Research On Target Path Selection Method Based On Dominance Relation Of Nodes

There are many factors affecting the test consumption of path coverage,including not only the number of paths to be covered,but also the coverage difficulty of paths,as well as the generation efficiency of test data.At present,many literatures have done a lot of research on test data generation methods,but few research results on target path selection have been seen.Based on path coverage test,this paper studies the method of target path selection,and achieves the goal of reducing the cost of path coverage test by choosing a few representative complete paths.This paper studies the method of path selection from two aspects: serial program and parallel program.Firstly,aiming at serial programs,by analyzing the dominance relation between nodes,an automatic generation strategy of target paths based on dominance relation is proposed.Firstly,the information of nodes and control edges of the program to be tested is analyzed statically.Then,the quantitative characteristics and mutual relations of nodes are analyzed,including the calculation of node outgoing and entry values,the generation of necessary nodes,non-dominant nodes and corresponding dominant node sets of the program to be tested,and the correlation between non-dominant nodes and the generation of similar adjacency matrix are analyzed.Then,using the method of permutation and combination,different combinations of non-dominant node sets satisfying the conditions are given,and corresponding complete paths are automatically generated for the combination of non-dominant node sets.Secondly,based on the research of serial program,aiming at message passing parallel program,an automatic target path selection method based on Latin hyper-cube sampling is proposed.First,the parallel program is divided into different processes,each process is regarded as a serial program;then,the correlation and similar adjacency matrix of the non-dominant nodes of the same process under different scheduling sequences are analyzed;finally,the non-dominant nodes are sampled by hyper-cube sampling method,and different combinations of non-dominant node sets satisfying the conditions are selected.The combination of the non-dominant node set and the corresponding scheduling sequence can automatically generate the corresponding complete path.Experiments show that for serial programs,the proposed strategy of automaticgeneration of target paths based on dominance relation is less than two traditional methods,and the reduction rate of path number is as high as 52.4%.For parallel programs,the strategy of selecting target paths based on Latin hyper-cube sampling is proposed.The reduction rate of path number is about 37.5% when the path is attainable.The path selection method proposed in this paper can not only reduce the number of target paths to be covered,but also ensure the accessibility of the selected paths.Finally,the goal of reducing the test consumption of path coverage is achieved.