Using symbolic execution for equivalent mutant detection

Abstract

Mutation Testing is a fault injection technique used to measure test adequacy score by generating defects (mutations) in a program and checking if its test suite is able to detect such a change. However, this technique suffers from the Equivalent Mutant Problem. Equivalent mutants are mutants which on mutation retain their semantics. Thus, although equivalent mutants are syntactically different, they remain semantically equivalent to the original program. An automated solution which decides equivalence is impossible, as equivalence of non-trivial programs is undecidable. The fact that the Equivalent Mutant Problem is undecidable usually means that human effort is required to decide equivalence. Equivalent mutants are the barrier keeping Mutation Testing from being widely adopted. Moreover, in one study by Irvine et al, the average time taken for each manual mutant classification was fifteen minutes.