Using program data-state scarcity to guide automatic test data generation

摘要: Finding test data to cover structural coverage criteria such as branch is largely a manual and hence expensive activity. A potential low cost alternative generate the required automatically. Search-based generation one approach that has attracted recent interest. This based on definition of an evaluation or function able discriminate between candidate cases with respect achieving given goal. The implemented by appropriate instrumentation program under test. then executed instrumented program. provides in terms "distance" computation achieved achieve Providing reliably tests are close far from covering goal feasible, search process converge solution, i.e., case satisfies For some programs, however, informative difficult define. operations performed these programs returns constant value for very wide range inputs. typical example this problem arises predicates depend Boolean-valued (flag) variable although not limited contain flag variables. Although methods known overcoming problems variables particular cases, more general near been tackled. paper presents new heuristic directing when at differentiate directs toward produce rare scarce states. Scarce inputs likely values. proposed method evaluated empirically number which existing inadequate.