On the right objectives of data flow testing

Abstract—This paper investigates the limits of current data flow testing approaches from a radically novel viewpoint, and shows that the static data flow techniques used so far in data flow testing to identify the test objectives fail to represent the universe of data flow relations entailed by a program. This paper compares the data flow relations computed with static data flow approaches with the ones observed while executing the program. To this end, the paper introduces a dynamic data flow technique that collects the data flow relations observed during testing. The experimental data discussed in the paper suggest that data flow testing based on static techniques misses many data flow test objectives, and indicate that the amount of missing objectives (false negatives) can be more limiting than the amount of infeasible data flow relations identified statically (false positives). This opens a new area of research of (dynamic) data flow testing techniques that can better encompass the test objectives of data flow testing. I