Limits to program testing with random number inputs

The principal feature of this paper is the illustration of a method which may be useful in the estimation of the number of execution sequences (i.e., realizable logical paths) of a given program. This is accomplished through use of event-altered rate models, particularly the Jelinski-Moranda purification process. A sequence of random numbers is used as input data or drivers for the program, and, on the basis of the spacings between the occurrences of "new" execution sequences, the total number which can be driven by such numbers can be estimated. An application to a particular program is described.