Abstract
Highly-configurable software systems can be easily adapted to address user's needs. Modelling parameter configurations and
their relationships can facilitate software reuse. Combinatorial Interaction Testing (CIT) methods are already often used
to drive systematic testing of software system configurations. However, a model of the system's configurations not conforming
with respect to its software implementation, must be repaired in order to restore conformance. In this paper we extend CIT
by devising a new search-based technique able to repair a model composed of a set of constraints among the various software
system's parameters. Our technique can be used to detect and fix faults both in the model and in the real software system.
Experiments for five real-world systems show that our approach can repair on average 37% of conformance faults. Moreover,
we also show it can infer parameter constraints in a large real-world software system, hence it can be used for automated
creation of CIT models.
[download the pdf file] [DOI]