Static Data Race Detection for Java Programs with Dynamic Class Loading

Abstract

Multi-thread programs are likely to have bugs in mutual exclusion among threads. Data race is one of the problems that occur in wrong mutual exclusion. There are two kinds of methods of data race detection: dynamic analysis detection and static analysis detection. Dynamic analysis detection is to detect data race by analyzing the results or processes of program executions. Static analysis detection is to detect data race by analyzing source codes of programs. A method of static analysis detection of data race for Java programs has been proposed, but this method cannot handle dynamic class loading. This paper proposes a method of static analysis detection of data race for dynamic class loading and implements this method. This paper also evaluates the implementation by experiments using Java programs that have dynamic class loading. This experiment shows that the proposed method detects data races that cannot be found by the previous data race detection method. The result of experiment shows advantages of the proposed method.