Introduction to Undodb
summary
A reversible debugger called UndoDB provides reverse execution via snapshots plus deterministic record-and-replay, works on unmodified Linux (not macOS or Windows), handles threads signals and shared memory, and integrates with GDB/totalview-style front ends. Backwards watchpoints and attach-to-process recording let you find where a cache slot was corrupted and why. Performance is roughly half‑speed on a 16MB gzip run versus native, far better than GDB process record which uses tens of gigabytes. A deployable flight recorder records at customer sites for postmortem replay.
AI-written summary · about
chapters (4)
related talks
01h 02mNext Generation Debugging
59mHow Debuggers Work
57mDTrace for Rust... and everything else too!
33mBeyond println!: Mastering Rust Debugging
01h 01mBack to Basics: Debugging in C++
59mAsync Stacks: Making Senders and Coroutines Debuggable
41mDemistifying Rust Debugging
16mAirport for DuckDB: Letting DuckDB take Apache Arrow Flights