Automatizando a formação de oráculos de testes de regressão constituídos por dados reais

Abstract

Over the decades, software malfunctions have caused incalculable losses, be they of a financial nature or not. Therefore, the search for practices that allow a possible decrease in the number of these defects has made the test process gain an important highlight. However, the implementation of this process is still considered costly, as it often corresponds to more than 50% of the monetary resources provided in the budget of a project. In order to reduce this cost, part of the scientific community has been striving to automate the activities contained in the testing process. Among the activities to be automated is the generation of test cases. One of the challenges for its automation is precisely to be able to automate one of its main sub-activities: the construction of test oracles. Among the existing barriers to this goal is the difficulty of automating the definition of a correct/expected output among other possibly incorrect results. This barrier is known as Oracle Problem. This work proposes an approach for the automation of the generation of oracles aimed at conducting regression tests, which circumvents the barrier known as Oracle Problem. In this approach, through an aspect-based architecture, the oracles generated are made up of real data, obtained directly from actions performed by end users. Also in this work, the Heimdall tool was developed, which implements the proposed approach.