[cfe-dev] naked attribute

[Rafael Espíndola]

> If you use Antons approach, it would not be an option for the backends for
> particular targets to ignore the naked attribute.

I am with Anton on this. I don't think we should have a user visible
switch for enabling/disabling the support in a particular
architecture. I think the complete decision tree is

if (llvm doesn't support the attribute on this architecture for some reason)
   warn about the ignored attribute and actually ignore it (i.e.,
don't put it on the IL).
else {
   put the attribute on the IL and produce unreachable instead of ret
as appropriate.
   if (gcc doesn't support this attribute in this arch)
     produce an externsion warning.
}

The only point there was disagreement was on the extension warning. In
any case, I am ok with compromising on that if you are ok with
"Anton's approach".

Cheers,
Rafael