Back to search

University of British Columbia

Detecting and explaining regression failures

Abstract

dc:description

Regression failures occur when changes to software, such as adding features or fixing security vulnerabilities, unintentionally break existing functionality. Debugging regression failures requires significant time and effort, especially in large-scale software systems, where developers need to review an extensive body of code in order to assess how the changes affected code functionality. Numerous debugging-assistance approaches have been designed to help developers with this task. One category of approaches focuses on detecting latent regressions by formally proving/refuting their presence. Such formal approaches typically rely on symbolic execution – a technique that models program behavior with symbolic inputs. Another category of approaches focuses on explaining known regressions, i.e., explaining how changes led to the failure. These approaches typically aim at identifying relevant program statements developers need to inspect. The most prominent of these approaches are based on program slicing, which maintains the flow of information between the relevant statements. The scalability of debugging-assistance approaches often hinders their practical applicability. Specifically, symbolic execution struggles with large code bases, as well as complex programming constructs, such as loops and non-linear arithmetic. Slicing often produces results that are too large for manual inspection. This thesis contributes to improving the scalability of both detection and explanation approaches via abstraction, thus, reducing the analysis complexity without sacrificing accuracy. First, we propose an approach, named ARDiff, for improving the scalability of symbolic-execution-based techniques in detecting regressions in complex code. Next, to explain regressions, we introduce the concept of summarizing unchanged code between versions, implemented in a tool called InPreSS. To understand how developers can utilize the summarization concept and what their general preferences in debugging a regression failure are, we survey over 50 practitioners from eight different countries. Guided by the insights from this study, we develop a novel slicing approach, named CoReX, which provides code representations more aligned with developers’ needs. To promote the adoption of slicing approaches, we introduce Slicer4D, an IntelliJ IDEA plugin that integrates slicing into the development environment, helping developers focus on relevant code.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Electrical and Computer Engineering
Grantor dc:publisher
University of British Columbia
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Badihi, Sahar
Contributors dc:contributor
  • Rubin, Julia

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/90519

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Badihi, Sahar. Detecting and explaining regression failures. doctoral thesis, University of British Columbia, 2025. http://hdl.handle.net/2429/90519