Alright, I think the PR is now in a merge ready state
Happy to get some feedback :)

Thanks for Your attempt to fix it. Very likelly we will use another approach. The main problem is in misleading AbstractIO::select() result and wrong exception, not the fact it was interrupted.

Hmm - so it is supposed to be interrupted?

We have an AMQP worker on k8s - when the sigterm comes, it's supposed to finish the last message

when the sigterm comes, while it's waiting for the pending acks for follow up queues, AMQP will throw an error and the message will be reprocessed

What would the approach here then be?
Catching the correct (when implemented) exception and then looping?

Thanks for the constructive feedback!
I changed the PR to instead throw the new Exception AMQPIOWaitInterruptedException

Hope that helps :)

Best,
Marvin

Just noticed what you mean - the error_handler does already catch the interrupt, but does ignore it
So the signal catcher is not needed

Thus I'll close this PR
At least learned a lot about the amqplib ^^

If you want I can create a PR for throwing the correct error within the error_handler