Search Based Test Suite Minimization For Fault Detection And Localization

Fault detection and localization are two important and independent processes in regression testing.For improving efficiency of regression testing,test suite optimization is usually used in the two processes.Since the goals of fault detection and localization are different,which means fault detection aims at ensure whether the programs under test has faults while fault localization aims at finding where the faults are,test suite optimized according to the two goals are usually different from each other.Early researches found that test suite optimized by fault detection is not suitable for fault localization and vice versa.To further improve the efficiency of regression testing,we firstly focus on test suite minimization which is one of the most popular kinds of test suite optimization methods and analysis the two optimized objectives for fault detection and localization.We then verify the gap between fault detection driven test suite minimization and fault localization driven test suite minimization,and do further analysis for combining fault detection and localization,to realize the aim of improving efficiency in regression testing.To verify the gap between fault detection and fault localization,this paper first analysis the differences between fault detection goals and localization goals,and then implements a detection-driven multi-objective test suite optimization named MoTSM-FD and firstly propose a localization-driven multi-objective test suite optimization namely MoTSM-FL for sufficient analysis.In the empirical study,the two method are compared to verify the gap and their objectives are analyzed to explain the relationship between fault detection and localization.Also,since the MoTSM-FL is firstly proposed in this paper,its effectiveness is simply verified in follows.By verifying the effects on fault detection and fault localization between MoTSM-FD and MoTSM-FL,confirming the difference of effects and goals between fault detection and localization,this paper proposes a co-driven test suite minimization for fault detection and localization,which are also used in an empirical study.This paper shows an empirical study under 70000 test suites and the results show that this method can keep good fault detection ability and high fault localization effectiveness,and further get a high suite reduction ratio.In this paper,there is also a further analysis on the correlation between fault detection and localization to verify the efficiency of Co-MoTSM.To further analyzing the effectiveness of Co-MoTSM and verifying the relationship between fault detection goal and fault localization goal,this paper also does a further discussion of the correlation between the two optimization goals.The theoretical and empirical study of this paper show that the proposed method can give consideration for both fault detection and localization,it can also keep a high test suite reduction ratio and a high efficiency on optimization phase.This is a good try on the combination between two processes in regression software testing.