[Mlir-commits] [mlir] [mlir][arith] Fix SelectOp unsafe int range inference with unfinalized range case (PR #173716)

[Jakub Kuderski]

kuhar wrote:

> > I don't remember the int range analysis details from the top of my head but can we just return uninitialized state in this case instead of falling back to conservative full range immediately?
> 
> I was wondering this as well, not sure should we return uninitialized range or max range. But in this scenario, I think the max range effectively behaves the same as an uninitialized range.

@Hardcode84 @krzysz00 do we have this documented anywhere?

https://github.com/llvm/llvm-project/pull/173716