Sources of Inter-package Conflicts in Debian

TL;DR

This paper analyzes inter-package conflicts in Debian, showing that more detailed meta-data could prevent many conflicts and proposing ideas for future detection methods.

Contribution

It presents a case study on Debian's inter-package conflicts and suggests that enhanced meta-data can significantly reduce undetected conflicts.

Findings

At least one third of conflicts could be prevented with better meta-data.

Another one third of conflicts can be detected through targeted testing.

Current meta-data is insufficient for detecting all common conflict types.

Abstract

Inter-package conflicts require the presence of two or more packages in a particular configuration, and thus tend to be harder to detect and localize than conventional (intra-package) defects. Hundreds of such inter-package conflicts go undetected by the normal testing and distribution process until they are later reported by a user. The reason for this is that current meta-data is not fine-grained and accurate enough to cover all common types of conflicts. A case study of inter-package conflicts in Debian has shown that with more detailed package meta-data, at least one third of all package conflicts could be prevented relatively easily, while another one third could be found by targeted testing of packages that share common resources or characteristics. This paper reports the case study and proposes ideas to detect inter-package conflicts in the future.