Sampling strategies for product lines with unbounded parametric real-time constraints

Combinatorial interaction testing (CIT) has been successfully applied to product-line testing for selecting from a usually very large configuration space a relatively small sample of test configurations sufficiently covering critical combinations of configuration options. As most recent CIT techniques like pairwise sampling are limited to finite configuration spaces (i.e., vectors of yes/no options), they are not directly applicable to configuration parameters with a priori unbounded value domains (e.g., for adjusting unlimited resources or non-functional properties). Applying existing sampling strategies to infinite configuration spaces therefore requires further heuristics for selecting a finite subset of parameter-value combinations to be covered. Nevertheless, applying purely black-box heuristics may produce inherently ineffective test suites in which particularly critical parameter-value combinations are missed. In order to tackle this problem, we present a novel methodology for effectively sampling product lines with infinite configuration spaces by means of freely configurable real-time behaviors. To this end, we employ solution-space information obtained from our new modeling formalism, configurable parametric timed automata, to generate samples for covering critical best-case/worst-case execution-time behaviors. We also present a tool implementation which we applied to a collection of subject systems to demonstrate the applicability of our approach.