Common use of Model-based Test Generation Clause in Contracts

Model-based Test Generation. We are now in a position to generate tests from the model that can be used to verify the implementation. Because, the pilot is free to move the handle backwards and forwards with complete freedom, we now need to constrain the model behaviour to ensure that we can generate a test suite of tractable size. We introduce, in a refinement a ghost variable, handle toggle count which limits the number of times the handle can change position. We can then restrict the value of this count in the ADVANCE test generation tool to limit the scope of the model search space. For instance, restricting the count to two, the pilot can only move the handle twice and a total of 48 tests are generated. If the count is restricted to three, 852 tests are generated. In a structured regression suite mechanism, these tests sets can be run first to detect any gross errors. For more thorough testing, the tests generated for higher handle counts can be run - setting the count to 5, for instance, results in 15400 tests. The coverage obtained by each set of tests is measured by using the ProB simulator to run the generated tests against the model.

Model-based Test Generation. We are now in a position to generate tests from the model that can be used to verify the implementation. Because, the pilot is free to move the handle backwards and forwards with complete freedom, we now need to constrain the model behaviour to ensure that we can generate a test suite of tractable size. We introduce, in a refinement a ghost variable, handle toggle count handle_toggle_count which limits the number of times the handle can change position. We can then restrict the value of this count in the ADVANCE test generation tool to limit the scope of the model search space. For instance, restricting the count to two, the pilot can only move the handle twice and a total of 48 tests are generated. If the count is restricted to three, 852 tests are generated. In a structured regression suite mechanism, these tests sets can be run first to detect any gross errors. For more thorough testing, the tests generated for higher handle counts can be run - setting the count to 5, for instance, results in 15400 tests. The coverage obtained by each set of tests is measured by using the ProB simulator to run the generated tests against the model.