O impacto do uso de rastros de execução em atividades de localização de características de software: um experimento controlado

Abstract

One of the most frequent problems faced by software maintainers is to find the location of the code related with specific features, which are important for understanding the software requirements. To find where a certain feature is located is a costly task because, in general, it tends to be scattered or tangled into the code. Considering this problem, an approach for locating features using dynamic analysis has been developed, which presents views generated from execution traces. The objective of this work is to help understanding the features, enabling faster location of features and providing higher accuracy rate in maintenance activities, achieved the information that leads directly to the understanding. To evaluate this approach and check whether the goal was achieved, a controlled study was performed with the participation of human beings, when they were performing actual maintenance activities on systems of different sizes. This study contributes with an appraisal about the impact of using information from execution traces on software maintenance. This study shows the benefits related with a systematic use of information from execution traces by reducing the execution time and increasing the rate of correction in searches for information during software maintenance. The views were useful for locating the features of interest and reducing the initial search space, providing better precision in the search for information. This feature resulted in a lower level of difficulty perceived by study participants during the proposed maintenance activities. Besides that, this study has indicated that the size and quality of the search spaces (presented in the views) can affect the rate of correction and execution time on maintenance activities using the presented approach. However, this study also reveals that there are some challenges when it comes to implement this approach on a large scale, system since the views may reveal many false positives and false negatives. This situations can impact the understanding of the systems based on the provided information. On the large-scale use, of the approach, it would be interesting to introduce more robust views, enabling more robust analysis. Finally, another conclusion is that this approach is not useful in comprehension activities where the problem of scattering and entanglement does not occur.