An Empirical Assessment of the Use of Execution Traces in Software Maintenance

One of the most frequent problems faced by software maintainers is to find the location of the code related with specific features. This is a costly task because features tend to be scattered in the code. Considering this problem, several approaches for locating features using dynamic analysis have been developed. Nonetheless, the use of dynamic analysis is not fully incorporated in the daily practice of software maintainers. Empirical studies that reveal strengths and weaknesses of the use dynamic analysis in maintenance tasks can bring better understanding of this problem. This study contributes with an empirical assessment of the use of execution traces on software maintenance. We report on a controlled experiment performed with developers when they were performing actual maintenance activities on systems of different sizes. There are benefits related with a systematic use of execution traces by reducing the execution time and increasing the correction of the developer task during software maintenance. Also, a lower level of difficulty of the tasks was perceived by the participants using dynamic information. Another conclusion is that the approach seems to be useless in maintenance activities where the problem of feature scattering does not occur clearly. This study also reveals that there are still some challenges related to false negatives.