Re: PATCH: Add ifunc attribute

[Jakub Jelinek <jakub at redhat dot com>]

On Sat, Jun 13, 2009 at 11:07:38AM -0700, H.J. Lu wrote:
> I don't think there are anything wrong with ifunc testcases in glibc.
> 
> void *foo (void) __attribute__((__ifunc__))
> 
> isn't a good idea since this isn't the real prototype for foo. It will lead
> to many problems down the road. foo is usually declared in a header
> file and you can't include such a header file in the implementations
> of foo if 2 prototypes don't match.

Yeah, I agree.  Having to tweak headers through ugly hacks to avoid the real
prototype to be visible in order to define ifunc function would be a very
bad idea.

	Jakub