Evaluating an Automatic Text-based Test Case Selection using a Non-Instrumented Code Coverage Analysis

During development, systems may be tested several times. In general, a system evolves from change requests, aiming at improving its behavior in terms of new features as well as fixing failures. Thus, selecting the best test plan in terms of the closeness between test cases and the changed code and its dependencies is pursued by industry and academia. In this paper we measure the coverage achieved by an automatic test case selection based on information retrieval that relates change requests and test cases. But instead of using off-the-shelf coverage tools, like JaCoCo, we propose a way of obtaining code coverage of Android apk's without instrumentation. This was a basic requirement of our industrial partner. We performed some experiments on this industrial partner and promising results were obtained.