Lecture 014

Scientific Debugging

Hypothesis: propose a possible defect explains the failure (Ockham's Razor: pick the simplest / closest to current work) based on the following

• Code

• Failing Run

• Other Runs

• Problem Description

Prediction: What data could confirm or refute hypothesis

• Experiment: collect data (running tests) to see if hypothesis is true

• Observation: interpret experiment to support / refute hypothesis

Develop a fix, and run experiment again to see if reproduced defect is fixed.

Brute Force

Other approach:

• Add compilation flags

• Try other optimization levels

• valgrind