Literate Tracing
summary
Argues that documentation built from annotated, concrete executions makes huge codebases understandable: literate traces walk real runs, show input→output, and attach visual state to source lines. Implements TReX that runs the program under GDB, extracts state into HTML/LaTeX, offers a single-stepper and Python visualization modules, and is faithful by construction. Applied to Linux kernel, Git, GCC and Minisat; it works but diagram generation lacks a good DSL and run-to-line fragility suggests CI integration or autofix for maintenance.
AI-written summary · about
chapters (13)
- 0:00Software complexity study
- 3:15Documentation for understanding
- 5:42Literate tracing concept
- 6:48First literate trace example
- 9:43Literate trace features
- 11:28Building literate traces
- 12:30Literate tracing vs programming
- 13:13Quick start to traces
- 16:47Custom visualizations
- 19:37Describing program state
- 22:03Documentation with code changes
- 22:43Value of literate tracing
- 24:29Audience questions and discussion
