Test case generation from mutated task models

Abstract

This paper describes an approach to the model-based testing of graphical user interfaces from task models. Starting from a task model of the system under test, oracles are generated whose behaviour is compared with the execution of the running system. The use of task models means that the effort of producing the test oracles is reduced. It does also mean, however, that the oracles are confined to the set of expected user behaviours for the system. The paper focuses on solving this problem. It shows how task mutations can be generated automatically, enabling a broader range of user behaviours to be considered. A tool, based on a classification of user errors, generates these mutations. A number of examples illustrate the approach.