13 messages in com.xensource.lists.xen-develRe: [Xen-devel] uint64_aligned_t not ...
FromSent OnAttachments
Jan Beulich28 Aug 2006 07:46 
Keir Fraser28 Aug 2006 08:01 
Jan Beulich28 Aug 2006 08:32 
Keir Fraser28 Aug 2006 08:35 
Jan Beulich29 Aug 2006 01:49 
Jan Beulich29 Aug 2006 03:12 
John Levon29 Aug 2006 05:37 
Jan Beulich29 Aug 2006 05:46 
Keir Fraser30 Aug 2006 09:18 
Keir Fraser30 Aug 2006 09:46 
Jan Beulich30 Aug 2006 23:41 
Jan Beulich30 Aug 2006 23:53 
Keir Fraser31 Aug 2006 11:03 
Subject:Re: [Xen-devel] uint64_aligned_t not compatible across gcc versions
From:Keir Fraser (Keir@cl.cam.ac.uk)
Date:08/30/2006 09:46:03 AM
List:com.xensource.lists.xen-devel

On 29/8/06 1:46 pm, "Jan Beulich" <jbeu@novell.com> wrote:

/* Structural guest handles introduced in 0x00030201. */ #if (defined(__XEN__) || defined(__XEN_TOOLS__)) && !defined(__ASSEMBLY__) -typedef uint64_t __attribute__((aligned(8))) uint64_aligned_t; +#define uint64_aligned_t uint64_t __attribute__((__aligned__(8)))

This is GCC-specific, in a public header. At the very least it needs to be conditional on gcc being used (#define __xen_aligned or whatever)...

That is a comment regarding the original change, then. I still agree this should be abstracted out in some way - the question is how, since I guess users of other compilers would need to provide input.

I'll admit there's still the question of whether this is worthwhile for just these two hypercalls in the first place. Jan: do you think much code will be saved by explicit alignment for domctl/sysctl, or do you think we're just as well to remove uint64_aligned_t and XEN_GUEST_HANDLE_64, and do compat shims for domctl/sysctl just as we are for all other hypercalls?

-- Keir