Fault Localization With Weighted Test Model in Model Transformations

Abstract

Model transformations and model-driven engineering (MDE) have been applied widely in service-oriented architecture based information systems. To support the development of such a service-oriented information system, it is necessary to guarantee the quality of model transformations. With the increasing complexity and scale of model transformations, debugging of model transformations becomes more and more time-consuming and difficult, there is an increasing need to rely on efficient and accurate fault localization approaches to help with debugging. Among the existing fault localization approaches, the spectrum-based fault localization (SBFL), as a dynamic analysis method, mainly used the coverage information and execution results of the rules of model transformation to estimate the probability of each rule may be faulty. However, there are many false-positive and false-negative results in the rule coverage information, the accuracy of the SBFL is not ideal, so we consider mining the impact of covered range in different test models to further improve the effectiveness of fault localization. In this paper, we propose an optimized strategy of fault localization based on the impact of the test model, according to the covered range of test models, weight values are assigned to different test models, and the statistical coverage information of rules are weighted and adjusted accordingly. We compare the proposed approach with the SBFL, the experimental results show that under the same techniques for computing the suspiciousness, our approach can help locate around 26% more faulty rules in the ranking Top-1 of the suspicious list than the SBFL, the effectiveness of fault localization techniques can be improved by 50.42% in the best case and 8.9% in the average case.

Model Transformation Testing and Debugging: A Survey

“…Later, Li et at. [75] present a similar approach, where they propose to use weighted test models as well as weighted rule coverage to improve the performance of SBFL.…”

A Survey of Challenges in Spectrum-Based Software Fault Localization

“…SBFL is a dynamic program analysis technique that is performed through program execution [10], [11]. The goal of SBFL is to address the problem of finding the root causes of bugs by utilizing information from program elements executed by test cases, in particular the outcomes of tests and their code coverage [12].…”

Revisiting Fault Localization Techniques for Model Transformations: Towards A Hybrid Approach.

“…Burgueño et al (Burgueño et al 2015) proposed a static fault localization approach based on the extraction of metamodel footprints from the contracts and the transformation implementation for locating the bug. In contrast, Spectrum-Based Fault Localization (SBFL) for model transformations (Troya et al 2018;Li et al 2020;Du et al 2020) is a dynamic approach that requires input models and to execute the model transformation for locating the bug. Both approaches rank the transformation rules according to the likelihood that they contain the bug.…”

“…This results in the creation of a program spectrum that details the execution information of the program from a certain perspective, such as branch or statement coverage (Harrold et al 2000). Troya et al presented an approach to apply SBFL to model transformations (Troya et al 2018)-other authors also focused on this technique recently (Li et al 2020;Du et al 2020). As explained in (Troya et al 2018), the approach requires test cases that comprise input models for exercising the model transformation.…”