Investigando estratégias de teste de regressão: um estudo de caso na STI-UFRN

Abstract

The Superintendência de Tecnologia da Informação - STI is responsible for developing and supporting UFRN’s computational systems. Each system has a testing team responsible for system testing. One of these systems is the Sistema Integrado de Gestão de Atividades Acadêmicas - SIGAA, it has more than 5 thousand use cases. The test team performs exploratory and regression tests manually, however, for a system of this size, performing manual regression tests is unfeasible, thus the regression tests end up being done in a negligible amount, but the task to increase the amount of regression tests is not easy, especially if the are performed manually, given that it will will take away the ability to do many other exploratory tests. Due to the high number of use cases, it is impractical to automate all regression tests. Therefore, the aim is to perform automated regression tests for features that are most likely to fail and then find indicators that will show the most relevant use cases at which the automated tests are justified, thus allowing the testing team to focus on exploratory tests, and, generally speaking, expand test coverage without having to increase the number of team members. Thus, this research sought strategies for selecting features for building automated system tests through a case study at STI-UFRN. This study identifies features from use cases of SIGAA which must have automated tests prioritized, in order to do this it was performed two studies which sought to measure the influence of criterias such as: different screen complexity and time from the last change, features that demanded more time of manual tests during the period of one month. Thus, the tests of the chosen features were automated with UI.Vision, considering the equivalence classes and happy path in their construction. Based on a daily execution of these tests, it was possible to note that the criterias which were more relevant in order to determine which tests would be automatized were: screens that are dependent on external services and features which required a large amount of manual tests recently.