MSFL: A Model for Fault Localization Using Mutation-Spectra Technique

Fault localization (FL) is the most time-consuming and tedious task, while debugging. Several good techniques have been proposed for effective fault localization. These effective techniques justify the Lean methodology, where the waste process usually been avoided. However, most of the techniques are suffering with the problem of limited accuracy. To overcome the weakness of a technique there is a need of refinement and up-gradation of that technique. To achieve this, we can hybridize two different techniques to take advantages of both the techniques. In this paper, we propose to hybridize Mutation based testing with Spectrum based fault localization. This is a fact that both the techniques are rich in their domains. In our work, we are combining best of these techniques. We first create several mutants and drive along with the test cases to produce spectra for each mutant. This process is accountable under Agile Software Testing. These generated spectra for all mutants are supplied to fault localization techniques such as Tarantula, Barinel, Ochiai, and DStar to generate the statement ranking sequence for each mutant. Similarly, we compute the spectra for faulty program and also the statement ranking sequence. Based upon the similarity between the statement ranking sequence of faulty program and mutants, the bug is localized to most similar mutated line. We have experimented with nine open-source programs and achieved 36.48% improvement over existing FL techniques.