Partition-based coverage metrics and type-guided search strategies in concolic testing for javascript applications

Abstract

JavaScript broadens its uses from client-side web applications to mobile and smart appliance ap- plications, but testing JavaScript programs is not yet satisfactory. Researchers have not spent much attention on testing JavaScript programs, and existing coverage metrics and testing mechanisms for C and Java may not be applicable to JavaScript because of its extremely dynamic semantics without any compile-time checks. Because, in JavaScript, any variable may have six kinds of types during program evaluation, test cases with 100% coverage levels in terms of the existing coverage metrics for statically typed languages may miss test cases revealing faults in JavaScript. In this thesis, we identify characteristics of JavaScript that make thorough testing of JavaScript ap- plications more difficult than testing C and Java programs. To address such characteristics, we propose new partition-based coverage metrics that expose implicit execution flows using varying types. To gen- erate test cases satisfying the coverage metrics effectively, we develop type-guided search strategies for concolic testing using static analysis results. We evaluate the new coverage metrics and search strategies for concolic testing with open-source Tizen sample web applications. The experimental results show that our partition-based coverage metrics are practically usable in JavaScript testing and the type-guided search strategies generate test cases that reach high levels of the coverage metrics effectively.