On the use of mutations and testing for debugging

Abstract

SUMMARYTools for automated fault localization usually generate too many bug candidates depending on the underlying technique. Hence, more information is required in order to further restrict the number of bug candidates. Approaches that rely on specific knowledge of the program to be debugged, such as variable values at specific positions in the source code, are not easily accessible for users especially in the case of software maintenance. In order to avoid this problem, we suggest integrating testing to restrict the number of bug candidates. In particular, we propose computing possible corrections of the program and from this distinguishing test cases. A distinguishing test case is a test that reveals different output values for two program variants given the same input values. Besides the formal definitions and algorithms, we extend the first empirical results of our approach in this paper. The obtained empirical results show that the use of mutations and distinguishing test cases reduces the number of bug candidates by more than 90% on average. Copyright © 2012 John Wiley & Sons, Ltd.

Learning to program using hierarchical model-based debugging

“…The use of a multimodal-reasoning approach during a single debugging session and ranking the diagnoses according to their likelihood to explain the fault was proposed in [21,23]. The introduction of abstractions was proposed in [18] and the idea to find a mutant for a diagnosis candidate that passes all test cases to localize the bug, and also give a potential correction, was proposed in [25]. In our approach, we use the value-based model in different levels of abstraction during a debugging session and leave for the student to select the component (or set of components) for a more detailed debugging, i.e.…”

“…In order to reduce the number of computed diagnoses in model-based debugging, some improvements were proposed in [18,21,23,25]. The use of a multimodal-reasoning approach during a single debugging session and ranking the diagnoses according to their likelihood to explain the fault was proposed in [21,23].…”

Metallaxis‐FL: mutation‐based fault localization

“…Other related approaches include the work of Nica et al . . In this approach, mutants were employed in order to generate tests and thus to augment the utilized test suite.…”

Deriving Test Suites with the Guaranteed Fault Coverage for Extended Finite State Machines

“…Несмотря на большое коли-чество публикаций об автоматическом построении такой модели, в опубликован-ных статьях все примеры обычно ограничиваются небольшими программами, и до-статочно часто такой расширенный автомат строится инженером по тестированию вручную , исходя из неформальных требований к функционированию тестируе-мой программы. Соответственно, практически невозможно сопоставить ошибки в построенной модели, относительно которых и будет гарантирована полнота постро-енного теста, с ошибками в исходной программе, и для генерации тестовой последо-вательности, обнаруживающей такую функциональную ошибку, приходится срав-нивать программу-спецификацию с программой-мутантом для построения соответ-ствующей различающей последовательности [3,4]. В настоящей работе мы предла-гаем метод построения по расширенному автомату множества тестовых последо-вательностей, которые обнаруживают все функциональные ошибки в шаблонной реализации расширенного автомата в языке Java.…”