How to debug Linux multi threaded code

on YouTube ↗01h 06m

" summary

Shows concrete debugging steps: compile with ThreadSanitizer to get a line‑accurate race (demo program returns 0 or 1 and TSan points to line 9 vs line 10), compare with slower binary tools like Helgrind/Valgrind, use GDB scheduler‑locking and thread‑named backtraces to find which thread holds a lock, install a lightweight L3 logger (≈1 ns single‑thread, ≈7 ns with atomic index) instead of printf, and create a LiveRecorder plus time travel debugging + thread‑fuzzing workflow that turns an hour‑rare heisenbug into a fast, reproducible failure.

AI-written summary · about

" chapters (16)

1. 0:00Debugging race conditions
2. 2:04What makes bugs hard to find
3. 4:14Race conditions are tricky
4. 7:05Debugging tools
5. 9:37Demo with ThreadSanitizer
6. 13:54Sanitizers vs Valgrind
7. 16:05Race conditions in action
8. 19:23False positives in tools
9. 24:01Data races vs race conditions
10. 26:55Debugging with GDB
11. 37:10GDB thread scheduling
12. 44:20ThreadSanitizer sync issues
13. 45:26Logging for race conditions
14. 51:42L3 logging avoids sync
15. 56:59Thread fuzzing
16. 1:01:37Q&A

" related talks

• 01h 00mInvestigating C++ Applications in Production on Linux and WindowsSasha Goldshtein · NDC Oslo 2017
• 42mWhen the OS gets in the wayMark Price · Strange Loop 2015
• 52mApplication-Level Debugging and Profiling: Gaps in the Tool EcosystemRosemary Francis · LinuxCon + CloudOpen Europe 2013
• 01h 29mMore GDB wizardry and 8 other essential Linux application debugging toolsGreg Law · ACCU 2019
• 01h 05mThe art and methodology of troubleshootingAlexey Diomin · JavaDay Kiev 2014
• 01h 01mDebugging and Profiling .NET Core Apps on LinuxSasha Goldshtein · NDC Oslo 2017
• 54mStrace - Swiss Army Knife to Trace, Analyze, RT DebugHarald König · Open Source Summit Europe 2023
• 37mKernel and userspace tracing with LTTng and friendsSuchakrapani Datt · FUDCon Pune 2015