Designation of Exception handling code is very important in defensive programming. But, there was little research on how to test the exception handling code. Recently, exception testing criteria which are based on structural testing has been proposed. And exception-related mutation operators based on mutation testing has been proposed, too. There are 6 different exception testing criteria. They are all-e-defs, all-e-uses, all-e-du-paths, all-e-acts, all-e-deacts, and all-e-ad-paths. Exception testing criteria covers definition and use node of exception variable and activation and deactivation node of exception objects. Exception-related mutation operators inject syntactic faults to the exception handling code. Test case which kill these mutants are judged as an effective test case. There was deep research on exception testing criteria, but the effectiveness of exception-related mutation operators was not substantially verified.
In this thesis, we compare the strength of data flow testing and mutation testing for detecting exception handling faults to evaluate the effectiveness of exception-related mutation operators. Comparison is conducted by empirical evaluation. We chose the sample programs generated by public users. Then we designed test suites. One is adequate to mutation testing, and the other is adequate to data flow testing. Finally, we measured the coverage of mutation adequate test suites in view of data flow testing, and the coverage of data flow adequate test suites in view of mutation testing. Experimental results showed that exception-related mutation operators cover the associations adequate to all-e-acts, alle-deacts, and all-e-ad-paths criteria based on the relationship of exception object activation. We can confirm the extent of coverage that mutation adequate test suites selected by random fault injection process. And this experimental results help tester to make a decision whether the test suites designed by mutation testing is need to improve to be adequate to exception testing criteria.
프로그램 개발자가 예상하지 못한 상황에서도 프로그램이 안정적으로 동작하기 위해서는 예외 상황 처리 부분이 올바르게 작성되어야 한다. 객체 지향 언어인 Java는 임베디드 소프트웨어를 개발할 때 많이 사용되는데, 임베디드 소프트웨어에서는 메모리 부족 등으로 인하여 예상하지 못한 상황이 자주 발생한다. 따라서 Java에서 예외 상황 처리가 더욱 중요하게 다루어져야 한다. 예외 상황 처리 코드를 테스트하는 방법으로서 프로그램의 제어 흐름 그래프 분석에 기반한 구조적 테스팅 기법과 프로그램에 예외 처리 뮤테이션 연산자를 적용시켜 보는 뮤테이션 테스팅 기법이 제안되었다. 본 논문에서는 뮤테이션 테스팅과 구조적 테스팅의 오류를 찾아내는 능력을 비교함으로써 예외 처리 뮤테이션 연산자의 효용성을 알아내고자 한다. 이 연구를 통해, 예외 처리 뮤테이션 연산자를 이용하여 설계한 테스트 케이스가 실제로 프로그램을 어느 정도로 테스트하는지 알아낼 수 있다.
