Theoretical insights into fault-based testing

Abstract

The author presents a framework that characterizes fault-based testing schemes based on how many prespecified faults are considered and on the contextual information used to deduce the absence of those faults. Established methods of fault-based testing are placed within this framework. Most methods either are limited to finite fault classes or focus on local effects of faults rather than global effects. A novel method of fault-based testing, called symbolic testing, is presented by which infinitely many prespecified faults can be proved to be absent from a program on the basis of the global effect the faults would have if they were present. Circumstances are discussed as to when testing with a finite test set is sufficient to prove that infinitely many prespecified faults are not present in program.<>