A dataset of strong dependency and conflict graphs generated from Boolean formulas encoding the variability of highly configurable software systems
TL;DR
This Data Descriptor documents a dataset of 5,709 records that makes indirect relationships between dependencies and conflicts indirectly cause in variability models explicit and helps test configuration and sampling methods and supports research on conflict detection, impact analysis, and automated reasoning.
Abstract
Variability models describe the configuration options of highly configurable software systems and the direct constraints between them. The dependencies and conflicts those constraints indirectly cause are hidden in the model and too hard to check manually, making it difficult to analyze very large systems like the Linux kernel. This Data Descriptor documents a dataset of 5,709 records that makes those indirect relationships explicit. Each record includes two graphs based on the Boolean formula that represents the variability model of one configurable system: a strong dependency graph and a strong conflict graph. We created each record from publicly available Boolean formulas that other researchers made from variability models and shared under an open license, using our own tool: a C++ program that runs two algorithms based on step-by-step satisfiability solving. The resulting graphs range from 99 to 35,907 options, with a median of 2,874. Provided in the Pajek NET format with clear documentation, and released with the input formulas and the code that builds and checks the graphs, the dataset helps test configuration and sampling methods and supports research on conflict detection, impact analysis, and automated reasoning.
