Using FMC for family-based analysis of software product lines

Abstract

We show how the FMC model checker can successfully be used to model and analyze behavioural variability in Software Product Lines. FMC accepts parameterized specifications in a process-algebraic input language and allows the verification of properties of such models by means of efficient on-the-fly model checking. The properties can be expressed in a logic that allows to correlate the parameters of different actions within the same formula. We show how this feature can be used to tailor formulas to the verification of only a specific subset of products of a Software Product Line, thus allowing for scalable family-based analyses with FMC. We present a proof-of-concept that shows the application of FMC to an illustrative Featured Transition System from the literature.

Static Analysis of Featured Transition Systems

“…In Fig. 6, we depict an example FTS modelling the behaviour of a configurable vending machine selling soda and tea from [27], an FTS benchmark which was used in numerous publications, among which [5,6,11,30,32,[42][43][44][45]47]. Its feature model can be represented by the formula s ∨ t over the 4 features { f , c, s, t }, thus resulting in 12 products (i.e., 2 4 −4, excluding ∅, { f }, {c}, { f , c}).…”

Family-Based Model Checking with mCRL2

“…Because for larger SPL models enumerative product-by-product approaches become unfeasible, dedicated family-based techniques have been developed, exploiting variability in product families in terms of features (cf., e.g., [2][3][4][5][6][7][8]). In this paper, we contribute to the field of family-based model checking.…”

“…Thus, an FTS incorporates the behavior of all eligible products, while individual behavior can be extracted as LTSs. Properties of such models can be verified with dedicated SPL model checkers [16][17][18] or, to a certain degree, with single system model checkers [7,12,14].…”

Towards a Feature mu-Calculus Targeting SPL Verification

“…In particular the latter have gained substantial popularity: they offer a compact representation of a family of product behaviors, individually modeled as labeled transition systems (LTS), in a single transition system model in which actions are guarded by feature expressions whose satisfaction (or not) indicate the presence (or absence) of these actions in product behaviors. This has resulted in dedicated SPL model checkers [6,14,19] as well as the application of existing model checkers like NuSMV [15], mCRL2 [8] and FMC [3] to SPL.…”

“…The modal µ-calculus µL, going back to [26], is used to express and model check properties interpreted over LTS, which subsumes more intuitive popular temporal logics like LTL and CTL. The model-checking approaches of [5,14,16,18,19] are based on LTL, those of [4,6,15,17,28] on CTL and those of [3,7,8,10,29,31] on the µ-calculus. In line with the recommendations from [1] to "adopt and extend state-of-the-art analysis tools" and to "analyze feature combinations corresponding to products of the product line", in this paper we propose two variants of µL, coined µL f and µL ′ f , for FTS.…”