CONFIG_NUMA and DISCONTIG inconsistencies

[Olof Johansson]

On Mon, Mar 21, 2005 at 11:24:48AM -0800, Dave Hansen wrote:
> > If you want to
> > remove it, please do it explicitly under a different patch.
> 
> I actually attached two plain-text patches in that message.  They were
> separated when I attached them :)

One patch per email, please. So, separate patch = separate email. And
that means separate patch description, which would have avoided this
chain of emails in the first place.

Anyway, back to your original question:

I think we have machines with discontigous memory (due to I/O holes)
without having NUMA support on the machines, i.e. some POWER3/RS64
boxes.

Unfortunately I don't have such a machine with sufficient memory
configured to take a look at.

> > How do you expect this to build if you remove ARRAY_INITIALISER?
> 
> Hehe.  Minor detail.  I think Joel is going to make a more presentable
> (read: compiling) version.  I'm sure he'll get that part.

Joel just commented your patch and acked it, with the same build error
in it. I guess that got me confused.


-Olof