Reducing the search space of concolic testing via static analysis

Abstract

Software testing is crucial for developing any reliable application. Concolic testing is a White-Box testing technique, which tries to create various inputs for a program automatically, in order to locate runtime errors. This technique though suffers from a vast search space of execution paths that is inefficient and time-consuming to fully traverse. In our work, we propose a method to reduce this search space by exploiting the static information in a program and pruning branches which will certainly not produce errors. As static information, we use the program's abstract syntax tree, as well as the type annotations provided. This method was implemented in CutEr, a concolic testing tool for the Erlang programming language. We have also tested our method with custom programs and real world code. Finally, we report the results of our testing and show that the time performance of concolic testing can be immensely increased by using our method for safe branch pruning.Software testing is crucial for developing any reliable application. Concolic testing is a White-Box testing technique, which tries to create various inputs for a program automatically, in order to locate runtime errors. This technique though suffers from a vast search space of execution paths that is inefficient and time-consuming to fully traverse. In our work, we propose a method to reduce this search space by exploiting the static information in a program and pruning branches which will certainly not produce errors. As static information, we use the program's abstract syntax tree, as well as the type annotations provided. This method was implemented in CutEr, a concolic testing tool for the Erlang programming language. We have also tested our method with custom programs and real world code. Finally, we report the results of our testing and show that the time performance of concolic testing can be immensely increased by using our method for safe branch pruning.