Research On Invariant Directed Random Test Case Generation

Computers have more and more important position in human life, as a very important part of computer, the quality of the software is also growing more important. The fault of software will cause very serious consequences, therefore, as a very important approach to find software fault and ensure the quality of software, software testing technology has become more and more important in software engineering and computer science, also become increasingly important. As an important part in software testing, test case generation is also a hotspot. However, the test case generation technique has some obvious problems: excellent technique needs manual intervention, it is time consuming and inefficient, but automatic technique can not generate high-quality test cases. How to generate high-quality test cases automatically is a key problem in test case generation field.This thesis gives a new method based on random testing, remains the advantages of random testing and uses program invariant to avoid the disadvantages, and also designs a fully automatic test case generation framework tool CRT. The study in this article mainly include: 1) Use invariant to determine the effectiveness of test cases, filter invalid cases, effectively reduce the final cases set redundancy; 2) Use program invariant and ineffective test cases to reduce test-case space, which is to reduce the value extent of tested function parameters, to enhance the possibility of selecting useful test cases, further improve the efficiency of the use case generation. 3) Put forward two improve methods: use genetic algorithm to improve test cases set convergence speed and translate invariant into assertions as the effectiveness of the new cases judgment, reduces the time cost, improve test case generation efficiency.The method in this thesis can be completely automatic. The experimental results indicate that, the test case random generation technique described in the thesis could generate test cases satisfied coverage demand without any manual intervention. Compared with the pure random method, this method has lower test cases redundancy, higher coverage and stronger ability to generate test cases which can find program bugs. In this paper, combined with genetic algorithm methods can be well improve the speed of generating useful test cases.