A Comparative Evaluation of Unit Testing Techniques on a Mobile Platform

Software testing is generally an expensive, ad hoc and unpredictable process. Therefore, a better understanding of and an improvement of unit testing methods may improve the overall software testing process leading to more efficient test method selection and increased fault detection accuracy. The aim of this study is evaluate Black Box and White Box unit testing methods for the Android operating system. While several investigations have looked at unit testing using Black Box and White Box techniques, our research focuses explicitly on unit testing for Android applications. Test cases were created using JUnit and the program under test was seeded with errors by an outside party. We evaluated the performance of these two test methods on the difficulty of developing test cases for each method and the effectiveness of finding software errors in a controlled environment. Our results showed benefits and challenges for both testing techniques. Inferences and implications are also presented.

1. Introduction

In software testing [1], [5], [15], [16], [17], [18], [19], [20], an element under test is exercised to identify the presence of errors and to ensure that the software performs as expected. Constrained by computational resources and time, exhaustive testing is impractical for all but the most trivial applications. As a result, selecting an optimum testing technique for identifying errors in a program at the unit testing [2], [6], [7], [8], [14] level is an important challenge in software engineering.