Re: [C++] fix alias and members

[Nathan Sidwell <nathan at codesourcery dot com>]

On 06/08/10 14:46, Jason Merrill wrote:
> On 06/08/2010 05:59 AM, Nathan Sidwell wrote:
> > + alias = lookup_attribute ("alias", DECL_ATTRIBUTES (decl)) != 0;
> > +
> > + if (alias && TREE_CODE (decl) == FUNCTION_DECL)
> > + record_key_method_defined (decl);
>
> This seems wrong for weak aliases.

I'm not sure.  aliases have to be to another symbol defined in the same 
translation unit.  So, for the target to be in multiple TU's it'd have to be 
weak/comdat, which implies either an out-of-line inline function definition or a 
template instantiation.  Is that something that makes sense to support?

My understanding of the use-case for weak aliases was to a non-weak unique 
symbol.  So it seemed ok to me for the aliased symbol to be a key method.

nathan

--
Nathan Sidwell    ::   http://www.codesourcery.com   ::         CodeSourcery