Abstract:

Approaches for fault localization based on test suites are now collectively called TBFL (Testing Based Fault Localization). However, current algorithms do not take advantage of the prior knowledge about test cases and program so that they waste these valuable "resources". Stochastic TBFL is a new kind of TBFL approach whose spirit is to combine the prior knowledge with actual testing activities under stochastic theory, so as to locate program faults. Stochastic TBFL may be regarded as a general pattern of this approach, from which people can develop various algorithms. We separate the prior knowledge about program and test suites from actual testing activities, and according to testing results we then add the prior knowledge. This new method is called adjusting the probability of error statements based on testing results. Comparison to stochastic TBFL and some instances shows that our improved approach is feasible. Moreover, the paper proposes three standards for comparing different TBFL approaches. And from the investigation of the three standards, results of our proposed approach are also good.
      

Key words: software test;fault localization;prior knowledge;random method