Exception Fault Localization in Android Applications

Abstract

In software programs, most of the time, there is a chance of error, even though they are tested carefully. Finding error-related pieces of code is one of the most complicated tasks and it can make incorrect results if done manually. Semi-automated and fully-automated methods have been introduced to overcome this issue. The rapid growth of developing Smart Mobile Applications (SMAs) in recent years, competition among the development teams and many other factors have increased the chance of errors, and hence, the quality of these applications have reduced. There are two approaches to test SMAs in order to reach a high degree of quality: (i) using existing traditional methods and adapting them to SMA environments and (ii) introducing new special methods for SMAs. In this paper, we introduce a semi-automated hybrid method to localize exception errors in Android applications. The proposed approach includes the following three phases: extraction, execution and evaluation. In the extraction phase, all the information about the application under the test (AUT) such as the activity and object properties will be extracted. In the execution phase, we generate a set of test cases and execute them on the AUT. In the evaluation phase, we use test case traces, variable value patterns, and backward static slicing techniques to rank lines of application source code with respect to their relevance to that fault. To support localization of multiple errors in a single run of the approach, we introduce a classification measure on test case traces. Evaluations on nine open source Android applications of different sizes show that our method is effective in practice.