Artificial systems for software engineering studies

Abstract

A description is given of techniques for the automatic generation of large artificial software systems that can be used for laboratory studies of testing and integration strategies, reliability models, etc. A prototype generator is described that produces code for such systems by constructing a large number of nearly identical modules. This generator has been used to construct a family of systems that in theory can be made arbitrarily large. Several experiments were conducted to explore the sensitivity of the Jelinski-Moranda model to violations of the assumption that all defects have equal probability of being discovered.<>