Finding Important Stack Frames in Large Systems

TL;DR

The paper addresses the challenge of efficiently triaging vast numbers of stack traces by identifying potentially informative frames within each trace. It implements a pre-highlighting feature in JetBrains' internal bug-processing tool, displaying candidate frames in bold with a red exclamation icon to guide developers without disrupting their manual selection. The method relies on inverse document frequency (IDF) calculated over the full corpus to pick the top-3 rare frames as likely sources of specific issues, with the option to substitute or extend with dynamic analysis or AI models in future work. A simple survey with 18 developers confirmed moderate usefulness (mean 3.6/5) and good visualization (mean 4.0/5), and feedback highlighted the need for tooltips and potential on-the-fly learning or AI-assisted ranking (Du2023ResolvingCB). The work demonstrates production deployment, user-centered evaluation, and points toward more sophisticated modeling for stand-alone recommendations.

Abstract

In this work, we developed, integrated, and tested a feature that automatically highlights potentially important frames in stack traces. The feature was implemented in the internal bug-processing tool at JetBrains that processes tens of millions of stack traces. We surveyed 18 developers at JetBrains who provided valuable feedback on the idea and the implementation.

Figures (1)

• Figure 1: The course of work (a-b) in the pre-existing system without the feature, (c-d) with the proposed feature. The top part shows the initial UI of viewing a stack trace, the bottom --- after manually selecting important frames.