patch for PR 2632

[Richard Henderson]

On Tue, Jul 16, 2002 at 07:29:06PM -0700, Ulrich Drepper wrote:
> What I actually meant was whether the return value of
> __builtin_frame_address(0) on Arm is useful for anything.  Note Philip's
> patch where he had to post-process the value before it could be used. 
> If the immediate return value isn't useful, why not change gcc to
> perform the operation which is now done in the glibc code?

Because then we'd have to have port specific "useful values"
of something that is by nature not a useful value to anyone
but the compiler itself.

Even on x86, the meaning of __builtin_frame_address(0) changes
with and without -fomit-frame-pointer.


r~