Abstract

When a software system crashes, users report the crash using crash report tracking tools. A crash report (CR) is then routed to software developers for review to fix the problem. A CR contain a wealth of information that allow developers to diagnose the root causes of problems and provides fixes. This is particularly important at Ericsson, one of the world’s largest Telecom company, in which this study was conducted. The handling of CRs at Ericsson goes through multiple lines of supports until a solution is provided. To make this possible, Ericsson software engineers and network operators rely on runtime data that is collected during the crash. This data is organized into files that are attached to the CRs. However, not all CRs contain this data in the first place. Software engineers and network operators often have to request additional files after the CR is created and sent to different Ericsson support lines, a problem that often delays the resolution process.
In this thesis, we conduct an empirical study of the runtime files attached to Ericsson CRs. We focus on answering four research questions that revolved around the proportion of runtime files in a selected set of CRs, the relationship between the severity of CRs and the type of files they contain, the impact of different file types on the time to fix the CR, and the possibility to predict whether a CR should have runtime data attached to it at the CR submission time. Our ultimate goal is to understand how runtime data is used during the CR handling process at Ericsson and what recommendations we can make to improve this process.