Alias analysis is a method for extracting sets of expressions which may possibly refer to the same memory locations during program execution. Although many researchers have already proposed analysis methods for the purpose of program optimization, difficulties still remain in applying such methods to practical software engineering tools in the sense of precision, extensibility and scalability. Focusing mainly on a practical use for program maintenance activities such as program debugging and understanding, we propose an alias analysis method for object-oriented programs and discuss our implementation. Using this method, we have developed a tool named JAAT. Our proposed method employs a two-phase, on-demand, and instance-based algorithm, in which intra-class analysis is done in phase 1 for whole programs and libraries, and inter-class analysis is done in phase 2 only for a user-demanded target. JAAT can analyze large programs or libraries such as JDK class library. Also, JAAT includes various features for program maintenance activities, such as GUI for displaying aliases, and an XML database for storing analysis information

Program slicing has been used for efficient program debugging activities. A program slice is computed by analyzing dependence relations between program statements. We can divide dependence analyses into two categories, static and dynamic; the former requires small analysis costs, but the resulting slices are large, and in the latter the cost is high but the slices are small. In this paper, we propose a program slicing method for object-oriented programs and evaluate its effectiveness with Java programs. Since object-oriented languages have many dynamically determined elements, static analysis could not compute practical analysis results. Our method uses static and dynamic analyses appropriately and computes accurate slices with small costs.

This paper describes two software maintenance support tools, CCFinder (Code Clone Finder) and JAAT (Java Alias Analysis Tool), for Java programs. CCFinder identifies code clones in Java programs, while JAAT executes alias analysis for Java programs.

In this paper, we propose four slicing methods using both static and dynamic analysis information. (1) Statement-mark slice removes the unnecessary statements using an execution history of the statements. (2) Partial program analysis reduces the static analysis cost using invocation history of procedures. (3) Dynamic data dependence analysis extracts precise data dependence relations using dynamic data dependence analysis. (4) Array and pointer analysis improves the efficiency of (3) by dynamically analyzing pointer and array variables only. Using both dynamic and static information, we show that the precision of the slicing is improved with smaller run-time overhead.