Mining change history for test-plan generation (doctoral symposium)

Abstract

Regression testing is an essential step in safeguarding the evolution of a system, yet there is often not enough time to exercise all available tests. Identifying the subset of tests that can reveal potential issues introduced by a change is a challenge. It requires identifying the tests that test the chan- ged part of the software. Furthermore, and more challeng- ing, it requires identifying the parts of the system that are potentially affected by that change, a task typically done by means of static program analysis. In this doctoral research, we investigate an alternative approach, using software repos- itory mining. We propose a method that mines the change history of a system to uncover dependencies, and uses these for test-selection and test-prioritization. By reducing the amount of test to exercise, and limiting time spend on test- plan creation (i.e., selecting and prioritizing tests), the aim of our approach is to increase cost-effectiveness of software regression testing.