Flip van Spaendonck: Efficient dynamic model based testing of models with run-to-completion semantics


Event Details


Model based testing (MBT) provides an efficient and automated approach to finding discrepancies between software models and their implementation. If we want to incorporate MBT into the software development process, i.e. code changes are only accepted if the implementation remains in conformance with the model, then MBT must be able to thoroughly test the entire model in as little time as possible. However, current academic MBT tools traverse models at-random. While this allows them to find most discrepancies relatively quickly, it is inefficient at testing all or almost all states in the software model and is thus lacking in thoroughness.

We provide a new traversal strategy and show that the at-random strategy needs to take significantly longer to test  >95 of all states. The benchmarks are performed on pseudo-randomly generated LTS’s, which we show to be representative of systems with run-to-completion semantics.