Hmm,
It always worked for me although sometimes my own kernel had more=20
problems than the standard one.
I just followed instructions here :

http://netbsd.org/Documentation/kernel/#how_to_build_a_kernel

Although as I never use -current I never build tools just use the ones=20
that come with the release.
I build new kernel to free up some room. IMHO, GENERIC has way too much=20
things in it. Qube has pretty limited hardware and there is no need to=20
include all of the SCSI things plus all other raid stuff, etc. But I=20
guess if you cater to everyone's taste you'd have to build 50 different=20
kernels.

Regards,
Alex

Peter Fr=F6hlich wrote:
> Hi all,
>=20
> So my Qube is running 3.0 fine, everything's working nicely, I got all=20
> the tools I need for my compiler course installed as well, and now I=20
> made the mistake of trying to build my own kernel again...
>=20
> This has *never* worked for me before on the Qube, neither on 1.6, 2.0,=
=20
> nor 3.0 as it turns out (last time it *did* work for me was 0.9 or=20
> something on an Amiga over 10 years ago, how sad is that? :-). Can=20
> someone please clearly answer this question:
>=20
>   Am I supposed to build the Qube kernel on the Qube using build.sh or =
not?
>=20
> If I am supposed to, then it's just not working. I can rebuild the tool=
s=20
> for hours, try to build the kernel for hours (which always rebuilds=20
> nbmake as well, what a strange waste to always do this whether I say -u=
=20
> or not), but it also always ends up breaking somewhere. Am I just so=20
> dense that I can't follow supposedly simple instructions from the NetBS=
D=20
> Guide?
>=20
> If I am *not* supposed to use build.sh, then could someone *please* add=
=20
> that to the NetBSD Guide? Clearly? Just say somewhere:
>=20
>   If you're *not* cross-compiling, stay the heck away from that weird=20
> build.sh stuff!
>=20
> Sorry, I am obviously slightly frustrated... :-/ I just can't believe=20
> that pkgsrc is *so* straightforward to use, but a kernel build can be=20
> *such* a pain...
>=20
> Cheers,
> Peter
>=20
> PS: Of course the "direct" approach doing a config followed by a make=20
> (without build.sh) works just fine. I am not sure yet that the kernel=20
> actually works, but at least I can build one that way.
> --=20
> Peter H. Froehlich <><><><><><> http://www.cs.jhu.edu/~phf/
> OpenPGP: ABC2 9BCC 1445 86E9 4D59  F532 A8B2 BFAE 342B E9D9
>=20
>=20
>=20