Projected state machine coverage for software testing

Abstract

Our research deals with test generation for software based on finite state machine (FSM) models of the program specification. We describe a set of coverage criteria and testing constraints for use in the automatic generation of test suites. We also describe the algorithms used to generate test suites based on these coverage criteria, and the implementation of these algorithms as an extension of the Murφ model checker[4]. The coverage criteria are simple but powerful in that they generate test suites of high quality and moderate volume, without requiring the user to have a sophisticated grasp of the test generation technology. The testing constraints are used to combat the endemic problem of state explosion, typically encountered in FSM techniques. We illustrate our techniques on several well-known problems from the literature and describe two industrial trials, to demonstrate the validity of our claims.

A quest for appropriate software fault models: Case studies on fault detection effectiveness of model-based test generation techniques

“…The commonly recommended test objectives are similar to (1) transition coverage for EFSMs which provides the worst fault detection effectiveness albeit with the fewest coverage tasks, and (2) transition-pair coverage which provides better fault detection effectiveness than transition coverage, but worse than BZ-TT. Furthermore, we also observed that the expressive power of the notations used to specify the test objectives was not adequate to Hartman et al [6] Murph/ User defined projections on states and transitions No describe our intentions thus providing arguments for a better fault modeling notation for software testing.…”

“…We use an adapted version of the simplified bounded buffer example from [6]. Producers can deposit items in a bounded buffer with a given capacity (2 in this example), and Consumers can withdraw items from the bounded buffer.…”

“…The techniques also provide notations for specifying the desired test objectives. Friedman et al [6] provide notations to specify projections on the underlying state machine. Such projections can be specified to either include or exclude certain states and transitions.…”

“…Specifically, in this paper we compare the methods reported in [9,6,10,11]. We modeled three applications, and generated test cases using all the four techniquesautomatically for 2 cases [6,11] and manually for the remaining two [9,10] due to lack of tool support.…”

“…We modeled three applications, and generated test cases using all the four techniquesautomatically for 2 cases [6,11] and manually for the remaining two [9,10] due to lack of tool support. We also produced implementations for the three applications, and seeded random faults using the known mutation operators for Java [7].…”