Bug 4581 – extern function pointers inside functions

import std.c.windows.windows; The following code compiles, but does not link: void main(string[] args) { extern (Windows) HINSTANCE ShellExecuteW(HWND, LPCWSTR, LPCWSTR, LPCWSTR, LPCWSTR, INT); HINSTANCE i = ShellExecuteW(null, "open", "http://www.example.com", null, null, SW_SHOW); } Error 42: Symbol Undefined __D4test4mainFAAyaZi13ShellExecuteWMWT4core3sys7windows7windows6HANDLEPxuPxuPxuPxuiZT4core3sys7windows7windows6HANDLE@24 As we can see, the symbol is D mangled, which it should not be.

It's a feature, not a bug. If the declaration is contained in something nested (like a struct, I guess it also applies to functions), name extern doesn't disable mangling. You can ask Walter for the why. However I didn't find this in the spec and I could be wrong, so I'll leave it to others to find this and mark the bug as invalid.

I cannot foresee any instance in which it is wanted that the compiler disregard the extern declaration, and even if this is the case, there should at least be a warning as it blithely ignores the programmer's intention.

You can't express intention with extern attribute, because it's ambiguous: whether it specifies an external function or internal function with specific calling convention. Currently there's no way to differentiate between these two cases.

(In reply to comment #3) > You can't express intention with extern attribute, because it's ambiguous: > whether it specifies an external function or internal function with specific > calling convention. Currently there's no way to differentiate between these two > cases. But is there really a difference? Currently, extern( foo ) means 'use foo's calling conventions and name mangling'. As long as the latter is also part of the definition, I do not see such a difference.

Walter can we get your opinion on this?

fixed by https://github.com/dlang/dmd/pull/15582. The problem was actually not the extern attribute but the mangle.