Issue 10249 - incorrect mangling for overloaded symbol

[9rnsr]

http://d.puremagic.com/issues/show_bug.cgi?id=10249

[9rnsr]

Don't merge this yet.

[9rnsr]

Ok. Ready to merge.

[andralex]

A brief documentation? At least for the new code :o).

[9rnsr]

I'll write.

[9rnsr]

OK. Updated code.

[andralex]

lgtm, anyone up for a semantic review?

[WalterBright]

What is the difference between mangle() and mangleExact()? I don't understand what is happening in the code, and how 10249 is fixed. Can you please provide some comments in the code?

[9rnsr]

@WalterBright
Currently mangle() returns the exact mangled name of the root function in overloads.

module test;
void foo(long) {}       // _D4test3fooFlZv
void foo(string) {}     // _D4test3fooFAyaZv
pragma(msg, foo.mangleof);
// Currently prints "_D4test3fooFlZv", which is calculated by
// `FuncDeclaration::mangle()` of void foo(long)

But this is not good. foo is overloaded name, so its mangled name should be neither _D4test3fooFlZv nor _D4test3fooFAyaZv. That's the core issue in 10249

After the change, mangle() returns the exact mangled name by calling mangleExact() only when actually the function has no overloads.

pragma(msg, foo.mangleof);
// After this change, prints "4test3foo" by calling `Dsymbol::mangle()`,
// instead of `FuncDeclaration::mangleExact()`.
pragma(msg, __traits(getOverloads, test, "foo")[0].mangleof);  // _D4test3fooFlZv
pragma(msg, __traits(getOverloads, test, "foo")[1].mangleof);  // _D4test3fooFAyaZv

If a function has no overloads, .mangleof property still returns exact mangled name.

void bar() {}
pragma(msg, bar.mangleof);
// still prints _D4test3barFZv by calling `FuncDeclaration::mangleExact()` from `FuncDeclaration::mangle()`.

[WalterBright]

Thanks for the explanation! Can you please add these as comments in the code?

[9rnsr]

@WalterBright OK. Updated code comment on FuncDeclaration::mangle()