Trace-Based Code Generation for Model-Based Testing

Model-based testing can be a powerful means to generate test cases for the system under test. However, creating a useful model for model-based testing requires expertise in the (formal) modeling language of the used tool and the general concept of modeling the system under test for effective test generation. A commonly used modeling notation is to describe the model through an existing programming language. This paper presents a technique to automatically generate an initial model describing the system from execution traces, using a common programming language notation. Turning this initial model into a full model to be used for model-based testing of the system under test then requires minimal effort compared to writing a model from scratch. This is illustrated by a case study application to a software component, which revealed real faults in its implementation.