[Mlir-commits] [mlir] [MLIR][LLVM] Allow importing of nameless globals (PR #101918)

[Christian Ulmann]

Dinistro wrote:

> Thanks!
> 
> I wonder if we want to rely on the prefix to be special enough to avoid collision with symbols in the input program or if we want to check this as well during the import? Are there any guarantees the input program cannot contain a symbol called mlir.llvm.nameless_global...?

Name mangling should ensure this for sane input langages.

https://github.com/llvm/llvm-project/pull/101918