Systematic Testing of Software Architectures in the C2 Style

Abstract

Abstract. The topic of software architecture (SA) based testing has recently raised some interest. Recent work on the topic has used the SA as a reference model for code conformance testing, to check if an implementation fulfills (conforms to) its specification at the SA level. In this context, on previous papers, we have analyzed: i) how suitable test cases can be "selected" from the SA specification and ii) how they may be "refined" into concrete tests executable at the code level. While the selection stage has been done systematically, the refinement step has been left to be done manually, based on the software engineer knowledge on how to map "abstract values of the specification to the concrete values of the implementation". In this paper, we extend previous approaches, by providing a systematic way to perform the refinement step. We show how choosing a specific architectural style, which supports implementation and facilitates the mapping among SA-based and code-based test cases, a completely systematic SA-based testing approach can be delivered.

Rethinking Software Testing Based on Software Architecture

“…As we can see from Figure 2 of software testing's dreams [9], anti-model-based testing for test-based modeling and controlling evolution for efficacy-maximized test engineering are involved in architecture-based testing [10], which is the direction of future work.…”

Towards software architecture-based regression testing

“…Figure 2 summarizes the activities required by SA-based conformance and regression testing. While SA-based conformance testing has been already analyzed in [17,18], goal of this paper is to focus on SA-based regression testing.…”

Towards software architecture-based regression testing

“…We have previously performed the SAbased conformance testing technique over this system, as reported in [18]. Now, going through this example, we illustrate how goals #1 and #2 are accomplished in our SA-based regression testing technique.…”

“…Given the strong relationship among C2 architectural specification and implementation, the mapping between architectural test cases and code-level test cases is straightforward. Applying the mapping (as further discussed in [18]), we created 12 different test frames for architectural test case #1 (e.g., BP.addCall at lower, same or higher floor level than ADT1; BP.addCall up and down; ADT1 up and down), and 6 to ATC #2. Figure 5 reports the 18 test cases used to test the elevator system.…”

“…Following the four steps depicted in Figure 2 (right side), we took into consideration two different implementations of the Elevator system, architecture version 2: P which is the regular implementation and P' a faulty version (the two implementations were produced for testing purposes in [18]). In order to regression test such implementations, we applied the activities described in Section 4.2: we built a graph representation of P and P' (Step A) and we compared the two implementations (Step B).…”