Nowadays, Worldline, a major IT company, develops applications that are dealing with an increasing amount of critical data with a source code is more and more complex. To the stay in the race against its competitors, Worldline has to increase the quality of its projects. One transversal team of the company has for main mission to enhance project quality. By joining this team, we performed an audit of several projects of the company to identify how to avoid project failure. Relying on project metadata analysis, interviews, survey, and literature study, this audit drew no final conclusion. However, it highlighted a need to enhance testing usage. To test every piece of code is compulsory to avoid project failure. In industry and more specifically in Worldline, automation of tests has begun to ensure the proper behavior of their applications. But, ensuring the execution of all tests after a change can be a costly operation requiring several hours. However, in a daily development process, developers cannot run all the tests after a change to check the impact of their modifications. Feedback on the changed code is delayed and developers have to spent time to discover the potential bugs. A late feedback can potentially lead to the failure of a project. The solution generally proposed in literature consists in reducing the number of tests to run by selecting only the ones related to the last changes made by the developer. The approach selects a suitable and small set of tests to detect potential regression in the application behavior. Test selection approaches have their own advantages and drawbacks. These approaches need to be studied in the context of Worldline and consequently adapted to suit developers’ habits. Before integrating such an adapted test selection approach, a study of the testing behavior of the developers has been done to get their current test selection usage. This study will monitor all the tests launched by the developers during their everyday development tasks. Based on the study of the Worldline environment, we developed a tool for the developers aiming to perform test selection. It is adapted to the developers’ habits and is directly usable by them. The goal is also to study the impact of this tool on their behavior. We hope that the test selection approach that is packaged in a tool adapted to the Worldline developers will bring a change in their development process.
- Thesis supervisor(s) : Nicolas Anquetil (Associate Professor - Université Lille 1),Anne Etien (Associate Professor - Université Lille 1) - Referees : Anthony Clève (Professor - Université de Namur), Andy Zaidman (Associate Professor - Delft University of Technology) - Examiners : Serge Demeyer (Professor - University of Antwerp), Clarisse Dhaenens (Professor – Université Lille 1),Henda Hajjami Ben Ghezala (Professor - ENSI) - Guests, Sylvain Cresson (Worldline), Pascal Croisy (Worldline)
Thesis of the team RMoD defended on 08/11/2017