Preserve native K1 error causes and distinguish Wi-Fi actions

2026-09-07 14:11:34 +03:00
parent 8f17a7e282
commit 4f715e41d2
6 changed files with 142 additions and 18 deletions
@@ -11,7 +11,7 @@ import sys


 ROOT = Path(__file__).resolve().parents[1]
-VERSION = "0.8.5"
+VERSION = "0.8.6"
 sys.path.insert(0, str(ROOT.parents[1] / "scripts/packaging"))
 from debian import package

@@ -0,0 +1,44 @@
+# Onboard K1 causal diagnostics R8
+
+The owner screenshot after R7 corresponds to a new network.provision operation,
+not connection.verify. The operation failed at connect-failed in approximately
+68 ms with BleakError, side_effect_status=none and network_not_applied. It did
+not cross the Wi-Fi dispatch boundary. The service remained active without a
+restart. This is distinct from the Linux route-probe TypeError corrected in R7.
+The Bluetooth failure origin remains unproven; a missing BlueZ device object
+is only a hypothesis and is not grounds for an automatic rescan or write retry.
+
+The existing journal logger only recorded PluginExecutionError and the facade
+wrapper stack. The facade retains the native exception as __cause__, but the
+logger did not follow it. The new bounded causal logger records exception class
+and basename/function/line for up to six causes and ten frames per cause. It
+respects suppressed context and terminates cycles. It never renders exception
+messages, source lines, frame locals or payloads. Both journalled and outer
+HTTP failure paths use it. Error responses and device admission are unchanged.
+
+A regression now passes through the actual NodeBridge and plugin facade with
+only the service endpoint replaced. It verifies that the native BleakError
+location survives the real PluginExecutionError wrapper, exactly one connect
+is followed by state observation, and generated secret material is absent from
+logs and the response. A second test covers cyclic and suppressed chains.
+The earlier test substituted NodeBridge.invoke and therefore never exercised
+the wrapper that obscured the installed failure.
+
+The existing wireless modal also distinguishes its two actions more clearly.
+Check K1 state appears before the network fields; the provisioning action is
+now labelled Apply Wi-Fi and check. No new controls, permissions, retries,
+state machine, profiles or host-network changes are introduced. Final Connect
+continues to require current device/session proof.
+
+Validation: 20 NodeBridge/package lifecycle tests, 23 focused frontend and
+architecture checks, 794 full Core frontend tests, Core TypeScript/production
+build and scoped Ruff passed. Node 0.8.6 and optional K1 0.1.5 are reserved for
+this source. Package, installation and owner UI results will be recorded after
+they are observed. This increment improves diagnosis and action clarity; it
+does not claim to fix the physical Bluetooth failure.
+
+The owner screenshot and exact operation result are preserved with UTC and
+monotonic collection timestamps and SHA-256 hashes under private/acceptance/
+k1-node085-20260907-bluetooth-1355. Browser cache clearing for that attempt has
+not been confirmed. No hardware commands were issued by the agent. The next
+physical check remains through the owner UI after clearing browser cache.
@@ -133,10 +133,14 @@ export function DeviceEnrollmentWindow({transport,onChange,onClose,renderWindow}
    {bluetoothFailure&&<div data-enrollment-reveal="result" role="status"><SettingsCard align="center" title={notice}/></div>}
    </SettingsCard>
    {selected&&<div data-enrollment-reveal="wifi"><SettingsCard eyebrow="Шаг 02" title="Wi-Fi"
-     description="Укажите сеть для K1. Проверка передаст настройки сканеру и проверит связь с БК." actions={!confirmed&&
+     description="Проверьте связь с K1 в уже настроенной сети или укажите новую сеть Wi-Fi." actions={!confirmed&&
      <Button disabled={pending} aria-busy={busy==='networks'} icon={busy==='networks'?<ActivityIndicator size="compact"/>:undefined}
       onClick={()=>void run('networks')}>{busy==='networks'?'Ищем сети':'Найти сети'}</Button>}>
     {!confirmed&&<>
+     {enrollmentAllowed(state,'verify')&&
+      <Button disabled={pending} aria-busy={busy==='verify'}
+       icon={busy==='verify'?<ActivityIndicator size="compact"/>:undefined}
+       onClick={()=>void run('verify')}>{busy==='verify'?'Проверяем состояние':'Проверить состояние K1'}</Button>}
     <div data-enrollment-reveal="networks">
     {!!networks.length&&<Select label="Сеть Wi-Fi" value={network}
      options={[{value:'manual',label:'Ввести название сети'},...networks.map((value,index)=>({value:String(index),label:value.ssid,description:`Сигнал ${value.signal}% · ${value.security||'Без защиты'}`}))]}
@@ -146,11 +150,7 @@ export function DeviceEnrollmentWindow({transport,onChange,onClose,renderWindow}
     <TextField label="Пароль Wi-Fi" type="password" value={password} onChange={event=>setPassword(event.target.value)} disabled={pending} autoComplete="new-password"/>
      <Button disabled={pending||!enrollmentAllowed(state,'connect')||!bridgeFormValid(ssid,password)} aria-busy={busy==='connect'}
       icon={busy==='connect'?<ActivityIndicator size="compact"/>:undefined}
-       onClick={()=>void run('connect')}>{busy==='connect'?'Проверяем подключение':'Проверить подключение'}</Button>
-     {enrollmentAllowed(state,'verify')&&
-      <Button disabled={pending} aria-busy={busy==='verify'}
-       icon={busy==='verify'?<ActivityIndicator size="compact"/>:undefined}
-       onClick={()=>void run('verify')}>{busy==='verify'?'Проверяем состояние':'Проверить состояние K1'}</Button>}
+       onClick={()=>void run('connect')}>{busy==='connect'?'Применяем Wi-Fi и проверяем':'Применить Wi-Fi и проверить'}</Button>
     </>}
     {notice&&!bluetoothFailure&&<div data-enrollment-reveal="result" role="status"><SettingsCard align="center" title={notice}/></div>}
    </SettingsCard></div>}
@@ -22,7 +22,7 @@ from credential_install import PROFILE_ID, validate  # noqa: E402
 from debian import package  # noqa: E402
 from runtime_payload import files as runtime_files  # noqa: E402

-VERSION = "0.1.4"
+VERSION = "0.1.5"
 RESOURCES = (
    "plugins/xgrids-k1/profile_loader.py",
    "plugins/xgrids-k1/plugin.manifest.json",
@@ -37,6 +37,29 @@ ATTESTATION = {
 }


+def failure_locations(error: BaseException) -> str:
+    """Bounded causal stack, excluding messages, source lines and frame locals.
+
+    The facade deliberately wraps native transport errors. Logging only the
+    outer stack loses the actual OS failure boundary on journalled responses.
+    """
+    chain: list[str] = []
+    seen: set[int] = set()
+    current: BaseException | None = error
+    while current is not None and id(current) not in seen and len(chain) < 6:
+        seen.add(id(current))
+        frames = list(traceback.walk_tb(current.__traceback__))[-10:]
+        locations = ";".join(
+            f"{Path(frame.f_code.co_filename).name}:{frame.f_code.co_name}:{line}"
+            for frame, line in frames
+        )
+        chain.append(f"{type(current).__name__}[{locations}]")
+        current = current.__cause__ or (
+            None if current.__suppress_context__ else current.__context__
+        )
+    return " <- ".join(chain)
+
+
 class NodeEnrollmentRejected(ValueError):
    """A host admission failure proven to precede any device invocation."""

@@ -152,14 +175,11 @@ class NodeBridge:
            # results, so the HTTP exception logger never sees them. Record
            # only source locations here; exception text may contain a frame.
            logging.getLogger(__name__).warning(
-                "K1 journalled invocation failed: action=%s operation=%s exception=%s locations=%s",
+                "K1 journalled invocation failed: action=%s operation=%s exception=%s chain=%s",
                action,
                identifier,
                type(error).__name__,
-                ";".join(
-                    f"{Path(frame.filename).name}:{frame.name}:{frame.lineno}"
-                    for frame in traceback.extract_tb(error.__traceback__)[-10:]
-                ),
+                failure_locations(error),
            )
            # Read the exact result after a failure; never repeat an invocation
            # or interpret an exception as evidence that a command was unsent.
@@ -346,15 +366,12 @@ def create_app(repository_root: Path):
            # text, source lines, local variables, payloads or credentials.
            action = body.get("action") if isinstance(body, dict) else None
            logging.getLogger(__name__).warning(
-                "K1 operation failed: action=%s exception=%s locations=%s",
+                "K1 operation failed: action=%s exception=%s chain=%s",
                action
                if isinstance(action, str) and action in {"scan", "networks", "connect", "verify"}
                else "invalid",
                type(error).__name__,
-                ";".join(
-                    f"{Path(frame.filename).name}:{frame.name}:{frame.lineno}"
-                    for frame in traceback.extract_tb(error.__traceback__)[-8:]
-                ),
+                failure_locations(error),
            )
            return JSONResponse(
                {
@@ -97,6 +97,69 @@ def test_journalled_failure_logs_source_without_secret_and_does_not_repeat(caplo
    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