Fast Fixes and Faulty Drivers: An Empirical Analysis of Regression Bug Fixing Times in the Linux Kernel

Abstract

Regression bugs refer to situations in which something that worked previously no longer works currently. Such bugs have been pronounced in the Linux kernel. The paper focuses on regression bug tracking in the kernel by considering the time required to fix regression bugs. The dataset examined is based on the regzbot automation framework for tracking regressions in the Linux kernel. According to the results, (i) regression bug fixing times have been faster than previously reported; between 2021 and 2024, on average, it has taken less than a month to fix regression bugs. It is further evident that (ii) device drivers constitute the most prone subsystem for regression bugs, and also the fixing times vary across the kernel's subsystems. Although (iii) most commits fixing regression bugs have been reviewed, tested, or both, the kernel's code reviewing and manual testing practices do not explain the fixing times. Likewise, (iv) there is only a weak signal that code churn might contribute to explaining the fixing times statistically. Finally, (v) some subsystems exhibit strong effects for explaining the bug fixing times statistically, although overall statistical performance is modest but not atypical to the research domain. With these empirical results, the paper contributes to the efforts to better understand software regressions and their tracking in the Linux kernel.