| Popis: |
One way to implement model-based software product lines (MBSPLs) is to use a transformational approach known as Delta Modeling (DM). Here, an MBSPL is implemented by one core model and a set of delta modules. Delta modules define model transformations using edit operations which add, remove or modify model elements. Editings of different delta modules can be in conflict or depend on each other, leading to conflict and dependency relations between delta modules. Conflicts and unfulfilled dependencies can cause the generation of a product to fail or to lead to invalid models. In order to spot such defects, one needs analysis tools for each modeling (sub-)language used. Existing generic approaches to statically detect such defects in a language-agnostic manner analyze pairs of delta modules. However, the pairwise approach can lead to false positives, i.e., conflicts and unfulfilled dependencies are reported although product generation does not fail. Following the idea of family-based analysis, this paper presents a new approach to detect pseudo defects resolved by "healing effects" implied by the network of dependencies. These effects typically occur when a delta module (partially) reverts the effect of a preceding delta module. We have implemented our approach within the SiPL framework and evaluated our family-based analysis using a realistic MBSPL known as Body Comfort System (BCS). |
