Automatic generation of random self-checking test cases 论文

摘要

A technique of automatically generating random software test cases is described. The nature of such test cases ensures that they will execute to completion, and their execution is predicted at the time of generation. Wherever possible the test cases are self-checking. At run-time their execution is compared with the predicted execution. Also described are implementations of the technique that have been used to test various IBM programs—/I language processors, sort/merge programs, and Graphical Data Display Manager alphanumeric and graphics support.