A technology for optimizing the process of maintaining software up-to-date

Abstract

In this paper we propose a solution for reducing the time needed to make changes in an application in order to support a new version of a software dependency (e.g., library, interpreter). When such an update is available, we do not know if it comes with some changes that can break the execution of the application. This issue is very serious in the case of interpreted languages, because errors appear at runtime. System administrators and software developers are directly affected by this problem. Usually the administrators do not know many details about the applications hosted on their infrastructure, except the necessary execution environment. Thus, when an update is available for a library packaged separately or for an interpreter, they do not know if the applications will run on the new version, being very hard for them to take the decision to do the update. The developers of the application must make an assessment and support the new version, but these tasks are time consuming. Our approach automates this assessment by analyzing the source code and verifying if and how the changes in the new version affect the application. By having such kind of information obtained automatically, it is easier for system administrators to take a decision regarding the update and it is faster for developers to find out which is the impact of the new version.