Assessing the quality of specification-based testing
Stephen P. Allen, Martin R. Woodward · 1996
This paper reports an experimental investigation into the quality of sets of test expressions automatically generated from algebraic specifications. Faults were introduced into implementations in C and Fortran equivalent to the chosen specifications, manually in the case of the C versions, and using the Mothra mutation system in the case of the Fortran versions. The algebraic specifications themselves were also subject to mutation analysis. A number of test generation strategies were considered, in order to circumvent the problem of handling prohibitively large numbers of test expressions. Results show that certain test sets detected the majority of faults and appear to perform better than random test sets of comparable sizes.