Software Reflexion Models
Many software design models,over time, becoming increasingly inconsistent with the system's source code. The reflexion model approach helps an engineer use a high-level model of a software system to show where the high-level model agrees with and differs from the source code. Our results (and Gail Murphy's dissertation!) include a use of reflexion models to aid in the understanding and experimental reengineering of the Microsoft Excel spreadsheet product.
Dependent Tests
Almost all results in software testing, and almost all software testing tools, assume that each test in a test suite will produce the same result regardless of the order in which the tests are executed. We have shown that this "test independence" assumption does not hold by identifying test suites in publicly available source where executing the tests in "isolation" (i.e., in a clean virtual machine) produces a different results from executing them one after another.
Detecting Behavioral Anomalies
Developers change a program with an explicit and an implicit intention: to realize a particular objective and to avoid affecting other program properties. Testing and analysis help developers satisfy both intentions, but subtle bugs are still common. We earlier showed how a simple comparison of the static and dynamic call graphs from two related program versions can identify anomalies that detect some of these subtle bugs.
Speculative Analysis
Software developers primarily rely on experience and intuition to make development decisions. We believe that developers can make even better decisions if they are informed of the consequences of their choices. Speculative analysis is a family of techniques that attempt likely developer actions in the background and deliver precise information to the developers about their choices right as they are making the relevant decisions.