An Automatic Test Data Generation Technique Based On Dataflow Analysis

Software Testing plays increasingly important role to ensure the software quality in software development.Dataflow Testing is a white-box test technique that has been widely applied to test object-oriented software with sufficient code coverage.But the existing research work mainly cares the computation of method sequences,rather than feasible test data to run these sequences.That's say they can't support automatic testing. Although the existing random test data generators can create test data automatically,the generation process is blind yet essentially non-goal driven,it may generate unfeasible test data and thus fail to cover the intended testing requirements.To solve this problem,this thesis presents an automatic test data generation technique that takes the merits of genetic algorithm(GA) and dataflow analysis.Our method extends the dataflow testing technique in that it exploits the searching capability of GA with the aim to sufficiently cover the def-use pairs.This thesis starts with some concepts about software testing and the introduction of current techniques for generating test cases.Next we present our automatic dataflow testing method,which consists of the dataflow computation and the DTGA algorithm. Then we describe the design and implement of our protype system.Experimental results show that our method can work well on well on a variety of programs,outperform random generation and pure dataflow testing technique in terms of running time and coverage rate.In the end we conclude and suggest our future work.