nanocoap_sock: only re-transmit CON messages

[benpicco]

Contribution description

If a NON confirmable message is sent with a callback function, not receiving a response in time would lead to a retransmission.

This is of course an error, as only CON messages are to be retransmitted.

Testing procedure

Issues/PRs references

[riot-ci]

Murdock results

✔️ PASSED

782910a nanocoap_sock: only re-transmit CON messages

Success	Failures	Total	Runtime
2000	0	2000	06m:56s

Artifacts

• Documentation preview

This only reflects a subset of all builds from https://ci-prod.riot-os.org. Please refer to https://ci.riot-os.org for a complete build for now.

[chrysn]

Tested with BOARD=native make -C tests/nanocoap_cli all term / put_non coap://[2001:db8::1]/ hello before-after, largely LGTM with one note:

This sets the application timeout (which on this API is fixed to whatever nanocoap decides) to the time it'd take for the first CON transmission to be sent. That's probably not too bad if the times are all configured well (with ACK_TIMEOUT larger than a typical RTT), but if RTTs do exceed ACK_TIMEOUT, responses get lost.

For comparison -- not that I'd be particularly convinced it's the best thing to do --, gcoap keeps waiting for as long as it'd wait for CONs if it retransmitted (at least when waiting for separate responses, which is a different thing, but still comparable; I don't know off my head what it does for NONs).

What's coded there is an OK choice, I just want to be sure it's understood that that choice is being made here.

[benpicco]

Thank you for the review!
Maybe it would make sense to have an optional sock_asnyc option for nanoCoAP.
But for separate response there will be a separate timeout as well, but that's a differnet PR 😉