Adding SH support to glibc-2.2 (2)

[kaz Kojima]

NIIBE Yutaka <gniibe@chroot.org> writes:
>>>  - you don't want to unify the size of the setjmp buffer?  Even if the
>>>    SH3 has no FPU the room can be reserved and so more compatibility
>>>    among the platforms can give achieved.
>> 
>> Sure. We should unify it for the kernel math emulation, for example.
>> Niibe-san, do you have any opinion about this?
> 
> Note that Ulrich says about the size, not the implementations of
> setjmp/longjmp.
>
> I agree Ulrich.  We can unify the _size_ of the setjump buffer, it's
> the way to go.  When someone will implement kernel math emulation for
> SH-3, he could use the reserved fields with his implementation of
> setjmp/longjmp.

ok. We need to change the content of sysdeps/sh/sh4/__longjmp.S
to suchlike

#define HAVE_FPU
#include <sysdeps/sh/__longjmp.S>

and rewrite sysdeps/sh/__longjmp.S using HAVE_FPU, don't we?

	kaz