Source publication: [SHT18]
Schuts et al [SHT18] describe a case study carried out at Philips in which models created with a legacy tool (Rhapsody) are transformed to models that can be used by another tool (Dezyne). The transformation is established by means of a DSL for the legacy models. Model learning was appied to increase confidence in the correctness of the generated code. Two versions of state machine code, generated by Rhapsody and Dezyne, were stimulated by all possible inputs and the resulting outputs were examined by LearnLib. The two models constructed by LearnLib were compared by the equivalence checker of the mCRL2 tool set. With this approach two errors were found in the Dezyne models that were not detected by the existing regression test set.