jeanPerier wrote: > What about the optional KIND argument? Thanks for pointing that out. It is taken into account via the "resultType" that was computed by semantic analysis. Still deserves a test, I added one. https://github.com/llvm/llvm-project/pull/89785