Hi all, Wanted to share something with you that we’ve built over the last few weeks and have been using with great success in our environments. We use a combination of memory sanitizers here at Backtrace, including valgrind.
The issues we faced were: 1. It’s quite difficult to aggregate the errors and feed them into our engineering workflows 2. It’s quite easy to miss the fact that you might have introduced a new non-benign leak or memory error. 3. No automation around creation of tickets. 4. No deduplication of call sites across reports. Details for all of this are up in https://engineering.backtrace.io/posts/sanitizers/ - essentially, all you have to do is HTTPS POST / curl your valgrind logs into a server and it will aggregate the data for you in a more actionable way. You can also drill down into the raw logs of an individual error. I set up a seeded environment with an example from our test corpus (mix of real world applications and synthetic) up at https://sanitizers.sp.backtrace.io/p/valgrind/triage (username: sanitizers / password: appleorangewat1). If you click on the “Saved Views” button (left-most icon on the top right) then “All”, you can see some example views we found particularly useful here, especially when triaging leaks. Currently, I’ve only added support for XML protocol version 4 and memcheck, but if folks find this useful for other tools, please let me know and I’m happy to add support! Lastly, feedback is greatly appreciated. Hope others find this useful as well. Regards. -- Samy Al Bahra CTO / Co-Founder Backtrace [https://backtrace.io]
_______________________________________________ Valgrind-users mailing list Valgrind-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/valgrind-users
