The question is in the title. I am struggling with a pretty nasty race condition (that I cannot locate) involving concurrent jump computations (having a copy of the model for each tasks).
There’s previous threads here on Discourse regarding sanitizers like asan and tsan, and open issues on Github. There’s also this devdoc page:
The situation is, as far as I understand, such that the developers sometimes use some of the sanitizers to help themselves find bugs in Julia (bugs in the C/C++ code, I guess), but there’s no user-friendly solution yet.
Ideally Julia would plug into LLVM’s sanitizer support, but that still seems to be a long way off.