Subject: Re: 64-bit ABIs
To: Ben Harris <firstname.lastname@example.org>
From: Steve Woodford <email@example.com>
Date: 04/29/2003 23:09:32
On Tue, 29 Apr 2003, Ben Harris wrote:
> On Tue, 29 Apr 2003, Steve Woodford wrote:
> > I'm not an authority on these matters, but I'm pretty sure that the sh5
> > ABI requires 32-bit values to be correctly sign-extended when held in
> > registers, regardless of the signedness of their type. I've certainly had
> > to tweak some sh5 asm to make sure unsigned 32-bit results are still
> > sign-extended.
> OK. That sounds like we're stuck with RENAME.
To be honest, if the only ABI to be affected by this change is sh5, then I
would say forget about RENAME and just do it. I know of only two people
with sh5 hardware running NetBSD; myself and Jason Thorpe. I think either
of us could recover from any shlib lossage which might ensue. :)
> <http://www.chiark.greenend.org.uk/~bjharris/inet.diff> is a simple patch
> to change us over. The important question is whether applications built
> after this patch work correctly with libc built before it.
Thanks. I'll try to check it out over the next few days.