An Improved Confo mance Testing Method

In this paper, we present a novel conformance test suite derivation method. Similar to the HIS method, our method uses harmonized state identifiers for state identification and transition checking and can be applied to any reduced possibly partial deterministic or nondeterministic specification FSM. However, in contrast with the HIS method, in the proposed method appropriate state identifiers are selected on-the-fly (for transition checking) in order to shorten the length of the obtained test suite. Application examples and experimental results are provided. These results show that the proposed method generates shorter test suites than the HIS method. Particularly, on average, the ratio of the length of the test suites derived using the proposed method over the length of corresponding suites derived using the HIS method is 0.66 (0.55) when the number of states of an implementation equals to (is greater than) the number of states of the specification. These ratios are almost independent of the size of specifications. Many FSM-based test derivation methods have been developed for conformance testing of communication protocols and other reactive systems [2,3,10,12,14,15,17]. Well-known methods are called the W [2, 14], partial W (Wp) [3], HIS [10,17], and generalized Wp (GWp) [7, 8] test derivation methods. For related surveys the reader may refer to [1,6,13,16]. In [2,3,10,14,15,17] testing methods, one usually assumes that not only the specification, but also the implementation can be modeled as a deterministic FSM, while in [7,8] the specification and the implementation are modeled as non-deterministic FSMs (NFSMs). If the behavior of a (deterministic/nondeterministic) implementation FSM is different than the specified behavior, the implementation contains a fault. The above methods, each provides the following fault coverage guarantee: If the specification can be modeled by a (reduced) FSM with n states and if a corresponding implementation can be modeled by an FSM with at most m states, where m is larger or equal to n, then a test suite can be derived by the method (for this given m) and the implementation passes this test suite if and only if it conforms (i.e. is equivalent) to the specification. A test suite is called m-complete [11] if it detects any nonconforming implementation with at most m states. Guessing the bound of m is an intuitive process based on the knowledge of a specification, the class of implementations which have to be tested for conformance and their interior structure [1]. All of the above methods assume that a reliable reset is available for each An Improved Confo mance Testing Method Rita Dorofeeva, Khaled El-Fakih, and Nina Yevtushenko 1 Tomsk State University, Russia 2 American University of Sharja, UAE [email protected], [email protected], [email protected]