Each test builds its own scene, runs it with the same engine as the editor, and compares the measurement with the textbook answer.