ThreadBasedCyclicSendTask stops when any error happens on CAN bus

[tysonite]

I would like to have a discussion over implementation of ThreadBasedCyclicSendTask. The issue there is that if any kind of error happens on the CAN bus, it terminate itself:

python-can/can/broadcastmanager.py

Lines 251 to 253 in b0d7460

	except Exception as exc:
	log.exception(exc)
	break

I see few issues there:

1. The user (application code actually) is not notified if error happens (bus.send() fires exception) during sending periodic messages in ThreadBasedCyclicSendTask, thus application can't handle it properly, for example, restart periodic task after some time. Do I miss something and is this possible?

2. It contradicts with a behavior of, for example, SocketCAN implementation of send_periodic as it moves responsibility of sending periodic messages to the kernel level, and it is expected that error handling will be done inside the kernel. As tests show, if error on CAN bus happens during usage of native periodic implementation, the periodic tasks are still there and it covers after some time.

Any thoughts on this?

[christiansandberg]

It could certainly be discussed. I’m not sure how to handle that situation though. Letting the developer decide is probably better than trying to force some recovery mechanism that won’t suit everyone.

[tysonite]

I would think about having a callback function passed as an argument to bus.send_periodic, it may accept an exception raised internally by ThreadBasedCyclicSendTask so application can be implement any custom logic based on that.

[tysonite]

I could close this issue as fix is merged already. Or is there anything else in your minds?