Instrumenting Programs With Flag Variables For Test Data Search By Genetic Algorithms
Leonardo Bottaci · 2002
Evolutionary search is potentially a powerful way of searching for software test data to satisfy various structural testing criteria. Specific test cases are evaluated by a fitness function constructed by instrumenting the program under test. The more discriminating the fitness function, the more efficient the search. When a program uses flag variables to store the results of predicate expressions, it is difficult to instrument the program effectively. The problem is examined and a solution is given for a special case. An approach for tackling the general cases is described.