-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

> Perhaps we can expand the interface to allow some kind of identifier
> to say, "No, really.  I mean it this time." and just override what's
> already there.
Yeah, something like that. Or some kind of 'preliminary attach' - like, 
use this for now so you can complain when something goes wrong but I'll 
attach the real one later on.

> It's kind of wierd because we really only want to have one thing
> attach (i.e., if machfb is there, don't attach ofb).  But ofb and
> machfb (or whatever) are on different busses, so the autoconf
> framework for that (match function return values) doesn't work.
I don't want to attach ofb at all, just something to output kernel 
messages until the real framebuffer driver gets attached ( which may be 
ofb ), mainly to allow them to be architecture independent. Right now 
macppc needs special handling - ofb_cnattach gets called in machdep.c, 
any replacement framebuffer driver would have to support that too and 
then you'd have to hack machdep.c anyway. That's what I want to get rid 
of.
Right now I just disabled the KASSERT() line which prevents 
wsdisplay_cnattach() to work a 2nd time - ugly, but so far it doesn't 
seem to have bad side-effects.

btw. is there a clean way to tell autoconfig in which order it should 
try drivers? ofb is kind of a catch-all, so it should be tried last, 
after all other framebuffer drivers we may have.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iQEVAwUBQf2m48pnzkX8Yg2nAQLoCQgAn5ozfIUq2EuzHYzwnhRlKuLTlc4PtjNW
ddZ6V+DcaeUk7v0/p1Om/Buv8XeIFG2CcxsYrLTv2PjyihnqQ7APKAY1K+eLgb76
NRhodJQ2YWaErWQuWHxDiyYna5/tcOBa3QFkTPVUiVkrpxS6zJHWxMvEqWsB/u7s
L3v5qNTr+r7Sc4IzIsuwl2Xu3bOaNG24fBKteGHQ6tOACmS4N8yDIULVKHpkYQ8B
oDjz7Cpxl01Y4r2WCfmS9oL/QENe2c2AlB7sDE5g9ZM+8GLlaGaYGOzQOQWIdiIE
cde9oO2iLzUUSjaSZHK0hMCw1eDVprDlQ42Vh9rbVgwPsf/XOecvMQ==
=pceW
-----END PGP SIGNATURE-----