[llvm] LangRef: storing poison in memory is UB (PR #141339)

[Nuno Lopes via llvm-commits]

nunoplopes wrote:

I'm totally against this. Widening in IR is a problem, yes, but this is not the way to go. Load widening can be solved by using vector loads or struct loads (if the new size is not a multiple of the old one).

There are implications about making stores UB, such as making movement of stores and function calls even harder.

https://github.com/llvm/llvm-project/pull/141339