Automatic Dependency Identification for Isolation of Software Errors

In a typical software development cycle, a large amount of resources is spent locating errors. To find the cause of an error, developers typically execute the programs step by step, checking many operations and variables that might not be relevant to the problem in question. It is difficult to detach the relevant parts that triggered the error from the source code. This paper presents a tool that provides the programmer with a view of the program execution, containing the statements and variables that might lead to an observed error. Thus, simplifying and facilitating the debugging process. This tool was developed in Java for Java programs. First, it instruments the application to save the information that we consider relevant: variable assignments, method invocations, and conditional statements. Then, the saved data from the execution that reproduces the problem is presented in a variable dependency graph and a graph of the statements that induced those dependencies. The tool introduces a performance penalty which is quite affordable considering the traditional process of software development and testing. Keywords—Software Errors, Debugging, Diagnosis, Instrumentation