Semantic based Cross-Language Clone Related Bug Detection

Code clones are widespread in software since programmers always reuse code to reduce programming effort. As programming languages are continuing to evolve and morph, code clones also widely exist across different languages for platform compatibility and adoption. Although code clones can improve development efficiency, they are prone to introducing bugs. Existing code clone detection technologies, however, mainly focused on single programming language or syntactical features of code. The syntax of different programming language are diverse because of syntax sugar, and many cloning pairs are semantic related instead of syntactic similar, such as Type 4 clones. To bridge the gap between syntax and semantic, and detect clone-related bugs more accurately, we explore an IR (Intermediate Representation) based method to represent code semantic representation information of multiple language code. We utilize graph neural network to learn code semantic representation. Through the semantic representation, we can detect more cross-language clone related bugs across multiple language.