[llvm-branch-commits] [llvm] Attributor: Deduce noundef returns if poison-generating attributes hold (PR #231797)

[Nikita Popov via llvm-branch-commits]

https://github.com/nikic commented:

Wouldn't Attributor want to do this based on its own inference results, rather than the ValueTracking queries?

https://github.com/llvm/llvm-project/pull/231797