Revision history [back]

I’m thinking that the important thing to test is the “do something” code, and not ROS’s publish and subscribe mechanisms (or the little bit of probably-not-going-to-change-very-often client code that calls them). To test the “do something” code, I’d be inclined to write a set of tests where each test constructs a different message (the data structure), passes it to your “do something” code as input, and verifies whatever the “do something” code produces in response. Would that serve your objectives?