Sirraide wrote: > Is it intentional that this case now also is disallowed? That seems like an accidental result of how this was implemented. I’m surprised we didn’t have `register void*` as a parameter anywhere in our tests. https://github.com/llvm/llvm-project/pull/115313