Subject: Re: pkg/10616: sparc/cpu.h includes non-installed kernel includes
To: John Hawkinson <jhawk@MIT.EDU>
From: Dr. Lex Wennmacher <wennmach@geo.Uni-Koeln.DE>
List: netbsd-bugs
Date: 07/31/2000 18:43:08
On Jul 29,  1:28pm, John Hawkinson wrote:
> | State-Changed-Why:
> | userland programs should not define _KERNEL or face the consequences.
> That's the crux of the question. I don't think that is correct.
> It seems to me that userland programs can and should/must define _KERNEL
> if they want to see kernel data structures.
> Fundamentally, I guess this is a style question. Where do we seek the answer?

After looking more into this ... I noticed that it has been discussed before
(see PR 5377).

> I'm going to dupe 10696 into this PR since it was first ;-)

Maybe 10616 should be duped into 5377 since it was first ;-)

> I'm not sure I can agree with you and mrg.

I'd like to see the requirement to have the full kernel source available under
/sys to build LKMs go away but I'm well aware of mrg's argument that this will
be a lot of work.

I will update the arla package to check for the presence of /sys. Still, I
consider this a (temporary) work-around as it is an absolutely sub-optimal
solution (think bulk-builders).

> Lex, that will downgrade your PR from critical/high to non-critical/medium.
> Can you justify the severity/priority you asserted?

Actually I think that PR 10616 should also be closed. It is neither a
port-sparc problem nor a pkg problem. jhawk, let's dupe it into kern/5377. I'd
like to see the severity of this PR go to "serious", though (the requirement to
have the full kernel sources available to build a package have quite some