In automotive software developments, since the types of software functionalities are depending on automotive engineering domains such as powertrain, body, and chassis, software logic and data processing, code complexity, and its reliability are also depending on them. Therefore, it has some challenges that monolithic code quality measures are applied to software code for all domains. In addition, imprecise criteria for the measures also can make software developers and testers confused whether their code verifications are enough or not.
This paper proposes domain-specific code quality measures and precise criteria by combining a new functionality model, named Abstract Function Model (AFM), and shows results of automotive software functionality analysis using the model. Using 8 real automotive software projects, we derived statistics of software code on specific automotive engineering domain and identified code quality measures from the statistics. For the identified code quality measures, we tactically modified the legacy software code quality measures to be more reasonable and suitable for domain-specific code verification in terms of coding rule checking, potential semantic error checking, and code structural metrics.
Our results consist of detailed code quality measures, quantitative code verification criteria, and its configuration information on each domain. At the end of this paper, we show that our result can fulfill the software functional safety requirements (ISO 26262-6). As a result, we expect that our result could reduce confusion of code verification and improve code quality for automotive software.