Research On Regression Testing For Finite State Machines

We are living in a rapidly changing world,software systems keep evolving to adapt to the continuously incoming requirements.Regression testing is an important activity of retesting a software after it has been modified to provide confidence that existing,changed,and new parts of the software behave correctly.In model based testing,the models of systems are also evolving to reflect changes in their specifications.Though the test cases can be automatically generated,it's not feasible in terms of cost to regenerate and execute all test cases every time a system is modified.Effective and efficient techniques are still needed to reduce the effort required for regression testing.However,we are not aware any research on regression testing for the widely-used model,Finite State Machines(FSMs).In this paper,we manage to address this problem.That is,when the specification of a system modeled as FSM is modified,how to construct and execute test cases based on the existing test suite effectively and efficiently to guarantee that the new implementation conforms to the modified specification.The main work and results includes:1.The notions,ideas,and methods of testing from FSMs are surveyed in detail to be used as the foundations of our research on regression testing for FSMs.2.A test suite augmentation method for FSMs is proposed to guarantee the complete fault detection power of the test suite for the modified model of the new version.It first identifies the differences between the original model and the modified model.According to the differences,obsolete test cases are removed,reusable test cases are kept;when some test cases are invalid or lacking,the test generation methods will be employed to create new ones to replace or complement them.3.Optimization techniques such as regression test selection,test case prioritization and test suite minimization are combined with the proposed test suite augmentation method.These techniques are originally applied just on the existing test cases.They don't consider if these test cases are adequate and leave the task of creating new test cases to the engineers.We apply them on the augmented test suite to reduce the effort required for regression testing without loss of fault detection power.It's also natural and straightforward to apply those optimization techniques since the test cases have already been classified during the augmentation process.4.The proposed approach is implemented and evaluated by experiments.The results on a large set of randomly generated FSMs and a group of real-world FSM specifications show that our approach is able to reuse a huge proportion of previous test cases during augmentation,thus reduce a large amount of effort required for regression testing by optimization,and can achieve this even in shorter time than plain test generation.The less a system is modified,the more test cases can be reused,and the more cost will be reduced.This is especially meaningful for evolving softwares.