| Popis: |
Context:\ud \ud In software testing, Failed Error Propagation (FEP) is the situation in which a faulty program state occurs during the execution of the system under test (SUT) but this does not lead to incorrect output. It is known that FEP can adversely affect software testing and this has resulted in associated information theoretic measures.\ud \ud \ud Objective:\ud \ud To devise measures that can be used to assess the testability of the SUT. By testability, we mean how likely it is that a faulty program state, that occurs during testing, will lead to incorrect output. Previous work has considered a single program point rather than an entire program.\ud \ud \ud Method:\ud \ud New, more fine-grained, measures were devised. Experiments were used to evaluate these and the previously defined measures (Squeeziness and Normalised Squeeziness). The experiments assessed how well these measures correlated with an estimate of the probability of FEP occurring during testing. Mutants were used to estimate this probability.\ud \ud \ud Results:\ud \ud A strong rank correlation was found between several of the measures and the probability of FEP. Importantly, this included the Normalised Squeeziness of the whole SUT, which is simpler to compute, or estimate, than most of the other measures considered. Additional experiments found that the measures were relatively insensitive to the choice of mutants and also test suite.\ud \ud \ud Conclusion:\ud \ud There is scope to use information theoretic measures to estimate how prone an SUT is to FEP. As a result, there is potential to use such measures to prioritise testing or estimate how much testing an SUT might require. |
Full Text from ScienceDirect