Research On The Models And Tools Of Automated Regression Testing Based On FitNesse

Software testing is one of the crucial steps in quality assurance engineering and robustness improvement. It plays a critical role in the whole software development process. In light of that test automation is trend in software testing, lots of relative work has been done. Many automated testing tools are developed for automated test. But most of early software testing tools display some common shortcomings, such as manual designation of system under test and testing methods, manual compilation and debugging of generated test scripts. It may cause poor reusability and maintainability about test scripts. In other words, original test scripts could not be performed if the system under test changed.To design and generate test case which play a important part in the process of software testing. It requires a lot of test cases for comprehensive testing, but the way of manually generate test case is very difficult. Therefore, the automatic generate of test case which has become an important issue in software testing.In this paper, the technology of automatic generate test cases and the algorithms of automatic test case generation have been researched. Inspired by the algorithm of optimization parameter's probability distribution proposed by Simon Poulding and John A.Clark, this paper proposes an approach for test case generation based on optimization of parameter and boundary value analysis(short for TCG-OPBVA). Compared with the the algorithm of optimization parameter's probability distribution proposed by Simon Poulding and John A.Clark, the key differences are as follows:(1) Bayesian network structure automatic derived by control flow graph. It construct the control flow graph of program by tools. Then, it automatic build the Bayesian network structure of program's parameter from control flow graph.(2) The fitness function. Because of the structure of bayesian network has automatic constructed, the search algorithm only need to the parameter of bayesian network, so the diversity constraint can be removed, while reducing the effect made by the coverage constraint and the diversity constraint.(3) The combination of boundary value analysis. After sampling the test set, it will combine with the right amount of test cases generated by boundary value analysis to form the final set of test cases, to form the final set of test cases.To implement better automated regression testing, this paper analyzes the acceptance testing framework, FitNesse. In this paper, a automated regression testing model is proposed. And then, a practical FitNesse-based tool for automated regression testing (Automated Regression Testing Tool, short for ARTT) in Java programs is designed and implemented. It has the following characteristics:(1) Test cases automatic generation. The tool parsed system under test and extracted information from those tested classes. Meanwhile, it generated the XML documents which used to store the information of the parameters. And then, ARTT edited the information of the parameter values. Applied the algorithm of TCG-OPBVA, ARTT could generate test cases with the parameter values automatically.(2) Fit tables automatic generation. With the information of tested classes and generated test cases, ARTT could produce Fit tables with test cases in the FitNesse environment.(3) Test code automatic generation. With the information of tested classes Fit tables, ARTT could generate test code in the FitNesse environment. It stored input values and expected output in the Fit table alone which similar to the data-driven script. It was easy to add, delete and modify the test data. In addition, test on an approach can be done repeatedly without test code modification. Moreover, ARTT could generate test code with more readability, reusability, maintainability and so on.Then, this paper used an example to verifying the function of ARTT. Experimental results show that the automated regression testing model proposed in this paper is advisable, and show that apply the algorithm of TCG-OPBVA, this tool could easily perform automation of regression testing for java programs.