Identifying Causes and Fixes of Linux Variability Anomalies

In order to prevent variability anomalies from occurring in the first place, we need to understand what causes them. In order to provide automated solutions for such anomalies, we need to understand how developers usually fix them. This project mines commit information from Linux’s git repository in order to identify causes and fixes of variability anomalies. Our results show that variability anomalies are often introduced through incomplete patches that change Kconfig definitions without properly propagating these changes to the rest of the system. Anomalies are then commonly fixed through changes to the code rather than to Kconfig files. For more details, read our MSR’13 paper.

To come up with some patterns for how anomalies get introduced and fixed, we analyze a set of patches that have been submitted by Tartler et al. in their EuroSys 2011 paper. By using undertaker, they detected a set of variability anomalies, and then submitted patches to fix some of the detected dead/undead code blocks. They tracked the responses of developers to these patches.

The following is the dataset of all these email exchanges. To open it, you can install the mail tool on Linux which is part of the mailutils package. After extracting the archive, use the following command:

mail -f linux-mbox/

You can also use mutt which has a slightly better user interface.

mutt -f linux-mbox/

The dataset of the detected defects on which these patches are based can be found in this SQLite database. You can browse this database visually using SQLite Browser. The defects table contains the detected defects/anomalies. The tickets table contains the submitted patches, their resolution, and our classification of what the patch was doing (patch_type).

The following are examples of how you can query the database:

Maximum Number of Defects Solved by Patch:

select max(defectCountPerTicket) from
(select ticket, count(filename) as defectCountPerTicket from defects, ticket
where defects.ticket = ticket.id
group by ticket)

Number of referential defects for which patches are submitted:

SELECT count(ticket) from defects where ticket IS NOT NULL and defects.filename LIKE '%missing%'

Number of tickets solving detected defects:

select COUNT(*) from ticket where id IN 
(SELECT ticket from defects where ticket IS NOT NULL)

The code used to analyze the evolution of variability anomalies in this project can be found in our GitHub repository.