I'm trying to transmit and receive small chunks of data in between data processing steps. I can only get the first chunk of data across, after that, calling tx.start() and rx.start() does not seems to work. The flowgraphs I'm starting have head() blocks in them that are responsible for calling tx.stop() and rx.stop(). So the series of call would look something like this:
gr::blocks::head, does not call stop(). It signals WORK_DONE (return -1) which GR's scheduler recognizes and lets the downstream blocks finish their processing. A flow graph in this state is merely idle, not stopped. You need to call stop() yourself (after wait()) in order to reuse the top block.
When the head blocks call tx.stop() and rx.stop() internally, does that make the flowgraphs unrecoverable for the next tx/rx.start()? If this is the case, how would we implement what I'm describing?
Besides the missing stop(), the other problem is the head block itself — it counts the number of items it's passed on, independent of any starts and stops. Call reset() on it to tell it to pass some more items.
When you're wondering about the exact behavior, I recommend consulting the source code of the blocks. You don't have to know everything about how blocks work to get some clues about what they do or don't support, or what happens in a particular condition.