Classification of Randomly Generated Test Cases

Random test case generation produces relatively diverse test sequences, but the validity of the test verdict is always uncertain. Because tests are generated without taking the specification and documentation into account, many tests are invalid. To understand the prevalent types of successful and invalid tests, we present a classification of 56 issues that were derived from 208 failed, randomly generated test cases. While the existing workflow successfully eliminated more than half of the tests as irrelevant, half of the remaining failed tests are false positives. We show that the new @NonNull annotation of Java 8 has the potential to eliminate most of the false positives, highlighting the importance of machine-readable documentation.