In-depth characterization of exception flows in software product lines: an empirical study

Abstract

Software Product Lines (SPLs) play an essential role in contemporary software development, improving program quality and reducing the time to market. However, despite its importance, several questions concerning SPL dependability did not get enough attention yet, such as: how the exception handling code has been implemented in SPLs? The characteristics of the exception handling code may lead to faulty SPL products? The Exception Handling (EH) is a widely used mechanism for building robust systems and is embedded in most of mainstream programming languages. In SPL context we can find exception signalers and handlers spread over code assets associated to common and variable SPL features. If exception signalers and handlers are added to a SPL in an unplanned way, products can be generated on which variable features may signal exceptions that remain uncaught or are mistakenly caught by common features or other variable features. This paper describes an empirical study that categorizes the possible ways exceptions flow through SPL features and investigates whether some of their characteristics can lead to faulty exception handling behavior. The study outcomes presented in this paper are helpful in several ways, such as: (i) enhancing the general understanding of how exceptions flow through mandatory and variable features; (ii) providing information about the potential problems related to specific kinds of flows detected in this study; and (iii) presenting how a static analysis tool can be used to support the identification of potentially faulty exception handling flows.

An investigation on the evolutionary nature of exception handling violations in software product lines

“…In fact, the tool presented in our study performs a feature-oriented analysis of the exception flows, which allows a more accurate and detailed analysis of how exceptions flow through the code assets implementing the mandatory and variable features of an SPL. Melo et al [10] performed an empirical study that categorized the possible ways exceptions flow through SPL features and investigated whether some of their characteristics can lead to faulty exception handling behavior. However, their work did not tackle the problems that may emerge from evolution scenarios in SPLs.…”

“…Due to the complexity of manually collecting information about EH design rules, we extended PLEA -a feature-oriented static analysis tool [10]. In our current version, PLEA uses the WALA framework (http://wala.sourceforge.net) to execute static analysis in order to build the set of EH design rules based on the WALA call-graph algorithms, specifically the k-CFA family.…”

“…In a recent study, Melo et al [10] characterized: (i) the kinds of exception flows that are thrown in the implementation of two SPLs in terms of interaction between core and variable assets, and (ii) which of them can potentially cause faulty exception handling flows. Nevertheless, this study did not explore how the exception handling policies have evolved over the different SPLs releases.…”

“…These EH design rules define exceptions which can be thrown by an SPL module (signaler) and modules (handlers) which are responsible for handling some thrown exception. Therefore, the system dependability relies on the obedience of such rules [10].…”