Bug 15843 – D-type mangling used for extern(C) (extern) function declaration inside function body, on LDC, GDC, and DMD.

Code: void main() { extern(C) extern int func(int); static assert(func.mangleof == "_D1x4mainFZ4funcUiZi"); } It is mangled as such on all 3 major compilers. Using "pragma(mangle, "func")" as an attribute for func gives: x.d(4): Error: unrecognized pragma(mangle) x.d(6): Error: undefined identifier 'func' x.d(7): Error: undefined identifier 'func' Seemingly, the only user-side fix is moving the declaration out of the function body.

I would err on the side of this is working as intended. Nested functions are not compatible with C, and so should not have a C name (by declaring the function inside main, it implicitly gets a "this" context parameter)

Perhaps we should error out on `extern` and `extern(XXX)` functions then?

Static nested functions with C calling convention are useful as system callbacks. They just don't need to be visible to other linked modules.

extern(C) specifies not only the mangling, but also the calling convention. So having extern(C) inner functions do serve a purpose. I think the mangling is really implementation defined, since C doesn't have any notion of inner functions. Note that `extern` on its own isn't doing anything here. If anything, I'd say pragma(mangle) should work in this context, that would at least give a more straightforward path (really issue 22682). Note that defining the prototype without `static` indeed suggests it should have a `this` pointer, but it doesn't work even if you add `static`.

Prgma mangle works locally since a year or so: ``` void main() { pragma(mangle, "func") extern(C) int func(int); static assert(func.mangleof == "func"); } ```