Proposal for STT_GNU_IFUNC and R_*_IRELATIVE

[H.J. Lu]

On Mon, Jun 1, 2009 at 1:00 PM, Roland McGrath <roland@redhat.com> wrote:
>
> To be a full description, it should say something about the ABI of the
> entry-point that returns the function pointer.  Is it "the prevailing
> normal ABI" for a function of no arguments returning a pointer to function?
> i.e., with whatever call-clobbered registers et al that generically entails.
> Or is it a special-case ABI to be defined precisely in the
> processor-specific spec?  However much or little the specification of the
> ELF feature per se wants to say about this, it should be explicit about
> what it does or doesn't specify for it.
>
>

How about this:

The calling convention of the STT_GNU_IFUNC function, which takes no
arguments and returns a function pointer, should follow the
processor-specific ABI. All rules for caller-saved and callee-saved
registers apply.


-- 
H.J.