An Empirical Study Of Input Domain Defect Location Technology

Software testing technology is one of the important way to guarantee software quality.Combinatorial testing technology can detect all kinds of interactive faults by using a small number of test cases.It is necessary to locate the fault after using combinatorial testing technology to find the fault.Because all the factor interactions in the minimal failure-causing schema(MFS)can cause system failure,so long as we find the MFS of many suspicious fault-causing schema,we can know which factor interaction will cause failure.In this thesis,the fault location method is divided into two kinds: adaptive method and non-adaptive method.Based on the TCAS benchmark dataset in the field of combinatorial testing research,twenty TCAS Boolean expessions and their variants are selected as the experimental subject.The main research work includes:(1)The localization effect of non-adaptive method by using high-strength combination coverage table(CA)and partial coverage table(PCA)as two different error location tables(ELA).(2)The localization effect of two different methods including non-adaptive method and adaptive method.The main research results of this thesis are listed as follows:(1)The necessary condition for non-adaptive method is that a safety value must be exised in the system under test.Taking the twenty TCAS Boolean expessions and their variants as the experimental subject,which verify whether the safety value exist universally;(2)Taking the twenty TCAS Boolean expessions and their variants as the experimental subject,the effects of four PCA generation algorithms are compared and analyzed,which are Greedy,Density RO,Density FO and IPO algorithm;(3)Taking the recall,precision,f-measure and the number of test cases as the experimental evaluation indexes,the twenty TCAS Boolean expessions and their variants as the experimental subject,which verify and analyze the localization effect of non-adaptive fault location method by using CA and PCA as different error location tables;(4)Taking the twenty TCAS Boolean expessions and their variants as the experimental subject,under the condition of large number of actual MFSs and high strength,how to select the dimension of CA and PCA;(5)Taking the recall,precision,f-measure and the number of test cases as the experimental evaluation indexes,the twenty TCAS Boolean expessions and their variants as the experimental subject,which measure the performance of both adaptive and non-adaptive methods.