fix: preserve live tasks during daemon recovery

tests/unit/admin_console/test_automation.py

@@ -300,6 +300,116 @@ def test_restart_never_repeats_started_work(isolated):
     assert isolated.jobs(retry["id"])[0]["session_id"] != job["session_id"]
 
 
+@pytest.mark.parametrize("liveness_source", ["pid", "device_lock"])
+def test_recovery_preserves_live_runner_and_batch(isolated, monkeypatch, liveness_source):
+    import os
+    from types import SimpleNamespace
+    from artemis.runtime import DeviceExecutionLock
+
+    batch = service.create_batch(payload())
+    service.batch_action(batch["id"], "resume")
+    job = isolated.jobs(batch["id"])[0]
+    isolated.save_job({**job, "status": "running"})
+    if liveness_source == "pid":
+        with sqlite3.connect(isolated.path) as conn:
+            conn.execute(
+                "INSERT INTO sessions(session_id,initial_goal,start_time,status,device_info,pid) "
+                "VALUES (?,?,?,?,?,?)",
+                (job["session_id"], "goal", 1, "running", "{}", os.getpid()),
+            )
+    monkeypatch.setattr(
+        DeviceExecutionLock,
+        "get_active_owners",
+        lambda: (
+            {"device": SimpleNamespace(session_id=job["session_id"])}
+            if liveness_source == "device_lock"
+            else {}
+        ),
+    )
+    service.recover_batches()
+    assert isolated.job(job["session_id"])["status"] == "running"
+    assert not isolated.job(job["session_id"]).get("error")
+    assert isolated.get("batch", batch["id"])["status"] == "running"
+
+
+@pytest.mark.asyncio
+async def test_natural_completion_resolves_old_recovery_error_in_job_report_and_history(
+    client, isolated
+):
+    job = isolated.jobs(service.create_batch(payload())["id"])[0]
+    message = "服务中断；未完成任务需手动重跑"
+    isolated.save_job({**job, "status": "interrupted", "error": message})
+    await service.on_terminal({"session_id": job["session_id"]}, "completed")
+    saved = isolated.job(job["session_id"])
+    assert saved["status"] == "completed"
+    assert not saved.get("error")
+    assert saved["recovery_events"][0]["message"] == message
+    assert reports.session_report(job["session_id"])["error"] == ""
+    assert not client.get("/api/automation/history").json()["items"][0].get("error")
+
+
+def test_recovery_repairs_completed_legacy_warning_but_preserves_real_failure(isolated):
+    batch = service.create_batch(payload())
+    first, second = isolated.jobs(batch["id"])
+    isolated.save_job({**first, "status": "completed", "error": "服务中断，需手动重跑"})
+    isolated.save_job({**second, "status": "failed", "error": "任务执行超时（30 秒）"})
+    assert reports.session_report(first["session_id"])["error"] == ""
+    service.recover_batches()
+    assert not isolated.job(first["session_id"]).get("error")
+    assert "服务中断" in isolated.job(first["session_id"])["recovery_events"][0]["message"]
+    assert "超时" in isolated.job(second["session_id"])["error"]
+
+
+@pytest.mark.asyncio
+async def test_surviving_worker_is_finalized_when_it_exits_without_resubmission(
+    isolated, monkeypatch
+):
+    import os
+    from artemis.runtime import DeviceExecutionLock
+
+    batch = service.create_batch(payload())
+    service.batch_action(batch["id"], "resume")
+    job = isolated.jobs(batch["id"])[0]
+    isolated.save_job({**job, "status": "running"})
+    monkeypatch.setattr(DeviceExecutionLock, "get_active_owners", lambda: {})
+    with sqlite3.connect(isolated.path) as conn:
+        conn.execute(
+            "INSERT INTO sessions(session_id,initial_goal,start_time,status,device_info,pid) "
+            "VALUES (?,?,?,?,?,?)",
+            (job["session_id"], "goal", time.time(), "running", "{}", os.getpid()),
+        )
+    service.recover_batches()
+    await service.reconcile_recovered_jobs()
+    assert isolated.job(job["session_id"])["status"] == "running"
+    with sqlite3.connect(isolated.path) as conn:
+        conn.execute(
+            "UPDATE sessions SET pid=NULL,status='completed' WHERE session_id=?",
+            (job["session_id"],),
+        )
+    await service.reconcile_recovered_jobs()
+    saved = isolated.job(job["session_id"])
+    assert saved["status"] == "completed"
+    assert saved["report_ready"]
+    assert not saved.get("recovery_waiting")
+    assert not saved.get("error")
+    assert len(isolated.jobs(batch["id"])) == 2
+    assert isolated.jobs(batch["id"])[1]["status"] == "pending"
+
+
+@pytest.mark.asyncio
+async def test_surviving_worker_that_dies_without_result_is_interrupted(isolated, monkeypatch):
+    from artemis.runtime import DeviceExecutionLock
+
+    job = isolated.jobs(service.create_batch(payload())["id"])[0]
+    isolated.save_job({**job, "status": "running", "recovery_waiting": True})
+    monkeypatch.setattr(DeviceExecutionLock, "get_active_owners", lambda: {})
+    await service.reconcile_recovered_jobs()
+    saved = isolated.job(job["session_id"])
+    assert saved["status"] == "interrupted"
+    assert "服务中断" in saved["error"]
+    assert not saved.get("recovery_waiting")
+
+
 @pytest.mark.asyncio
 async def test_batch_admission_idempotent_and_bound_device(isolated, monkeypatch):
     from apps.admin_console import server