zerortt test counts 1-RTT traffic from first connection

[Ralith]

In the criteria used to judge whether a sufficiently large proportion of traffic is 0-RTT, the 1-RTT traffic used purely to get 0-RTT secrets is included by tr.get_1rtt(Direction.FROM_CLIENT) and counted against the ratio of 0-RTT. In the presence of PLPMTUD probes, this can cause spurious failures.

quic-interop-runner/testcases.py

Lines 691 to 700 in b15dfbb

	zeroRTTSize = self._payload_size(tr.get_0rtt())
	oneRTTSize = self._payload_size(tr.get_1rtt(Direction.FROM_CLIENT))
	logging.debug("0-RTT size: %d", zeroRTTSize)
	logging.debug("1-RTT size: %d", oneRTTSize)
	if zeroRTTSize == 0:
	logging.info("Client didn't send any 0-RTT data.")
	return TestResult.FAILED
	if oneRTTSize > 0.5 * self.FILENAMELEN * self.NUM_FILES:
	logging.info("Client sent too much data in 1-RTT packets.")
	return TestResult.FAILED

[marten-seemann]

Want to send a PR to fix this logic?

[marten-seemann]

You don't need to go by CID to associate 1-RTT packets to a connection. You can count packets sent after the (beginning of the) second handshake as belonging to the second connection.