Using information about functions in selecting test cases

We consider the problem of generating a set of test cases from a black box specification. We focus on stress testing, i.e. picking test cases that seem most likely to reveal program bugs. Our approach assumes that so-called interesting points, i.e. points in a function's domain where properties change, e.g. maxima, are likely to reveal any problems and examine how we can determine the interesting points for a function defined by a complex expression if we know the interesting points for the functions named in that expression.