Unsupported IOT.IPCAMERA devices raise a bare KeyError from get_device_class_from_sys_info during discovery (fix written in #1571, closed by stale bot) #1753

Summary

A Kasa camera/doorbell that answers legacy discovery on UDP/9999 causes
get_device_class_from_sys_info to raise a bare KeyError, which escapes
Discover.datagram_received entirely and surfaces as an unhandled asyncio callback
exception in the consuming application.

This is the library-side cause of home-assistant/core#138866 (open since 2025-02-19,
labelled waiting-for-upstream).

This is not a request for camera support — #467 already tracks that. This is only a
request that an unsupported device type fail the way the library's own callers already
expect it to fail.

Reproduction

python-kasa 0.10.2, Python 3.14.

Any device reporting "type": "IOT.IPCAMERA" in its legacy get_sysinfo response. Here
it is a Kasa KD110(US) video doorbell, hw 4.0, fw 2.3.26 Build 20240508. Three of
them on the network produce three tracebacks per discovery run, indefinitely.

  File "kasa/discover.py", line 377, in datagram_received
    device = device_func(info, config)
  File "kasa/discover.py", line 742, in _get_device_instance_legacy
    device_class = cast(type[IotDevice], Discover._get_device_class(info))
  File "kasa/discover.py", line 722, in _get_device_class
    return get_device_class_from_sys_info(info)
  File "kasa/device_factory.py", line 145, in get_device_class_from_sys_info
    return TYPE_TO_CLASS[IotDevice._get_device_type_from_sys_info(sysinfo)]
KeyError: <DeviceType.Camera: 'camera'>

Cause

kasa/device_factory.py:135-145 — DeviceType.Camera was commented out of
TYPE_TO_CLASS in #1480 ("Disable iot camera creation until more complete", merged
2025-01-25), but the lookup remained a bare subscript:

    TYPE_TO_CLASS = {
        ...
        # Disabled until properly implemented
        # DeviceType.Camera: IotCamera,
    }
    return TYPE_TO_CLASS[IotDevice._get_device_type_from_sys_info(sysinfo)]

_get_device_type_from_sys_info still maps IOT.IPCAMERA to DeviceType.Camera, so the
subscript raises KeyError.

datagram_received guards only UnsupportedDeviceError (discover.py:378) and
KasaException (:385). KeyError is neither, so it propagates out of the protocol
callback. Because the raise is at :377, the code at :391-394 that records the device
and invokes on_discovered never runs — so downstream consumers cannot even see, filter
or ignore the device.

The fix already exists

#1571 ("Raise UnsupportedDeviceError for unsupported legacy device types", @rytilahti,
2025-08-31) does exactly this, and adds the fixture
incomplete_iotcamera = _make_unsupported("IOT.IPCAMERA", "XOR").

It was stale-labelled 2025-11-30 and auto-closed 2025-12-08 with no human review.
kasa/device_factory.py is byte-identical between 0.10.2 and master today.

Its author flagged one concern in the description — "The test might not be correct, as
this was also raising even before adding the new raise, needs to be verified." — which
looks like the only thing standing between it and merge.

Impact

Downstream (Home Assistant) this produces a 19-line traceback per responding camera per
15-minute discovery cycle, indefinitely, for devices the user has deliberately chosen not
to integrate. On my instance that was ~288 tracebacks/day and about 94% of the retained
log file — it evicts unrelated evidence, which is how I found it.