Program segmentation for controlling test coverage

We present a new control flow based approach to dynamic testing of sequential software. A practicable number of test cases is generated by using the boundary interior path testing strategy (J.B. Goodenough and S.L. Gerhard, 1975) and by dividing the test units into test segments (program fragments composed of one statement or a sequence of statements). The size of the test segments can be adjusted by means of a parameter, i.e. the thoroughness of the test coverage can be adapted to the needs of the tester. The selection of test cases is performed by constructing path classes for each test segment. The coverage criteria constructed by means of our approach (test segment coverage criteria) are fulfilled if at least one path from each path class is covered. A validation of our approach is given by comparing the fault detection capabilities of test segment coverage criteria with the fault detection capabilities of branch testing, multiple condition testing, LCSAJ testing and all-uses testing using n test cases for each item (e.g. branch) to be covered. The comparison demonstrates that, compared with the other testing criteria, greater fault detection probabilities can be achieved if a test segment coverage criterion is used.