Re: [PATCH][C++] Fix PR29433, make C++ use a lot less time/memory

[Jason Merrill <jason at redhat dot com>]

Nathan Sidwell wrote:
> Jason Merrill wrote:
>
> > When we see the first T in "T T();" we have no idea that it's supposed 
> > to be a type-specifier, so we can't look it up specially.
>
> I don't understand, we must be looking it up specially, otherwise the 
> struct stat hack wouldn't work.

No.  We are looking it up normally, but the member function declaration 
has not been processed yet, so lookup finds the outer declaration.

Without a special rule, this would be well-formed.  I assume that the 
rule is there to allow for compilers to use some sort of delayed parsing 
of member declarations.

I guess that we were missing this diagnostic before by accident, because 
of the different DECL_NAMEs for class template instantiations.  If we 
want to avoid the memory usage for different DECL_NAMEs, I'd think you'd 
have to hack around a lot to avoid giving this (correct) diagnostic here.

My inclination would be to make the change, and downgrade the error to 
pedwarn.  There's no need for it to be a hard error.

Jason