I kind of suspected that.

In reality, __using should have been part of the data type, just as __stdc, __fastcall, __pascal etc have existed as attribute fields to tell how parameters are passed and cleaned out by a function, or many compilers have had __far, __huge, __large, ... to specify size of pointers.

But it is quite common that smaller compilers gets their own extensions as more "experimental" features where the compiler don't get the full orthogonal implementation of the extension.

If the compiler can't bind __using(1) to a type, then you can probably not remove that warning without skipping your function pointer approach.