Description
We have no way of composing a TokenRange of type tokens with a name. For example, if the type is struct foo *, then composing this with a name would give us struct foo *name, however, composing the type void (*)(void) with a name would give us void (*name)(void). I think we would "engineer" for the type tokens to determine the name offset via the pretty printer. It would require quite the hack, but the idea would be to synthesize a fake clang::ParmVarDecl, set its type to the relevant clang::QualType, set its name/identifier to a synthesized name, and then adopt this and actually rely on the underlying decl printer. Then, when we're post-processing the pasta::PrintedTokenRange, we'd look for the special name token, and then use keep track of its offset in rpc::Type. This would then let us compose type tokens with other tokens to form semantically correct typed names, e.g. for rendering in the UI.