基于集合进化与占优关系的变异测试用例生成

发布时间：2018-02-28 01:55

本文关键词： 软件测试变异测试测试用例生成变异分支集合进化占优关系占优度　出处：《中国矿业大学》2017年博士论文　论文类型：学位论文

【摘要】：软件测试用于检测软件中可能存在的缺陷,从而改善软件质量。软件测试的一个核心问题是生成高质量的测试用例集,从而检测软件中的缺陷。作为一种面向缺陷的测试技术,变异测试不仅能有效评价现有测试用例集的质量,也能辅助生成高质量的测试用例集,其准则是杀死一定数量的变异体。然而,为数众多的变异体导致高昂的测试代价,严重影响这一技术在实际测试中的广泛应用。已有的基于变异测试的测试用例生成方法,其效率十分低下,因此,研究新的测试用例生成方法是非常必要的。通过约简变异体数量,也非常有助于提高测试用例的生成效率。本文基于变异测试研究新的测试用例生成方法,以高效生成具有很高缺陷检测能力的测试用例集。首先,针对变异测试用例生成问题,提出基于集合进化的测试用例生成方法。所提方法首先基于弱变异测试转化方法,将变异前后的语句组合成能够反映杀死对应变异体的必要性条件,并利用组合后的条件语句构建变异分支。因此,覆盖变异分支真分支的测试用例集,能够以弱变异测试准则杀死对应的变异体。然后,将所有变异分支集成到原程序中,形成新的被测程序,从而将杀死变异体的变异测试问题,转化为覆盖变异分支真分支的分支覆盖测试问题。接着,基于转化后的新程序,建立新的测试用例生成问题的数学模型,对该模型的一次求解,能够生成杀死所有变异体的测试用例集。最后,在利用集合进化方法求解上述模型时,设计具有针对性的适应度函数和进化策略。所提方法能够显著降低基于变异测试的测试用例生成成本。然后,转化后的新程序中集成了为数众多的变异分支,显著增加了转化后新程序的复杂度,从而降低了测试用例生成的效率。针对这一问题,提出基于占优关系的变异体约简方法。所提方法在转化后的新程序中,分析变异分支之间的占优关系,由于覆盖非被占优变异分支的测试用例集,也能够覆盖被占优变异分支,因此,被占优变异分支对应的变异体是冗余的,而非被占优变异分支对应的变异体则是约简后的变异体。在分析非被占优变异分支时,构建占优关系图,并利用图中入度为0的顶点,识别非被占优变异分支。由于仅需覆盖非被占优变异分支,生成的测试用例集就能够以弱变异测试准则杀死所有变异体,因此,所提方法能够约简变异体数量,从而提高测试用例的生成效率。接着,为了提高分析占优关系的效率,增强基于占优关系约简变异体方法的实用性,提出统计占优分析方法。所提方法通过建立统计分析模型,利用测试用例集自动检测新程序中变异分支之间的占优关系,并构建占优关系图。根据占优关系图中入度为0的顶点,得到非被占优变异分支,其对应的变异体就是约简后的变异体。所提方法能够自动检测占优关系,从而有效约简变异体。最后,研究表明具有优先级排序的测试用例集,使得缺陷检测能力强的测试用例能够优先执行,从而降低实际软件测试中缺陷检测的成本。鉴于此,提出基于占优度的测试用例生成方法。所提方法提出占优度的概念,以确定非被占优变异分支的顺序,占优度在数值上等于非被占优变异分支占优的变异分支数量,即占优图中对应顶点的出度;并基于占优度提出两种策略,以选择非被占优变异分支,优先生成缺陷检测能力强的测试用例。此外,针对建立的测试用例生成问题的数学模型,设计和实现了集合进化测试用例生成的原型系统,并详细介绍该系统的使用方法。本文研究成果丰富了变异测试理论,降低了变异测试的成本,提高了测试用例的生成效率。
[Abstract]:Software testing for possible defect detection software, thus improving the quality of software. One of the core issues of software testing is to generate test cases with high quality, so as to detect defects in the software. As a kind of defect oriented testing technology, not only can effectively evaluate the quality variation test of existing test suite, the test case is can generate high quality set, its principle is to kill a certain number of variants. However, lead to high cost of testing for a number of variants, extensive application of this technology in the practical test has been seriously affected. Some test case generation method based on mutation testing, its efficiency is very low, therefore, the method of test case study on the new generation is necessary. Through the reduction of number of variants, also helps to improve the efficiency of test generation. Based on the variation of study to test new measurement Test case generation method, test case generation, has very high detection ability set. Firstly, aiming at the problem of mutation testing case generation, put forward the method of test case generation set. Based on the evolution of the proposed method based on weak mutation testing transformation method, variation before and after the statements together can reflect the necessary conditions of the corresponding kill variants, and the use of conditional statements after the combination of construction branch variation. Therefore, test cases covering the variant branch branch really set, to kill the weak mutation testing criteria. Then the variants, are integrated into the original program branch variation, the formation of the new program to be measured, which will kill the mutation testing problem variants the transformation for branch coverage branch coverage branch variation really test problem. Then, the new program transformation based on the mathematical modeling of the test case generation of new problems For a time, the solution of the model, can generate test cases to kill all variants of the set. Finally, in the ensemble evolution method solving the model, the design of targeted fitness function and evolution strategy. The proposed method can significantly reduce the cost of test case generation based on mutation testing. Then, the new program transformation after the integration of the branch number variation, significantly increases the complexity of the transformed new program, thereby reducing the efficiency of test case generation. To solve this problem, proposed variants of the dominant relationship reduction method. Based on the new program in the modified method, analysis of variation between the dominant branch the relationship, because the coverage is non dominant variation branch of the test case set, can also be covered by the dominant variation of branches, therefore, is the dominant branch variation corresponding variants are redundant, rather than being dominant mutation Variant branch is the corresponding variant after reduction. In the analysis of non dominant mutation branches, construct dominance relationship graph, and use the graph degree of 0 vertices, identification of non dominated variation. Because only non branch coverage is the dominant mutation branch, test case generation set to weak mutation testing criteria kill all variants, therefore, the proposed method can reduce the number of variants, so as to improve the efficiency of test generation. Then, in order to improve the efficiency of the dominant relationship, enhance the practicability of the dominant relationship reduction method based on the statistic variant, dominance analysis method. The proposed method by establishing statistical analysis model, using the test suite of variation the new branch program in automatic detection of dominance relations, and construct the dominant relationship diagram. According to the diagram of 0 dominant degree vertices, non dominant variation was the branch. The corresponding variant variant is after the reduction. The proposed method can automatically detect the dominant relationship, thus effectively reduction variants. Finally, research shows that the test cases with the priority set, the defect detection capability of test cases can give priority to the implementation of strong, thereby reducing the defects of low detection cost in the actual software testing. In view of this, put forward the test test case generation method based on dominance probability. The proposed method put forward the concept of dominant degree, to determine the non dominant branch variation order of dominance degree is equal in value to the non variation branch number of dominant branch variation of dominant, the dominant figure in the corresponding vertex degree; and based on the proposed two dominant strategy. To select the non dominated variation branch, test case generation preferred detection ability strong. In addition, the mathematical model for the test case generation set, design and Implementation The prototype system of set up evolutionary test case generation is introduced, and the use of the system is introduced in detail. The research results enrich the theory of mutation test, reduce the cost of mutation testing, and improve the efficiency of test case generation.

【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP311.53

【相似文献】