This moves the onMessage handler out of useEffect and it fixes
appending the the array. The .unshift() method does not work
well using React.useState. It would prepend, but since it would not
change the array (it adds an item to the array), the update would
not propagate.
This refactor the UI components from class based element into function
based elements. This makes it possible to use hooks that are used now
by most React components. This also updates React and Ant to the latest
versions (+ other dependencies).
This change makes it possible to use the DeviceModeInd mac-command to
switch between Class-A / C for LoRaWAN 1.0.x devices. In such a case the
Supports Class-C would be disabled for LoRaWAN 1.0.x devices, meaning
that on OTAA the device is marked as Class-A. If the device then sends a
DeviceModeInd mac-command, this will cause the to be marked as
Class-C, event.
This was (unintentionally) possible in v3, but this part of the Class-B
beacon locked code would prevent to make the same work in v4.
This reduces the number of external services that must be running to
succesful run 'cargo test'. In this case one the requirements that
ChirpStack has must be present and there is no need to run local
Kafka and RabbitMQ instances.
These integrations can still be tested using the 'test-all-integrations'
feature flag (which is included in the CI workflow).
The dev-nonce validation updates the used dev-nonces in the database.
Therefore, we should first check the MIC. This also removes the
ctx.save_device_keys step, and puts the join-nonce increment in the
device_keys::validate_incr_... function, reducing the SQL queries by one.
The frequency-hopping ping-slots are equal to the beacon frequencys.
With the US915 and AU915 channels using the Class-A downlink
channels as beacon-frequencies, this was wrongly implemented in the
same way, but in case of CN470 the beacon-frequencies are not equal
to the Class-A downlink frequencies.
