On Introducing Automatic Test Case Generation in Practice: A Success Story and Lessons Learned

Abstract

The level and quality of automation dramatically affects software testing activities, determines costs and effectiveness of the testing process, and largely impacts on the quality of the final product. While costs and benefits of automating many testing activities in industrial practice (including managing the quality process, executing large test suites, and managing regression test suites) are well understood and documented, the benefits and obstacles of automatically generating system test suites in industrial practice are not well reported yet, despite the recent progresses of automated test case generation tools. Proprietary tools for automatically generating test cases are becoming common practice in large software organisations, and commercial tools are becoming available for some application domains and testing levels. However, generating system test cases in small and medium-size software companies is still largely a manual, inefficient and ad-hoc activity.This paper reports our experience in introducing techniques for automatically generating system test suites in a medium-size company. We describe the technical and organisational obstacles that we faced when introducing automatic test case generation in the development process of the company, and present the solutions that we successfully experienced in that context. In particular, the paper discusses the problems of automating the generation of test cases by referring to a customised ERP application that the medium-size company developed for a third party multinational company, and presents ABT 2.0 , the test case generator that we developed by tailoring ABT , a research state-of-the-art GUI test generator, to their industrial environment. This paper presents the new features of ABT 2.0 , and discusses how these new features address the issues that we faced.

More Effective Test Case Generation with Multiple Tribes of AI

“…Although automated TCG is becoming more common in large software companies, the widespread adoption of these techniques is lacking behind [10]. One of the key challenges in automated TCG is the size of the input space [13].…”