Given we don't checksum the length field, I think the best we could do is to verify if the length falls off the end of the segment like you said. The only annoyance is that making that check at this layer could be tricky given we don't have access to `LogSegment`. But I agree that if we have a way to assert on reading corrupted records, the warning becomes useless at that point.
[ Full content available at: https://github.com/apache/kafka/pull/5629 ] This message was relayed via gitbox.apache.org for [email protected]
