Regression test selection based on intermediate code for virtual machines

Abstract

Regression testing is testing applied to software that has been modified. It basically entails re-testing the software with previous test cases to confirm that the modifications made to the software do not have an adverse effect. But re-executing all test cases is normally cost prohibitive, and thus much research has been done on selecting test cases from a test suite without compromising the reliability of the software. These regression test selection techniques find test cases that will not detect any bugs in the modified software, and delete those from the next regression test suite. However, these techniques are based on analysis of source code. Recent programming environments have seen a proliferation of virtual machines. For example, programs written in Java and with the Microsoft .Net Framework are compiled into a platform-independent intermediate code which is executed. Such code could also be used for regression test selection. This especially holds for the Microsoft .Net Framework which handles various programming languages, such as Visual Basic and C++. Thus, this paper presents a safe regression test selection technique for virtual machine based programs. We especially target the Microsoft .Net Framework. Evaluation on 10 different examples resulted in an average of a 40.4% decrease in the cost of regression testing.