Scalable, Flow-Sensitive Type Inference for Statically Typed Object-Oriented Languages

An important problem in the analysis of programs written in object-oriented languages like C ++ /Java is to determine the values that a pointer variable ( a reference to an object in Java ) can have at run-time. This information can be used for applications such as virtual function resolution, side-effect analysis, detecting memory errors etc. We address issues involved in designing a scalable, flow-sensitive solution for this and related compile-time analysis questions. First we present a flow-sensitive solution which uses conditional points-to analysis. Although we show that it is better than a solution using alias analysis, it has several limitations which affect its scalability and precision. Next we identify properties necessary for any scalable, flow-sensitive algorithm solving these problems. Using this characterization, we propose two techniques: preprocessing using types and analysis-usingabstract -values, which remove many of these limitations. We show that these techniques..