"Linux Testing made better with DATA"
summary
Builds a pair of databases that link functions (from ctags) to per-test code coverage (collected and parsed into a test DB), then parses a patch to select only tests that touch modified files. Automation runs in Jenkins and yields dynamic suites: claimable reduction up to 99% and initial-environment results showing 93.5% reduction (example: ~1,000 tests reduced to 129 while retaining fault-detection), with kselftest and avocado already covering many files at 100%.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:27Agenda
- 5:20Overview
- 5:37Source Database Creation
- 9:30How To Know What Test Cases To Be Improved
- 11:42Disclaimer Chart
- 13:36Is There any Way We Can Measure Path Coverage Rather than Line Coverage
related talks
15mLinux Testing made better with DATA
50mTips for Writing Good Tests for Linux
19mChecking your work: Linux kernel testing and CI
45mUsing AI, statistics, and set theory to know what to test
30mIntegrating testing with maintainer flows
26mKeeping Your Tests Lean
35mSmarten Up Your Pipeline and Reduce Test Execution Time｜
26mA Unit Test Would Have Caught This...