[cfe-dev] Should we build semantically invalid nodes?

[Argiris Kirtzidis]

Argiris Kirtzidis wrote:
> To get more specific about Sebastian's patch, I tried it out and it 
> works quite well, Clang is pretty robust:
>
> typedef int fn(int,int);
> int f() {
>  int a;
>  reinterpret_cast<fn>(a)(1,2,3);
>  int (*fx)(int,int) = reinterpret_cast<fn>(a);
> }

I mean, Clang builds a nice AST tree of everything for the above snippet.