[Mlir-commits] [mlir] fix a concurrent destruction issue for FallbackTypeIDResolver::registerImplicitTypeID due to function static variable (PR #85471)

[Stella Laurenzo]

stellaraccident wrote:

> ccing @MaskRay : Fangrui, WDYT? I feel that we do not have the control over the program exit sequence inside google?

I'm willing to be overridden on this if the evidence leads there, but I've worked many bugs like this, and I think the fault is elsewhere. Imagine for example if the background thread was not querying the type id registry but was just getting the value of a string flag that had been destructed. Then the request would be to heap allocate that thing... And the list goes on.

https://github.com/llvm/llvm-project/pull/85471