Re: c++/8897: Demangling of template conversion operators

[Carlo Wood <carlo at alinoe dot com>]

On Mon, Feb 24, 2003 at 11:45:07PM -0600, Benjamin Kosnik wrote:
> Did you get a chance to look at:
> 
> // c++/8897
> void test05()
> {
>   // cplus-dem FAIL  
>   // icc FAIL
>   // new __cxa_demangle FAIL
>   test("_ZN1AIfEcvT_IiEEv", "A<float>::operator float<int>(void)");
>   // XXX better "A<float>::operator int()"
>   // XXX better "A<float>::operator T()[T=int]"
> }
> 
> I'm not quite sure what's up with this one.

It's

template<typename T1>
  struct A {
    template<typename T2>
      operator T2();
  };

A<float> a;
(int)a;		<-- that function.

The demangling is obvious a bug.  A first "fix" would
be to print:

A<float>::operator int<int>()

compare

A<float>::templatememberfunction<int>()

or say, A<float>::operator +=<double>(double), being
  template<typename T>
    operator+=(T);

But, when the cast type is a class, that could be
confusing:

A<float>::operator B<B>()

Note 'B' cannot be a template, and thus this is NOT ambigious.
If B was a template, you'd get:

A<float>::operator B<int><B<int> >()

for example.

But, the question is: how do you WANT it to look?

I don't like to use "operator T() [with T = int]"
because it would be the only instance that the demangler
would use that format.  I think I do prefer the
"operator int<int>()".

-- 
Carlo Wood <carlo at alinoe dot com>