An Analysis Of The Coupling Effect In Weak Mutation Testing

In the field of mutation testing,the coupling effect hypothesis suggests that a test data set that detects all simple faults in a program will detect a high percentage of the complex faults.The coupling effect hypothesis is one of basic assumption of mutation testing.Fault-based Boolean-specification testing is an important weak mutation testing techniques,since the running paths of the program are usually dependent on boolean expressions in predicate statements.However,the question whether the coupling effect in the Boolean-specification testing exists or not is still unknown.This paper takes 20 Boolean expressions as experimental objects and uses 10 common fault types as mutation operators.Experiments are carried out to verify the existence of the coupling effect hypothesis in the Boolean specification test.Firstly,1-order mutants and sets of test case for 1-order mutants are generated by tools.On the basis of all 1-order mutants,2-order mutants and the sets of test case for 2-order mutants also be obtained.By analyzing the test sets of 1-order mutants and test sets of 2-order mutants,the detection difficulty of first-order mutants and second-order mutants are compared.For coupling effect analysis,the adequate test sets of 1-order mutants are reduced to detect the 2-order mutants.The percent of the 1-order adequate test sets killing 2-order mutants are calculated.The work of the paper is mainly in the following three aspects:1)After analysis of 1-order mutants and 2-order mutants,it is found that 1-order mutants are more difficult to be detected than 2-order mutants in Boolean-specification testing.2)Using the common reduction algorithm to reduce the 1-order test sets and calculating the percent of 1-order reduction test sets detecting 2-order mutants.Results show that: for any fault type,the coupling effect can not be guaranteed to exist with absolute certainty.Especially,when high-order mutants involve fault type ENF and ORF,the percentage of the coupling effect existing is obviously low.3)Analyzing the experimental results to summarize which types of fault type require higher-order mutation in future mutation testing.Then putting forward methodological guidance for Boolean high-order mutation testing.