Exact distribution for independent two-sample problems with only two observations was wrong, spotted by Fritz Scholz .
