Tracking down Exceptions in Standard ML Programs

We describe our experiences with an exception anal-ysis tool for Standard ML. Information about excep-tions gathered by the analysis is visualized using pam, a program visualization tool for emacs. We study the results of the analysis of three well-known programs, classifying exceptions as assertion failures, error excep-tions, control- ow exceptions, and pervasive exceptions. Even though the analysis is often conservative and re-ports many spurious exceptions, we have found it useful for checking the consistency of error and control- ow exceptions. Furthermore, using our tools, we have un-covered two minor exception-related bugs in the three programs we scrutinized.