On the diffuseness of code technical debt in Java projects of the apache ecosystem

Background. Companies commonly invest majorBackground. Companies commonly invest major effort into removing, respectively not introducing, technical debt issues detected by static analysis tools such as SonarQube, Cast, or Coverity. These tools classify technical debt issues into categories according to severity, and developers commonly pay attention to not introducing issues with a high level of severity that could generate bugs or make software maintenance more difficult. Objective. In this work, we aim to understand the diffuseness of Technical Debt (TD) issues and the speed with which developers remove them from the code if they introduced such an issue. The goal is to understand which type of TD is more diffused and how much attention is paid by the developers, as well as to investigate whether TD issues with a higher level of severity are resolved faster than those with a lower level of severity. We conducted a case study across 78K commits of 33 Java projects from the Apache Software Foundation Ecosystem to investigate the distribution of 1.4M TD items. Results. TD items introduced into the code are mostly related to code smells (issues that can increase the maintenance effort). Moreover, developers commonly remove the most severe issues faster than less severe ones. However, the time needed to resolve issues increases when the level of severity increases (minor issues are removed faster that blocker ones). Conclusion. One possible answer to the unexpected issue of resolution time might be that severity is not correctly defined by the tools. Another possible answer is that the rules at an intermediate severity level could be the ones that technically require more time to be removed. The classification of TD items, including their severity and type, require thorough investigation from a research point of view.effort into removing, respectively not introducing, technical debtissues detected by static analysis tools such as SonarQube, Cast, or Coverity. These tools classify technical debt issues intocategories according to severity, and developers commonly payattention to not introducing issues with a high level of severitythat could generate bugs or make software maintenance moredifficult. Objective. In this work, we aim to understand the diffuseness ofTechnical Debt (TD) issues and the speed with which developersremove them from the code if they introduced such an issue. The goal is to understand which type of TD is more diffusedand how much attention is paid by the developers, as well asto investigate whether TD issues with a higher level of severityare resolved faster than those with a lower level of severity. Weconducted a case study across 78K commits of 33 Java projectsfrom the Apache Software Foundation Ecosystem to investigatethe distribution of 1.4M TD items. Results. TD items introduced into the code are mostly relatedto code smells (issues that can increase the maintenance effort). Moreover, developers commonly remove the most severe issuesfaster than less severe ones. However, the time needed to resolveissues increases when the level of severity increases (minor issuesare removed faster that blocker ones). Conclusion. One possible answer to the unexpected issue ofresolution time might be that severity is not correctly definedby the tools. Another possible answer is that the rules at anintermediate severity level could be the ones that technicallyrequire more time to be removed. The classification of TD items, including their severity and type, require thorough investigationfrom a research point of view.