[Bug libstdc++/15910] can't compile self defined void distance(std::vector<T>, std::vector<T>)

["giovannibajo at libero dot it" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From giovannibajo at libero dot it  2004-08-03 12:48 -------
Subject: Re:  can't compile self defined void distance(std::vector<T>, std::vector<T>)

gdr at integrable-solutions dot net wrote:

>> Is there *anything* in the standard saying that the code is actually
>> illegal?
>> This is a rejects-valid IMO.

>   Can you elaborate on why you believe this is a reject a valid?

Because I think that picking up std::distance by Koenig lookup and having it
*abort compilation* instead of being reject as an invalid overload is just an
unintended side effect of the implementation. If you want to push it, it might
be an unintended side-effect of the standard - fine, the standard is broken.
But I won't never buy that the code is rightfully ill-formed as-is. It breaks
every single kind of orthogonality in the language: for instance, if I define a
::distance() with 3 parameters, it will compile.

When and if we agree there is *at least* a QoI issue here, we can discuss the
implementation details.


>A class is not an iterator just because it happens to have
>a nested type named iterator_category.

Sure, I'm just making the QoI issue less worse. There are other possible
implementations that come to mind. I can enable_if using a ConceptCheck on the
iterators - still much better than just aborting on *any* class.

Or maybe, we could have a way to use enable_if using as condition "does
iterator_traits<T> have a nested difference_type?".

Giovanni Bajo




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15910