A case study of coverage-checked random data structure testing

We study coverage-checked random unit testing (CRUT), the practice of repeatedly testing units on sequences of random function calls until given code coverage goals are achieved. Previous research has shown that this practice can be a useful complement to traditional testing methods. However, questions remained as to the breadth of its applicability. In this paper, we report on a case study in which we applied CRUT to the testing of two mature public-domain data structures packages. We show that CRUT helped in identifying faults, in debugging, in extracting and specifying actual behaviour, and in achieving greater assurance of the correctness of the debugged software