[Issue 22021] pragma(mangle) not accepted in function body

d-bugmail at puremagic.com d-bugmail at puremagic.com
Mon Jun 14 19:58:28 UTC 2021


https://issues.dlang.org/show_bug.cgi?id=22021

--- Comment #2 from Max Samukha <maxsamukha at gmail.com> ---
(In reply to kinke from comment #1)
> I'm not sure this should work - `extern(<lang>)` doesn't affect the mangling
> of nested functions either, only ABI/calling convention. The reason is
> simple, nested functions are D only, and the D mangling makes sure there are
> no duplicates/conflicts (as it includes the parent function(s)).
> 

I needed that when I was trying to implement a poor man's decorator in the
least ugly way:

mixin template ISR(int number, alias impl) {
        static struct Impl {
            extern(C) static
            pragma(mangle, "__vector_" ~ number.stringof)
            @signal @used /* @externally_visible */ void isr() {
                impl();
            }
        }
    }

    /// ditto
    mixin template ISR() {
        enum dummy = 0;
        alias impl = __traits(parent, dummy);
        enum number = mixin(__traits(identifier, impl), "_vect");
        mixin ISR!(number, impl);
    }

@always_inline void USART_UDRE() {
    mixin ISR; // magic
    ...
}


> The IMO more intuitive workaround is a freestanding function. Exposing a
> nested D function for consumption from C[++] is probably not very common; I
> guess the main usage would be consuming an external function from D, and
> declaring it as nested function in D for locality.

Nested functions are conceptually very similar to member functions. It is a
matter of good taste to try and make them consistent.

--


More information about the Digitalmars-d-bugs mailing list