Research Of Regression Testing Techniques Incorporating Defect Correlation

This paper mainly considers research of regression testing techniques incorporating defect correlation in order to optimize test cases suits and to simple the process of regression testing. This can save required human, material and financial resources in regression testing, reduce the time and price of regression testing and improve efficiency of regression testing. Techniques of test case suits in regression testing are consisted of test case selection, test case optimization and test case reduction. This paper will discuss techniques of test case optimization and test case reduction with defect correlation that incorporated in regression testing.For defect correlation, although it is acknowledged that defects are not independent existence and there is a certain correlation between two defects no matter in academic and engineering, the research of defect correlation is still at an early stage both at home and aboard. It is great less to apply defect correlation in engineering industry. Currently, there is only a software development management system ---- BMS XP published by Wicresoft Co., Ltd in Shanghai, which uses defect correlation. The techniques of test case suits in regression testing are mature and there are some excellent ideas and detailed algorithms for test case suit selection, optimization and reduction. These ideas and algorithms have combined practical applications that spur on their constant enhancing and updating. As a result, they can be applied to solve real problems effectively in actual testing process.Then the paper introduces defect correlation and gives the formulate description about defect correlation and the definition of positive correlation, negative correlation and defect affected degree for defect correlation. Besides, this paper analyzes defect correlation on the basis of test data using Space software. This indicates that positive correlation defects are hidden greatly, which is the biggest barrier in software testing. And they should be checked out in testing process. Still, it explains that the greater the affected degree of a defect is the higher priority it has in testing. Those independent defects can be sorted at last testing. After this analysis, the paper has discussed defect correlation with the impact of system test, reliability assessment, system design and regression testing.Next, this paper divides two chapters to focus on test case suits optimization of regression testing and test case suits reduction of regression testing. The former mainly considers test case sorting. Chapter three presents two optimization strategies for test case suits and fourteen optimization ways for test case suits. In real cases, different ways depend on different situations. In the last of this chapter, it shows a typical way to measure the pros and cons of test case suits optimization methods ---- APFD measure.In Chapter four, this paper discusses test case suits minimization technique of regression testing. This technique is the extent of test case suits reduction technique of regression testing. Firstly, this paper introduces test case minimization and greed algorithm. Secondly, it gives a mathematic model of test case suits minimization. Finally, it studies two minimization methods in details, that is, HGS algorithm and 0-1 integrate programming method.Chapter five is the focus of the full text and it will incorporate defect correlation into techniques of test case suits sorting and minimization. Regression testing implies that the relationship between defects and test cases has been determined through previous testing. So test cases can be sorted according to it. Positive correlation defects should be in front. Then sort defects in the light of their defect affected degree. At last are independent defects. Having this sequence, test cases also have their own order in test case suits. The experiment result of Space software shows that this sorting way is a little better than additional FEP method and proves that this test case suits sorting way incorporated defect correlation into can improve the speed of finding defects in regression testing and can overall detect existing defects in software.Considering the regression test suite reduction techniques with defect correlation, this paper has two situations to discuss: (1) the regression test suit minimization techniques for a certain;(2) HGS algorithm incorporating defect correlation. The idea of the former: given the certain defect is Dx . Before regression testing, compute Dx and the defects which are in the same component with Dx , then fine the defects that have positive correlation with Dx and get defect affected degrees. When doing regression testing, the test cases for defect Dx firstly is executed;then,the test cases for the defects with positive correlation with Dx ;next,the test cases for the defect according to their affected degrees, the highest the first. If the time for regression test is limit and not all test cases can be executed, the test cases for the independent defects and the defects with lower affected degree can be discarded in order to reduce the price.Considering the HGS algorithm incorporating defect correlation, the paper has presented two ways to introduce defect correlation. One is to use defect correlation to reduce test case suit firstly and then to use HGS algorithm to reduce the test case suit secondly. Another way is to add a variable to mark positive correlation defect and a binary array to record two defects that have positive correlation relationship. Every time when selecting defect basis on the cardinal number of their test case suit, if the defect is not positive correlation defect, the algorithm should check whether it has positive correlation relationship with other defects. If has, check whether its positive correlation defect has been marked, that is, whether the test cases of its positive correlation defect have been executed. If executed, run the test cases of this defect and mark this defect. If not, this defect should wait until the test cases of its positive correlation defect have been executed. Using two ways above to do the same experiment, the result shows that the first way is a little better at test case minimization than the second one and a little worse at running speed than the second one.