Constrained Mutation in C programs

Abstract

Software development is always under the pressure of time and budget constraints before release. A good testing strategy should not only be effective and economical but also incremental. Although mutation testing has been empirically found to be effective in detecting faults, it remains unused for reasons of economics. A major obstacle to the use of mutation testing is its high computational cost. In this paper we report results from experiments designed to investigate six different constrained mutation mechanisms. Our data indicate that these alternatives not only reduce the cost of mutation significantly in terms of the number of test cases required and the number of mutants to be examined, but also maintain very good fault detection effectiveness. Effects of incremental mutation testing examining different sets of mutants are also discussed. Furthermore, our experiments are unique in that constrained mutation was performed directly on C programs. This eliminates the possible bias experienced by earlier mutation studies because of the programming language translation between the Fortran, Pascal, and C.