Re: [C++ Patch] 28705 & koenig lookup

[Mark Mitchell <mark at codesourcery dot com>]

Nathan Sidwell wrote:

> This patch fixes bug 28705, but it impacts koenig lookup and DR 218.  
> The actual ICE reported in 28705 is somewhat incidental.

> This patch changes G++ to ignore non-function non-variable entities 
> during koenig lookup.  This makes G++ accept strictly more programs than 
> it did before.  I don't think it causes G++ to reject a program it 
> previously accepted.

That's a good thing.

However, we might still be changing G++ to accept a program that the 
committee might later deem invalid, meaning that we'll have introduced 
an extension, just to remove it again.  The way that might happen is 
that the committee might decide to include types in the overload set, 
thereby making programs that we would deem unambiguous with your patch 
ambiguous.

We ICE now if we end up finding a type.  If we fix it so that types 
actually work, then the committee might decide that types don't count, 
and we've again added an extension.

But, what if we just issue an error if the only thing we find is a type? 
 Wouldn't that be even more conservative?  That program ICEs now, so 
that's no worse.  If finding types becomes OK, then we can add that 
behavior.

Thanks,

--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713