CellularStateMachine can generate spurious attach events #9737
Description
Description
In the following code...
UARTSerial * serial = new UARTSerial(....);
UBLOX_AT * dev = new UBLOX_AT(serial);
CellularContext* iface = dev->create_context(serial, "stream.co.uk", false, false);
iface->set_credentials(..., ..., ...);
iface->connect();
dev->set_timeout(5000);
The intended timeout of 5000 can be overwritten and reset to 60000. This appears to be because the CellularStateMachine is driven by a separate thread and so the "connect" call can return before all events are processed. Specifically, the "state_attaching" can be called twice, setting the timeout both times. This appears to be because it uses the current _network_status as a guard but fails to set it upon successfully attaching. The following version of the code behaves a bit better though I'm not sure if it completely addresses the issue.
void CellularStateMachine::state_attaching()
{
if (_network_status != ATTACHED_TO_NETWORK) {
_cellularDevice.set_timeout(TIMEOUT_CONNECT); //only set the timeout if we are trying to attach
_cb_data.error = _network->set_attach();
}
if (_cb_data.error == NSAPI_ERROR_OK) {
_network_status |= ATTACHED_TO_NETWORK; //ensure we know we're attached.
if (_event_status_cb) {
_cb_data.status_data = CellularNetwork::Attached;
_event_status_cb(_current_event, (intptr_t)&_cb_data);
}
} else {
retry_state_or_fail();
}
}
Issue request type
[ ] Question
[ ] Enhancement
[X] Bug