Effective test scrubbing
summary
A ten‑month project ingests ~20,000 daily system test logs (3–7% failing → hundreds of failures), tokenizes and removes noisy runtime fields, converts text to vectors with count vectorizer and n‑grams, then groups runs using clustering in scikit‑learn inside Jupyter notebook. The prototype clusters similar failures so one inspected example fixes whole groups, cutting manual triage dramatically and enabling more automation; safety nets include deterministic identical‑hash checks and human review before automatic actions.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:39Continuous Integration
- 3:56Analyzing Test Failures
- 4:22Automatic Crash Detection
- 9:53Test Logs
- 18:03Scikit-Learn
related talks
49mCrash and test failure analysis with AI... no wait, ehm, with regexes!
48mUsing Machine Learning to find Linux bugs
18mNavigating Through Test Failures | From Challenges to Solutions
55mThe art of living with broken things
01h 30mAdvanced Software Testing for Data Scientists
24mAutomated Regression Testing with Splinter and Jupyter
44mThe Future is Today: Leveraging AI in Software Testing
25mGotta catch'em all: recognizing sloppy work in crowdsourcing tasks