What Makes Finite-State Models More (or Less) Testable?

Finite-state machine (FSM) models are commonly used to represent software with concurrent processes. Established model checking tools can be used to automatically test FSM models, but this approach can be very resourceintensive and may not scale to larger models. In this paper we use a partial random search technique for testing FSMs. Such a search has been shown previously to be a surprisingly effective and scalable to very large models. Random search is also very fast and so can be used to gather testability data representing a wide range of FSM models. The TAR2 treatment learner can then summarize that data to determine what FSM attributes characterize models that are easiest to test.

• Type: Conference
• Year: 2002
• Where: KBSE
• Authors: David Owen, Tim Menzies, Bojan Cukic
• Keywords: FSM Attributes Characterize, FSM Models, KBSE 2002, Random Search, Software Engineering