Research On Multiple Fault Localization Based On Density Clustering

Software debugging includes the work of judging the source of the fault,analyzing and repairing the fault etc.The location of the faulty statement is an essential prerequisite for repairing the fault of programs.Traditional fault localization needs to be done by software debuggers manually.Therefore,to reduce the time and labor costs,automatic fault location methods came into being.SBFL(Spectrum Based Fault Localization)is a critical automatic fault localization method.It is widely used and studied because of its lightweight and high fault localization accuracy.Relevant research on fault localization techniques based on the program spectrum has gradually shifted from a single fault to multiple faults.The localization accuracy is affected by factors such as interference between bugs and CC(Coincidental Correct)test cases.Existing studies generally use clustering to reduce the interference between multiple bugs.The clustering-based method assumes that,ideally,all failed test cases in a single cluster are caused by the same single faulty statement,so failed test cases are clustered into several clusters.However,such clustering algorithm usually needs to define the number of clusters of clustering results.Besides,related research has shown that the presence of coincidental correct test cases in a single-fault program will decrease the accuracy of fault localization.Still,the current multi-fault localization research rarely considers the impact of such test cases on fault localization.In response to the above problems,this paper focuses on density clustering to develop multi-fault localization methods,and on this basis,considers the problem of coincidental correct test cases.First of all,this paper proves through empirical research that the existence of multiple errors will have a negative impact on the effect of fault localization based on the program spectrum,and in the multi-fault localization methods based on parallel debugging,the higher quality clusters can achieve higher fault localization accuracy.Based on this empirical study,this paper proposes a multi-fault localization method based on density clustering,locating only one faulty statement in the program in a single iteration.Finally,based on the fault localization of the multi-fault fault program,this topic further identifies and processes CC test cases,thereby improving the efficiency of the multi-fault location.The experimental results of 804 multi-fault versions of 9 real programs show that,compared with other multi-fault localization methods,the method proposed in this paper can achieve better clustering accuracy and fault localization efficiency,and on this basis,identify and handling with coincidental correct test cases,the efficiency of fault localization can be further improved.