NODEDC_MISSION_CORE/tests/test_node_k1_bridge.py

import asyncio
import copy
import json
import secrets
from datetime import UTC, datetime, timedelta
from pathlib import Path

import pytest
from missioncore_plugin_sdk.v0alpha2.session import DeviceSessionSnapshot

from k1link.device_plugins.xgrids_k1.linux_host import nm_fields, route_fields
from k1link.device_plugins.xgrids_k1.node_bridge import NodeBridge, plugin_operation_id
from k1link.device_plugins.xgrids_k1.node_sensor import NodeK1Sensor, project_sensor
from k1link.viewer.node_rerun import NodeRerunHub


def state():
    return {
        "snapshot_runtime_id": "runtime-one",
        "snapshot_revision": 1,
        "devices": [
            {"device_id": "AA:BB:CC:DD:EE:FF", "name": "K1-test", "likely_k1": True},
            {"device_id": "unrelated", "name": "Other", "likely_k1": False},
        ],
        "ble_discovery_generation": 1,
        "desired_connection_mode_revision": 0,
        "active_connection_mode": "bridge",
        "connection_lifecycle": {"connection_ready": True, "ready_to_start": True},
        "selected_device_id": "AA:BB:CC:DD:EE:FF",
        "source_mode": "idle",
        "device_session": {
            "device_id": "synthetic-k1",
            "device_session_id": "session-test",
            "opened_at": datetime.now(UTC).isoformat(),
        },
        "application_control_session": {
            "session_generation": 1,
            "state_revision": 1,
            "state": "connection-ready",
        },
    }


class Facade:
    def __init__(self):
        self.current = state()
        self.actions = []

    async def invoke(self, request):
        self.actions.append((request.action_id, copy.deepcopy(request.parameters)))
        if request.action_id == "application-control.workspace.enter":
            self.current["application_control_session"]["state"] = "workspace-ready"
        elif request.action_id == "acquisition.prepare":
            self.current["acquisition"] = {
                "acquisition_id": "acquisition-test",
                "state": "prepared",
                "state_revision": 1,
            }
            self.current["application_control_session"]["state"] = "project-ready"
        elif request.action_id == "acquisition.start":
            self.current["application_control_session"]["state"] = "initializing"
            self.current["acquisition"]["state"] = "running"
            self.current["source_mode"] = "live"
        return copy.deepcopy(self.current)


def bridge():
    result = NodeBridge(Path.cwd(), service=object())
    result.facade = Facade()
    return result


@pytest.mark.parametrize("available", [False, True])
def test_startup_credential_check_does_not_authorize_control(monkeypatch, available):
    from k1link.device_plugins.xgrids_k1 import node_bridge
    from k1link.device_plugins.xgrids_k1.protocol.application_authority import (
        ApplicationAuthorityLoadError,
    )

    calls = []

    class Loader:
        def load(self):
            calls.append("load")
            if not available:
                raise ApplicationAuthorityLoadError("Unavailable")
            return object()

    monkeypatch.setattr(node_bridge, "_validate_installed_compatibility_profile", lambda _: None)
    monkeypatch.setattr(node_bridge, "LinuxApplicationAuthorityLoader", Loader)
    monkeypatch.setattr(node_bridge, "XgridsK1CompatibilityService", lambda *a, **kw: object())
    device = NodeBridge(Path.cwd())
    device.facade = Facade()
    device.facade.current["connection_lifecycle"] = {"connection_ready": False}

    async def read():
        for _ in range(2):
            snapshot = await device.state()
            assert snapshot["application_authority_available"] is available
            assert snapshot["connected"] is False

    asyncio.run(read())
    assert calls == ["load"]


def test_journalled_failure_logs_source_without_secret_and_does_not_repeat(caplog):
    from bleak.exc import BleakError

    secret = secrets.token_hex(20)
    device = bridge()
    calls = []
    identifier = "op_" + "c" * 32
    native = plugin_operation_id(identifier)

    async def invoke(action, payload, _identifier):
        calls.append(action)
        if action == "network.provision":
            raise BleakError(secret)
        return {"operations": [{"operation_id": native, "status": "failed"}]}

    device.invoke = invoke
    result = asyncio.run(device.invoke_journalled(
        "network.provision", {"operation_id": native, "password": secret}, identifier,
    ))
    assert result["operations"][0]["status"] == "failed"
    assert calls == ["network.provision", "state.read"]
    assert "exception=BleakError" in caplog.text
    assert "test_node_k1_bridge.py:invoke:" in caplog.text
    assert identifier in caplog.text
    assert secret not in caplog.text


def test_real_facade_preserves_native_failure_location_in_journalled_log(caplog):
    from bleak.exc import BleakError

    secret = secrets.token_hex(20)
    calls = []

    class Service:
        current = state()

        def bind_runtime_event_loop(self, _loop):
            pass

        def require_snapshot_runtime_id(self, expected):
            assert expected == self.current["snapshot_runtime_id"]

        def state(self):
            calls.append("state")
            return copy.deepcopy(self.current)

        async def connect(self, request):
            calls.append("connect")
            self.current["operations"] = [{
                "operation_id": request.operation_id, "status": "failed",
                "error": {"code": "BleakError"},
            }]
            raise BleakError(secret)

    async def run():
        device = NodeBridge(Path.cwd(), service=Service())
        command = {
            "operation_id": "op_" + "c" * 32, "runtime_id": "runtime-one",
            "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
            "action": "connect", "parameters": {
                "device_id": "AA:BB:CC:DD:EE:FF", "discovery_generation": 1,
                "mode_revision": 0, "ssid": "test-network", "password": secret,
            },
        }
        result = await device.deliver(command)
        assert result["command_result"]["status"] == "failed"
        assert calls == ["state", "connect", "state"]
        assert "password" not in command["parameters"]
        assert secret not in json.dumps(result)

    asyncio.run(run())
    assert "PluginExecutionError[" in caplog.text
    assert "BleakError[" in caplog.text
    assert "test_node_k1_bridge.py:connect:" in caplog.text
    assert secret not in caplog.text
    assert "test-network" not in caplog.text


def test_failure_locations_respects_suppression_and_bounds_cyclic_chains():
    from k1link.device_plugins.xgrids_k1.node_bridge import failure_locations

    first = RuntimeError(secrets.token_hex(20))
    second = ValueError(secrets.token_hex(20))
    first.__context__ = second
    second.__cause__ = first
    assert failure_locations(first) == "RuntimeError[] <- ValueError[]"
    first.__suppress_context__ = True
    assert failure_locations(first) == "RuntimeError[]"


def test_node_scan_reaches_service_through_real_facade_with_runtime_fence():
    """Exercise the actual admission boundary, replacing only the BLE service."""
    from k1link.device_plugins.xgrids_k1.facade import SnapshotRuntimeConflict

    class Service:
        def __init__(self):
            self.scans = []
            self.fences = []
            self.current = state()

        def bind_runtime_event_loop(self, _loop):
            pass

        def state(self):
            return copy.deepcopy(self.current)

        def require_snapshot_runtime_id(self, expected):
            if expected != self.current["snapshot_runtime_id"]:
                raise SnapshotRuntimeConflict()
            self.fences.append(expected)

        async def scan_ble(self, request):
            self.scans.append(request.operation_id)
            return self.state()

    async def run():
        service = Service()
        device = NodeBridge(Path.cwd(), service=service)
        command = {
            "operation_id": "op_" + "b" * 32,
            "runtime_id": "runtime-one",
            "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
            "action": "scan",
            "parameters": {},
        }
        output = await device.deliver(command)
        assert service.fences == ["runtime-one"]
        assert service.scans == [plugin_operation_id(command["operation_id"])]
        assert len(output["candidates"]) == 1
        command["runtime_id"] = "retired-runtime"
        rejected = await device.deliver(command)
        assert rejected["command_result"]["status"] == "rejected"
        assert service.scans == [plugin_operation_id(command["operation_id"])]

    asyncio.run(run())


@pytest.mark.parametrize("radio_failure", [False, True])
def test_node_scan_real_service_preserves_operation_identity_and_replay(
    tmp_path, monkeypatch, radio_failure,
):
    from bleak.backends.device import BLEDevice
    from bleak.backends.scanner import AdvertisementData

    from k1link.device_plugins.xgrids_k1 import facade
    from k1link.device_plugins.xgrids_k1.ble import scanner

    calls = []

    class Radio:
        def __init__(self, detection_callback):
            self.callback = detection_callback

        async def __aenter__(self):
            calls.append("scan")
            if radio_failure:
                raise RuntimeError("synthetic radio unavailable")
            self.callback(
                BLEDevice("AA:BB:CC:DD:EE:FF", "XGR-TEST", None),
                AdvertisementData(
                    local_name="XGR-TEST", manufacturer_data={}, service_data={},
                    service_uuids=[], tx_power=None, rssi=-50, platform_data=(),
                ),
            )
            return self

        async def __aexit__(self, *_args):
            pass

    # Keep the actual Node adapter, facade, service, journal and BLE arbiter.
    # Replace only the OS radio endpoint and shorten its observation window.
    monkeypatch.setattr(scanner, "BleakScanner", Radio)
    monkeypatch.setattr(scanner, "BLE_SCAN_INITIAL_WINDOW_SECONDS", 0.01)
    service = facade.XgridsK1CompatibilityService(tmp_path)

    async def run():
        device = NodeBridge(tmp_path, service=service)
        current = await device.state()
        command = {
            "operation_id": "op_" + "b" * 32,
            "runtime_id": current["runtime_id"],
            "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
            "action": "scan", "parameters": {},
        }
        result = await device.deliver(command)
        if not radio_failure:
            assert scanner.capture_discovered_device("AA:BB:CC:DD:EE:FF") is not None
        repeated = await device.deliver(command)
        assert len(calls) == 1
        assert result["discovery_generation"] == 1
        if radio_failure:
            assert result["candidates"] == []
        else:
            assert result["candidates"][0]["id"] == "AA:BB:CC:DD:EE:FF"
        assert result["command_result"] == repeated["command_result"] == {
            "operation_id": command["operation_id"],
            "action": "scan",
            "status": "failed" if radio_failure else "succeeded",
            "error_code": "RuntimeError" if radio_failure else None,
        }

    try:
        asyncio.run(run())
    finally:
        service.close()


def test_node_attempt_projection_keeps_host_correlation_without_mutating_plugin_state():
    current = state()
    identifier = "op_" + "a" * 32
    current["connection_attempt"] = {
        "schema_version": "missioncore.xgrids-k1-connection-attempt/v1",
        "attempt_id": plugin_operation_id(identifier),
        "recovery_operation_id": plugin_operation_id("op_" + "b" * 32),
        "status": "failed", "public_error_code": "wifi-ssid-not-found",
    }
    result = NodeBridge.project(current)
    assert result["connection_attempt"]["attempt_id"] == identifier
    assert result["connection_attempt"]["recovery_operation_id"] == "op_" + "b" * 32
    assert current["connection_attempt"]["attempt_id"] == plugin_operation_id(identifier)


def test_node_bridge_forces_reviewed_bridge_and_clears_input_secret():
    async def run():
        device = bridge()
        command = {
            "operation_id": "op_" + "a" * 32,
            "runtime_id": "runtime-one",
            "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
            "action": "connect",
            "parameters": {
                "device_id": "AA:BB:CC:DD:EE:FF",
                "discovery_generation": 1,
                "mode_revision": 0,
                "ssid": "test-network",
                "password": secrets.token_urlsafe(24),
                "allow_host_wifi_switch": True,
            },
        }
        secret = command["parameters"]["password"]
        output = await device.execute(command)
        payload = next(v for a, v in device.facade.actions if a == "network.provision")
        assert payload["connection_mode"] == "bridge"
        assert payload["allow_host_wifi_switch"] is False
        assert payload["password"] == secret
        assert "password" not in command["parameters"]
        assert secret not in json.dumps(output)
        assert [v["id"] for v in output["candidates"]] == ["AA:BB:CC:DD:EE:FF"]

    asyncio.run(run())


def test_stale_discovery_prevents_provisioning():
    async def run():
        device = bridge()
        with pytest.raises(ValueError):
            await device.execute(
                {
                    "operation_id": "op_" + "a" * 32,
                    "runtime_id": "runtime-one",
                    "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
                    "action": "connect",
                    "parameters": {
                        "device_id": "AA:BB:CC:DD:EE:FF",
                        "discovery_generation": 0,
                        "mode_revision": 0,
                    },
                }
            )
        assert all(action != "network.provision" for action, _ in device.facade.actions)

    asyncio.run(run())


def test_sensor_projection_binds_native_sdk_to_board():
    value = project_sensor(state(), "node-test")
    snapshot = DeviceSessionSnapshot.model_validate(value["snapshot"])
    assert snapshot.context.execution.node_id == "node-test"
    assert snapshot.context.device.device_id == value["id"]
    assert value["kind"] == "k1"


def test_applied_wifi_does_not_grant_control_or_start_authority():
    current = state()
    current["connection_lifecycle"] = {
        "connection_ready": False, "ready_to_start": False,
        "allowed_actions": ["verify-control-read-only"],
    }
    current["connection_attempt"] = {
        "phase": "network_applied", "public_error_code": "application_authority_unavailable",
    }
    item = project_sensor(current, "node-test")
    assert not item["online"] and not item["verified"]
    assert item["control"]["network_applied"]
    assert item["control"]["reason_code"] == "application_authority_unavailable"
    assert not item["control"]["can_start"]
    assert item["control"]["can_verify"]
    current["connection_lifecycle"].update(connection_ready=True, ready_to_start=True)
    item = project_sensor(current, "node-test")
    assert item["online"] and item["verified"] and item["control"]["can_start"]
    assert item["control"]["reason_code"] is None


def test_one_start_intent_preserves_canonical_enter_prepare_start_sequence():
    async def run():
        device = bridge()
        sensor = NodeK1Sensor(device, None)
        item = project_sensor(state(), "node-test")
        command = {
            "operation_id": "op_" + "a" * 32,
            "action_id": "start",
            "session": {"device_id": item["id"], "session_id": "session-test"},
            "deadline_at": (datetime.now(UTC) + timedelta(seconds=60)).isoformat(),
            "parameters": {
                "operator_confirmed": True,
                "control_generation": 1,
                "acquisition_id": None,
            },
        }
        result = await sensor.execute(command, "node-test")
        actions = [a for a, _ in device.facade.actions if a != "state.read"]
        assert actions == [
            "application-control.workspace.enter",
            "acquisition.prepare",
            "acquisition.start",
        ]
        from k1link.web.device_lifecycle import OperationJournal

        journal = OperationJournal()
        for action, payload in device.facade.actions:
            if action in {"acquisition.prepare", "acquisition.start"}:
                row, created = journal.begin(
                    action, operation_id=payload["operation_id"],
                    idempotency_key=payload["idempotency_key"],
                )
                assert created
                assert journal.begin(action, operation_id=row.operation_id)[1] is False
        assert result["snapshot"]["acquisition"] == "streaming"
        command["parameters"]["control_generation"] = 2
        with pytest.raises(ValueError):
            await sensor.execute(command, "node-test")
        assert len([a for a, _ in device.facade.actions if a == "acquisition.start"]) == 1

    asyncio.run(run())


def test_networkmanager_ssids_are_not_split_at_escaped_colons():
    assert nm_fields(r"field\:network\\name:88:WPA2") == ["field:network\\name", "88", "WPA2"]


def test_wired_board_bridge_does_not_require_host_wifi_association(tmp_path, monkeypatch):
    from k1link.device_plugins.xgrids_k1 import linux_host

    (tmp_path / "enp1s0").mkdir()

    def forbidden(*_args, **_kwargs):
        raise AssertionError("Ethernet must not request Wi-Fi association")

    monkeypatch.setattr(linux_host, "_run", forbidden)
    value = linux_host.LinuxWifiAssociationProbe(tmp_path).observe(interface_name="enp1s0")
    assert value["association_state"] == "not-wifi"
    assert value["continuity_proven"] is True
    assert value["reason_code"] is None


@pytest.mark.parametrize("route_changes", [False, True])
def test_wired_board_real_service_correlates_route_and_control_endpoint(
    tmp_path, monkeypatch, route_changes,
):
    from dataclasses import replace

    from k1link.device_plugins.xgrids_k1 import facade, linux_host

    sys_net = tmp_path / "sys-net"
    (sys_net / "enp1s0").mkdir(parents=True)
    probe = linux_host.LinuxWifiAssociationProbe(sys_net)
    service = facade.XgridsK1CompatibilityService(
        tmp_path / "runtime", host_wifi_association_probe=probe,
    )
    path = facade.HostPathProbeResult(
        available=True, fingerprint="synthetic-route", interface="enp1s0",
        source_ipv4="192.168.1.2", route_class="direct", reason_code=None,
    )
    monkeypatch.setattr(facade, "_inspect_host_path", lambda _target: path)
    calls = []

    def tcp(_target):
        calls.append("tcp")
        if route_changes:
            monkeypatch.setattr(facade, "_inspect_host_path", lambda _target: replace(
                path, fingerprint="new-route", source_ipv4="192.168.1.3",
            ))
        return True

    monkeypatch.setattr(facade, "_control_endpoint_reachable", tcp)

    def forbidden(*_args, **_kwargs):
        raise AssertionError("Wired control bootstrap must not request host Wi-Fi")

    monkeypatch.setattr(linux_host, "_run", forbidden)
    try:
        observed = service._sample_host_path("192.168.1.7")
        assert observed.available is True
        assert observed.interface == "enp1s0"
        assert observed.reason_code is None
        assert observed.fingerprint != path.fingerprint
        assert service._sample_host_path("192.168.1.7").fingerprint == observed.fingerprint
        endpoint = service._probe_control_endpoint("192.168.1.7")
        assert calls == ["tcp"]
        assert endpoint.reachable is (not route_changes)
        assert endpoint.reason_code == (
            "host-path-changed-during-tcp-probe" if route_changes else None
        )
    finally:
        service.close()


def test_linux_kernel_route_is_matched_route_not_resolved_host(monkeypatch):
    from k1link.device_plugins.xgrids_k1 import linux_host
    from k1link.device_plugins.xgrids_k1.facade import _classify_host_route

    calls = []

    def run(args):
        calls.append(args)
        return '[{"dst":"default","dev":"wlp2s0","gateway":"192.168.1.1"}]'

    monkeypatch.setattr(linux_host, "_run", run)
    result = route_fields("192.168.2.7")
    assert calls[0] == ["ip", "-j", "route", "get", "192.168.2.7", "fibmatch"]
    assert _classify_host_route(result["interface"], result["destination"])[0] == "default"
    assert _classify_host_route("wlp2s0", "192.168.1.0/24")[0] == "direct"
    assert _classify_host_route("tailscale0", "192.168.1.7")[0] == "tunnel"


def test_native_node_rrd_opens_no_grpc_listener(monkeypatch):
    from k1link.viewer import rerun_bridge

    def forbidden(*_args, **_kwargs):
        raise AssertionError("Node must not expose a gRPC listener")

    monkeypatch.setattr(rerun_bridge, "_select_available_grpc_port", forbidden)
    hub = NodeRerunHub()
    publisher = hub.create()
    publisher.begin_session()
    subscriber = hub.subscribe()
    try:
        data = subscriber.read()
        assert data[:4] == b"RRF2"
    finally:
        subscriber.close()
        subscriber.thread.join(6)
        publisher.close()


@pytest.mark.parametrize("replacement", [False, True])
def test_preview_offer_cannot_cross_acquisition_boundary(replacement):
    async def run():
        device = bridge()
        device.facade.current["acquisition"] = {"acquisition_id": "acq-one", "state": "running"}
        device.facade.current["source_mode"] = "live"

        class Peers:
            opened = 0
            closed = []

            async def offer(self, _):
                self.opened += 1
                if replacement:
                    device.facade.current["acquisition"]["acquisition_id"] = "acq-two"
                return {"peer_id": "synthetic-peer"}

            async def close(self, identifier):
                self.closed.append(identifier)

        peers = Peers()
        sensor = NodeK1Sensor(device, peers)
        item = project_sensor(device.facade.current, "node-test")
        command = {
            "operation_id": "op_" + "a" * 32,
            "action_id": "offer",
            "session": {"device_id": item["id"], "session_id": "session-test"},
            "parameters": {"acquisition_id": "acq-one" if replacement else "acq-old"},
        }
        with pytest.raises(ValueError):
            await sensor.execute(command, "node-test")
        assert peers.opened == int(replacement)
        assert peers.closed == (["synthetic-peer"] if replacement else [])
        assert all(action == "state.read" for action, _ in device.facade.actions)

    asyncio.run(run())