Comments on modified condition/decision coverage for software testing [of flight control software]

This paper considers the modified condition/decision coverage method of testing software. There is an explanation of the method and why it appears reasonable; the probability model offered as a justification is closely examined; and several experiments are described that check the underlying assumptions of the method and its probability model. The method checks the implementation of Boolean expressions by checking the correctness of certain entries in the truth table. These entries are the ones for which a change in a single variable changes the evaluation of the entire expression. The probability model offered as a justification, however, is shown to be equivalent to a binomial distribution of errors, which says all entries are equally likely to be incorrect, Hence, there is not a match between the method and the model. For this reason, a number of experiments are proposed to test the assumptions of the method and the model.