Better testing through oracle selection

Abstract

In software testing, the test oracle determines if the application under test has performed an execution correctly. In current testing practice and research, significant effort and thought is placed on selecting test inputs, with the selection of test oracles largely neglected. Here, we argue that improvements to the testing process can be made by considering the problem of oracle selection. In particular, we argue that selecting the test oracle and test inputs together to complement one another may yield improvements testing effectiveness. We illustrate this using an example and present selected results from an ongoing study demonstrating the relationship between test suite selection, oracle selection, and fault finding.