current_pid": 4242,
+    }
+    kwargs.update(overrides)
+    return issue_lock_renewal.assess_exact_owner_lease_renewal(
+        make_lock() if lock is None else lock, **kwargs
+    )
+
+
+class ExactOwnerRenewalGranted(unittest.TestCase):
+    """AC1/AC3-AC7: the positive path."""
+
+    def test_expired_lease_alive_pid_exact_owner_is_renewable(self):
+        result = assess()
+        self.assertEqual(result["outcome"], issue_lock_renewal.RENEWAL_SANCTIONED)
+        self.assertTrue(result["renewal_sanctioned"])
+        self.assertTrue(result["is_candidate"])
+
+    def test_renewal_holds_when_owning_pr_head_matches(self):
+        result = assess(pr_number=999, pr_head_sha=HEAD)
+        self.assertTrue(result["renewal_sanctioned"])
+
+    def test_evidence_records_both_sides_of_the_transition(self):
+        result = assess()
+        evidence = result["evidence"]
+        self.assertEqual(evidence["prior_pid"], os.getpid())
+        self.assertTrue(evidence["prior_pid_alive"])
+        self.assertEqual(evidence["replacement_pid"], 4242)
+        self.assertTrue(evidence["prior_expires_at"])
+
+
+class ExactOwnerRenewalRefused(unittest.TestCase):
+    """AC3-AC8: every near-match must fail closed, one reason at a time."""
+
+    def _refused(self, **overrides):
+        result = assess(**overrides)
+        self.assertEqual(result["outcome"], issue_lock_renewal.REFUSED)
+        self.assertFalse(result["renewal_sanctioned"])
+        self.assertTrue(result["reasons"])
+        return result
+
+    def test_different_branch_refused(self):
+        result = self._refused(branch_name=f"fix/issue-{ISSUE}-other")
+        self.assertTrue(any("branch" in r for r in result["reasons"]))
+
+    def test_different_worktree_refused(self):
+        result = self._refused(worktree_path="/scratch/somewhere-else")
+        self.assertTrue(any("worktree" in r for r in result["reasons"]))
+
+    def test_different_claimant_refused(self):
+        result = self._refused(identity="someone-else")
+        self.assertTrue(any("claimant" in r for r in result["reasons"]))
+
+    def test_different_profile_refused(self):
+        result = self._refused(profile="other-author")
+        self.assertTrue(any("profile" in r for r in result["reasons"]))
+
+    def test_different_remote_org_or_repo_refused(self):
+        self._refused(remote="dadeschools")
+        self._refused(org="OtherOrg")
+        self._refused(repo="OtherRepo")
+
+    def test_dirty_worktree_refused(self):
+        result = self._refused(porcelain_status=" M gitea_mcp_server.py\n")
+        self.assertTrue(any("uncommitted" in r for r in result["reasons"]))
+
+    def test_missing_worktree_refused(self):
+        result = self._refused(worktree_exists=False)
+        self.assertTrue(any("does not exist" in r for r in result["reasons"]))
+
+    def test_worktree_on_wrong_branch_refused(self):
+        self._refused(current_branch="master")
+
+    def test_local_and_remote_head_mismatch_refused(self):
+        result = self._refused(remote_head_sha=OTHER_SHA)
+        self.assertTrue(
+            any("does not equal remote head" in r for r in result["reasons"])
+        )
+
+    def test_unpublished_branch_refused(self):
+        result = self._refused(remote_head_sha=None)
+        self.assertTrue(any("remote branch head" in r for r in result["reasons"]))
+
+    def test_pr_head_mismatch_refused(self):
+        result = self._refused(pr_number=999, pr_head_sha=OTHER_SHA)
+        self.assertTrue(any("does not equal local" in r for r in result["reasons"]))
+
+    def test_unobservable_pr_head_refused(self):
+        self._refused(pr_number=999, pr_head_sha=None)
+
+    def test_competing_live_lock_on_same_issue_refused(self):
+        result = self._refused(
+            competing_live_locks=[
+                {"issue_number": ISSUE, "branch_name": BRANCH, "pid": 777}
+            ]
+        )
+        self.assertTrue(any("live lock" in r for r in result["reasons"]))
+
+    def test_competing_live_lock_holding_the_branch_refused(self):
+        self._refused(
+            competing_live_locks=[
+                {"issue_number": 111, "branch_name": BRANCH, "worktree_path": ""}
+            ]
+        )
+
+    def test_competing_branch_claim_refused(self):
+        result = self._refused(candidate_branches=[BRANCH, f"feat/issue-{ISSUE}-rival"])
+        self.assertTrue(any("issue marker" in r for r in result["reasons"]))
+
+    def test_malformed_durable_lock_refused(self):
+        lock = make_lock()
+        lock["worktree_path"] = ""
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.REFUSED)
+
+    def test_lock_without_recorded_claimant_refused(self):
+        lock = make_lock()
+        lock["work_lease"]["claimant"] = {}
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.REFUSED)
+
+
+class NotARenewalCandidate(unittest.TestCase):
+    """AC12 and scope: situations renewal must decline to judge at all."""
+
+    def test_live_foreign_lease_is_never_a_candidate(self):
+        lock = make_lock(expires_at=future_ts())
+        result = assess(lock, identity="someone-else")
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+        self.assertFalse(result["renewal_sanctioned"])
+
+    def test_unexpired_lease_is_never_a_candidate(self):
+        lock = make_lock(expires_at=future_ts())
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+
+    def test_dead_pid_under_unexpired_lease_stays_with_753(self):
+        """The opposite trigger; #760 must not re-own it."""
+        lock = make_lock(expires_at=future_ts(), pid=dead_pid())
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+
+    def test_absent_lock_is_not_a_candidate(self):
+        result = assess({})
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+
+    def test_different_issue_is_not_a_candidate(self):
+        lock = make_lock()
+        lock["issue_number"] = ISSUE + 1
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+
+    def test_different_operation_type_is_not_a_candidate(self):
+        lock = make_lock()
+        lock["work_lease"]["operation_type"] = "review_pr_work"
+        result = assess(lock)
+        self.assertEqual(result["outcome"], issue_lock_renewal.NO_CANDIDATE)
+
+
+class DaemonPidIsNotTaskLiveness(unittest.TestCase):
+    """AC16: a live recorded PID is never, by itself, authorization."""
+
+    def test_alive_pid_alone_does_not_authorize_renewal(self):
+        # Every ownership fact except the live PID is wrong.
+        result = assess(identity="someone-else", branch_name="fix/issue-1-nope")
+        self.assertEqual(result["outcome"], issue_lock_renewal.REFUSED)
+        self.assertTrue(result["evidence"]["prior_pid_alive"])
+
+    def test_renewal_does_not_require_a_dead_pid(self):
+        result = assess()
+        self.assertTrue(result["evidence"]["prior_pid_alive"])
+        self.assertTrue(result["renewal_sanctioned"])
+
+    def test_dead_pid_does_not_block_an_otherwise_exact_owner(self):
+        lock = make_lock(pid=dead_pid())
+        result = assess(lock)
+        self.assertTrue(result["renewal_sanctioned"])
+
+
+class ConflictGateOrdering(unittest.TestCase):
+    """AC2: the same-owner allowance is reachable on an expired lease.
+
+    These cases need a worktree that genuinely exists on disk. The #601 reclaim
+    affordance already permits takeover when the recorded worktree is missing,
+    so a fictional path would satisfy the gate for the wrong reason and never
+    exercise the ordering defect this issue is about.
+    """
+
+    @classmethod
+    def setUpClass(cls):
+        cls._tmp = tempfile.TemporaryDirectory()
+        cls.worktree = cls._tmp.name
+
+    @classmethod
+    def tearDownClass(cls):
+        cls._tmp.cleanup()
+
+    def present_lock(self, **overrides):
+        return make_lock(worktree_path=self.worktree, **overrides)
+
+    def test_expired_same_owner_is_allowed_when_renewal_is_sanctioned(self):
+        block = issue_lock_store.assess_same_issue_lease_conflict(
+            self.present_lock(),
+            issue_number=ISSUE,
+            branch_name=BRANCH,
+            worktree_path=self.worktree,
+            renewal_sanctioned=True,
+        )
+        self.assertIsNone(block)
+
+    def test_expired_same_owner_still_blocks_without_the_waiver(self):
+        """Regression for the ordering defect: no waiver, no change in behavior.
+
+        Live PID and a present worktree, so the #601 reclaim affordance refuses;
+        before #760 this was the permanent dead end for an exact owner.
+        """
+        lock = self.present_lock()
+        self.assertFalse(
+            issue_lock_store.assess_expired_lock_reclaim(lock)["reclaim_allowed"]
+        )
+        block = issue_lock_store.assess_same_issue_lease_conflict(
+            lock,
+            issue_number=ISSUE,
+            branch_name=BRANCH,
+            worktree_path=self.worktree,
+        )
+        self.assertIsNotNone(block)
+        self.assertIn("Recovery review is required", block)
+
+    def test_waiver_does_not_unlock_a_different_owner(self):
+        """AC11: the waiver is scoped by same_owner, not merely by its own flag."""
+        block = issue_lock_store.assess_same_issue_lease_conflict(
+            self.present_lock(),
+            issue_number=ISSUE,
+            branch_name=f"fix/issue-{ISSUE}-someone-else",
+            worktree_path=self.worktree,
+            renewal_sanctioned=True,
+        )
+        self.assertIsNotNone(block)
+        self.assertIn("Recovery review is required", block)
+
+    def test_live_lease_disposition_is_unchanged(self):
+        """AC12: a live foreign lease still blocks, waiver or not."""
+        block = issue_lock_store.assess_same_issue_lease_conflict(
+            self.present_lock(expires_at=future_ts()),
+            issue_number=ISSUE,
+            branch_name=f"fix/issue-{ISSUE}-someone-else",
+            worktree_path="/scratch/other",
+            renewal_sanctioned=True,
+        )
+        self.assertIsNotNone(block)
+        self.assertIn("already has an active", block)
+
+    def test_dead_pid_reclaim_path_is_unchanged(self):
+        """AC11: expired + dead PID still reclaims through the #601 affordance."""
+        lock = self.present_lock(pid=dead_pid())
+        reclaim = issue_lock_store.assess_expired_lock_reclaim(lock)
+        self.assertTrue(reclaim["reclaim_allowed"])
+        block = issue_lock_store.assess_same_issue_lease_conflict(
+            lock,
+            issue_number=ISSUE,
+            branch_name=BRANCH,
+            worktree_path=self.worktree,
+        )
+        self.assertIsNone(block)
+
+
+class RenewalRecordAndDownstream(unittest.TestCase):
+    """AC9/AC10: durable audit trail, and a renewed lock that actually works."""
+
+    def test_record_captures_prior_and_replacement_state(self):
+        assessment = assess()
+        record = issue_lock_renewal.build_renewal_record(
+            assessment,
+            renewed_at="2026-01-01T00:00:00Z",
+            new_expires_at="2026-01-01T04:00:00Z",
+        )
+        self.assertTrue(record["renewed"])
+        self.assertEqual(record["prior_pid"], os.getpid())
+        self.assertEqual(record["new_expires_at"], "2026-01-01T04:00:00Z")
+        self.assertEqual(record["renewed_at"], "2026-01-01T00:00:00Z")
+        self.assertEqual(record["identity"], IDENTITY)
+        self.assertEqual(record["profile"], PROFILE)
+        self.assertTrue(record["prior_expires_at"])
+        self.assertTrue(record["proof"])
+
+    def test_renewed_lock_satisfies_verify_lock_for_mutation(self):
+        renewed = make_lock(expires_at=future_ts())
+        renewed["session_pid"] = os.getpid()
+        renewed["lease_renewal"] = {"renewed": True}
+        verdict = issue_lock_store.verify_lock_for_mutation(
+            renewed,
+            issue_number=ISSUE,
+            branch_name=BRANCH,
+        )
+        self.assertTrue(verdict["proven"])
+        self.assertFalse(verdict["block"])
+
+    def test_refusal_message_names_the_missing_evidence(self):
+        assessment = assess(porcelain_status=" M gitea_mcp_server.py\n")
+        message = issue_lock_renewal.format_renewal_refusal(assessment)
+        self.assertIn("refused", message)
+        self.assertIn("uncommitted", message)
+
+
+class NoCallerControlledRenewalFlag(unittest.TestCase):
+    """AC14: renewal eligibility is never declarable by a caller."""
+
+    def test_lock_issue_tool_exposes_no_renewal_parameter(self):
+        import gitea_mcp_server
+
+        target = gitea_mcp_server.gitea_lock_issue
+        target = getattr(target, "fn", getattr(target, "__wrapped__", target))
+        params = set(inspect.signature(target).parameters)
+        for forbidden in ("renewal_sanctioned", "renew", "allow_renewal", "is_owner"):
+            self.assertNotIn(forbidden, params)
+
+    def test_store_defaults_to_no_waiver(self):
+        params = inspect.signature(
+            issue_lock_store.assess_same_issue_lease_conflict
+        ).parameters
+        self.assertIs(params["renewal_sanctioned"].default, False)
+        bind_params = inspect.signature(issue_lock_store.bind_session_lock).parameters
+        self.assertIs(bind_params["renewal_sanctioned"].default, False)
+
+
+class NoIssueNumberSpecialCasing(unittest.TestCase):
+    """AC17: no repository issue or PR number is special-cased."""
+
+    def test_module_contains_no_hardcoded_issue_special_cases(self):
+        source = inspect.getsource(issue_lock_renewal)
+        code = "\n".join(
+            line for line in source.splitlines() if not line.strip().startswith("#")
+        )
+        for literal in ("757", "759", "760"):
+            self.assertNotIn(f"== {literal}", code)
+            self.assertNotIn(f"issue_number == {literal}", code)
+
+
+if __name__ == "__main__":
+    unittest.main()
@@ -871,9 +871,13 @@ class TestAc6McpUnpublishedClaimRecovery(_UnpublishedMcpBase):
        save_calls: list[dict] = []
        real_save = mcp_server._save_issue_lock

-        def tracking_save(data, *, expected_generation=None):
+        def tracking_save(data, *, expected_generation=None, renewal_sanctioned=False):
            save_calls.append({"expected_generation": expected_generation, "data": dict(data)})
-            return real_save(data, expected_generation=expected_generation)
+            return real_save(
+                data,
+                expected_generation=expected_generation,
+                renewal_sanctioned=renewal_sanctioned,
+            )

        with patch("mcp_server._save_issue_lock", side_effect=tracking_save):
            result = self.run_lock_issue()
@@ -926,18 +930,33 @@ class TestAc6McpUnpublishedClaimRecovery(_UnpublishedMcpBase):
        real_bind = issue_lock_store.bind_session_lock
        bind_calls: list[int | None] = []

-        def racing_bind(data, lock_dir=None, expected_generation=None):
+        # #760 added the renewal waiver keyword; the double forwards it verbatim
+        # so this race still exercises the real compare-and-swap.
+        def racing_bind(
+            data, lock_dir=None, expected_generation=None, renewal_sanctioned=False
+        ):
            bind_calls.append(expected_generation)
            if expected_generation is None:
-                return real_bind(data, lock_dir=lock_dir, expected_generation=None)
+                return real_bind(
+                    data,
+                    lock_dir=lock_dir,
+                    expected_generation=None,
+                    renewal_sanctioned=renewal_sanctioned,
+                )
            # First concurrent writer wins.
            if len([c for c in bind_calls if c is not None]) == 1:
                return real_bind(
-                    data, lock_dir=lock_dir, expected_generation=expected_generation
+                    data,
+                    lock_dir=lock_dir,
+                    expected_generation=expected_generation,
+                    renewal_sanctioned=renewal_sanctioned,
                )
            # Second concurrent writer still holds the pre-race generation.
            return real_bind(
-                data, lock_dir=lock_dir, expected_generation=expected_generation
+                data,
+                lock_dir=lock_dir,
+                expected_generation=expected_generation,
+                renewal_sanctioned=renewal_sanctioned,
            )

        # First recovery succeeds and advances generation.