Abstract

Testing is the most common activity to validate software systems and plays a key role in the software development process. In general, the software testing phase takes around 40-70% of the effort, time and cost. This area has been well researched over a long period of time. Unfortunately, while many researchers have found methods of reducing time and cost during the testing process, there are still a number of important related issues such as generating test cases from UCM scenarios and validate them need to be researched.
As a result, ensuring that an embedded software behaves correctly is non-trivial, especially when testing with limited resources and seeking compliance with safety-critical software standard. It thus becomes imperative to adopt an approach or methodology based on tools and best engineering practices to improve the testing process. This research addresses the problem of testing embedded software with limited resources by the following.
First, a reverse-engineering technique is exercised on legacy software tests aims to discover feasible transformation from test layer to test requirement layer. The feasibility of transforming the legacy test cases into an abstract model is shown, along with a forward engineering process to regenerate the test cases in selected test language.
Second, a new model-driven testing technique based on different granularity level (MDTGL) to generate test cases is introduced. The new approach uses models in order to manage the
complexity of the system under test (SUT). Automatic model transformation is applied to automate test case development which is a tedious, error-prone, and recurrent software development task.
Third, the model transformations that automated the development of test cases in the MDTGL methodology are validated in comparison with industrial testing process using embedded software specification. To enable the validation, a set of timed and functional requirement is introduced. Two case studies are run on an embedded system to generate test cases. The effectiveness of two testing approaches are determined and contrasted according to the generation of test cases and the correctness of the generated workflow. Compared to several techniques, our new approach generated useful and effective test cases with much less resources in terms of time and labor work.
Finally, to enhance the applicability of MDTGL, the methodology is extended with the creation of a trace model that records traceability links among generated testing artifacts. The traceability links, often mandated by software development standards, enable the support for visualizing traceability, model-based coverage analysis and result evaluation.