core: services: wifi: WifiManager: Start without a wlan interface

[patrickelectric]

wpa_supplicant fails when there is no wlan adapter.

Fix #4210

[github-actions]

Automated PR Review

0. Summary

• Verdict: MINOR SUGGESTIONS ✏️

Makes the wifi service tolerate startup with no wlan interface: the wpa_supplicant WifiManager no longer raises when the UDP fallback socket also fails, instead flagging itself as unavailable and short-circuiting every public method that requires a live socket. The NetworkManager handler mirrors the same state="unavailable" sentinel, and the frontend renders "No wifi adapter" plus backs the status polling off to 30 s while that state is set. Adds a small pytest module for the new guards.

1. Correctness & Implementation Bugs

• 1.1 [minor] core/services/wifi/wifi_handlers/wpa_supplicant/WifiManager.py:573 — after the change, start() returns normally on UDP-connect failure, but still schedules auto_reconnect(60) and start_hotspot_watchdog() on the loop. Both tasks eventually call methods that end up in WPASupplicant.send_command, which triggers assert self.sock because wpa.run(...) never succeeded; the resulting SockCommError gets re-raised as RuntimeError in enable_saved_networks and propagates out of the task, killing the watchdog silently. Consider skipping (or short-circuiting) the two loop.create_task(...) calls when self.wpa_path is None, so the service degrades cleanly instead of running for 60 s and then dying quietly.
• 1.2 [minor] core/services/wifi/wifi_handlers/wpa_supplicant/WifiManager.py:97 — self.wpa_path = str(path) uses str() on either a socket path or a (host, port) tuple. For the UDP fallback that yields the literal string "('127.0.0.1', 6664)", which is fine as a truthy marker but confusing given the attribute is typed Optional[str] and elsewhere is treated as a filesystem path. Consider using a dedicated self._connected: bool (or self.wpa_target: Union[str, Tuple[str, int], None]) so the semantic and the name line up.
• 1.3 [nit] core/services/wifi/wifi_handlers/wpa_supplicant/WifiManager.py:98 — the added comment reads "run only returns when the socket is connected, before that there is no adapter to talk to". WPASupplicant.run() only guarantees the wpa_supplicant control socket is connected, not that an actual wlan adapter is present (the UDP fallback path succeeds even when wpa_supplicant is talking to nothing on the far side). Reword or drop.

6. Code Quality & Style

• 6.1 [nit] core/services/wifi/test_wifi_manager.py:46 — _no_networks(_self: Any) -> list[Any] uses PEP 585 builtin-generics syntax while the rest of the wifi service consistently uses typing.List (List[ScannedWifiNetwork], List[SavedWifiNetwork], etc.). Switch to List[Any] (or list unparameterized) for consistency; also worth typing the return as List[ScannedWifiNetwork] to match what get_wifi_available() actually returns.
• 6.2 [nit] core/services/wifi/wifi_handlers/wpa_supplicant/WifiManager.py:481 — set_hotspot_credentials now persists the new ssid/password to SettingsV1 even when there is no adapter, then returns before touching the hotspot. That is likely intentional (settings stick around for when an adapter appears), but is a behavior change worth an inline note or a mention in the PR body.
• 6.3 [nit] core/frontend/src/components/wifi/WifiUpdater.vue:43 — the two setDelay(...) calls hard-code 30000/5000 ms. Lifting them to const UNAVAILABLE_POLL_MS = 30_000 / const NORMAL_POLL_MS = 5_000 keeps the intent readable next to the equally magic constants in data().

7. Tests

• 7.1 [minor] core/services/wifi/test_wifi_manager.py:13 — the sys.modules dance (_saved = ...; sys.modules.pop(...)) drops exceptions, settings, typedefs from the cache but, unlike test_hotspot.py, does not stub out commonwealth, pyroute2, or fastapi before importing WifiManager. If pytest collects test_hotspot.py first in the same worker (pytest-xdist can and does batch by module), the previously-stubbed commonwealth MagicMock has already been restored, so the import path likely works — but it is genuinely order-dependent. Consider mirroring the _STUBBED block from test_hotspot.py:15 to make this hermetic.
• 7.2 [minor] core/services/wifi/test_wifi_manager.py:26 — test_reports_available_after_connecting only asserts manager.wpa_path is not None at the end. Given the PR's core invariant is that wpa_path is set immediately after a successful wpa.run(...) and cleared only in the UDP-failure branch, it would be worth also asserting the value round-trips (assert manager.wpa_path == str(("127.0.0.1", 6664))) and that a subsequent await manager.status() no longer returns state="unavailable" — the current test would still pass even if a future refactor set wpa_path to a placeholder like "connected".

8. Documentation

• 8.1 [nit] The PR body ("wpa_supplicant fails when there is no wlan adapter. Fix bug: WIFI_STATUS_FETCH_FAIL spam with no wlan #4210") does not mention the frontend or the NetworkManager handler changes, though both are user-visible. Expanding the description would help reviewers and future git log readers.

Generated by PR Review Bot. This is advisory, a human reviewer must still approve.

[patrickelectric]

sorry, it should be ready now @bluerobotics/blueos-team

[patrickelectric]

ping @bluerobotics/blueos-team

[patrickelectric]

ping @bluerobotics/blueos-team

[joaoantoniocardoso]

I'm unable to use/test wifi on this branch (from the UI).

[patrickelectric]

@joaoantoniocardoso check now

[patrickelectric]

I found a problem, let me investigate further

[patrickelectric]

@joaoantoniocardoso check now