Lukas Güthing, Mathis Weiß, Ina Schaefer, Malte Lochau
Technical Track
 | Haus der Universität, Schlösslistrasse 5, 3008 Bern, Switzerland |
 | 8 February 2024, 14:00 CET |
 | Mathis Weiß |
 | Tobias Heß |
 | https://dl.acm.org/doi/10.1145/3634713.3634719 |
The goal of sample-based testing of variant-rich software systems is to reduce usually very large configuration spaces to significantly smaller, yet still representative subsets of configurations to be tested for quality assurance. Recent sampling techniques and tools are restricted to finite-dimensional, Boolean configuration spaces specified by a feature model. However, in many modern application domains like cloud computing and cyber-physical systems, customers not only decide about the presence or absence of features in a configuration but also about the multiplicity (number of instances) of configurable resources. Cardinality-based feature models extend Boolean feature models by cardinality annotations and respective constraints to enable multiple, and even potentially a-priori unbounded, copies of features and their respective sub-trees. The resulting infinite and inherently non-convex configuration spaces are no longer tractable by established sampling criteria and corresponding sampling algorithms for Boolean feature models like pairwise feature interaction coverage. In this paper, we first revisit the subtleties of the configuration semantics of cardinality-based feature models.We propose novel sampling criteria explicitly taking multiplicity of feature selections into account. Finally, we present evaluation results gained from applying our tool implementation to a collection of example models, showing applicability of the proposed approach.