Automated Test Data Generation with SAT

We present a novel technique for automatically generating a suite of test inputs to an object­oriented procedure. The suite is guaranteed to kill all mutant procedures produced from a given catalog of mutation operators, so long as those mutants could be detected by some test within user­provided bounds. Our test input generator constructs a mutation­parameterized version of the procedure, whose behavior may differ from the original by a single mutation. It encodes the original and mutation­parameterized procedures in a first­order rela­ tional formula, a solution to which includes both a mutation and a test input that detects it. Our tool iteratively invokes a constraint­solver to find such tests and adds them to the test suite. After each iteration, it amends the formula to ensure the next mutant found was not killed by a previous input. This process is repeated until the formula becomes unsatisfiable, at which point all mutants have been detected. We evaluate an implementation of this technique on a series of small benchmarks.