Iterative execution-feedback model-directed GUI testing

摘要

Current fully automatic model-based test-case generation techniques for GUIs employ a static model. Therefore they are unable to leverage certain state-based relationships between GUI events (e.g., one enables the other, one alters the other's execution) that are revealed at run-time and non-trivial to infer statically. We present ALT - a new technique to generate GUI test cases in batches. Because of its "alternating" nature, ALT enhances the next batch by using GUI run-time information from the current batch. An empirical study on four fielded GUI-based applications demonstrated that ALT was able to detect new 4- and 5-way GUI interaction faults; in contrast, previous techniques, due to their requirement of too many test cases, were unable to even test 4- and 5-way GUI interactions.