> 
> > > My changes don't in any way remove the serial console option.  All it does
> > > is make sure that mrg_init() and friends get called regardless of whether
> > > the boot is from serial console or not.
> >
> > The fact that mrg_init wasn't called if you were on a serial console made
> > serial console run even if mrg_init hung the machine. With this change,
> > if mrg_init doesn't work, we're hosed.
> 
> It might be a good idea to simply have MRG check a global that says if we're
> serial booted or not.

It does. Paul's talking about removing that test.

> I'm going to need to back up Paul here :) Some day we may have a grand
> unified grf driver and we're going to need his MRG routines to drive all
> the Video NuBus cards that don't specifically have a driver for them.

We're not talking about removing mrg_init if you're booting w/ the normal
keyboard (though maybe we need to do it even if we're using the ADBDIRECT
stuff; I'm not sure there). :-)

> > I think we need some way of passing generic parameters to the kernel, so that
> > we can customize a boot w/o adding everything and the dog to the booter.
> 
> :-)
> 
> That's a different issue.

I agree. Though with the extra parameters bit, an advanced user could boot
w/ a serial console, and tell the mrg_init stuff to still run. :-)

Take care,

Bill