A Domain Strategy for Computer Program Testing

This paper presents a testing strategy desiged to detect errors in the control flow of a computer program, and the conditions under which this strategy is reliable are given and characterized. The control flow statements in a computer progam partition the input space into a set of mutually exclusive domains, each of which corresponds to a particular program path and consists of input data points which cause that path to be executed. The testing strategy generates test points to examine the boundaries of a domain to detect whether a domain error has occurred, as either one or more of these boundaries will have shifted or else the corresponding predicate relational operator has changed. If test points can be chosen within e of each boundary, under the appropriate assumptions, the strategy is shown to be reliable in detecting domain errons of magnitude greater than ∈. Moreover, the number of test points required to test each domain grows only linearly with both the dimensionality of the input space and the number of predicates along the path being tested.