> I think Colin has missed the fact that we are not dealing with a monitor,
> but the actual video card on the IIcx. Remember, the IIcx is special in
> that it does *not* have any internal video. It must have some type of
> video card in it. I don't see anyway around this. You could buy a really
> cheap used video card to pop into the IIcx just to boot it.  Then, the
> paperclip trick should work. Personally, I use a Mac <-> VGA adapter on my
> Centris 650, just in case I want to plug in a monitor to see what's going
> on.

The kernel should be able to boot with no video devices present at all.
If it can't I'd be willing to bet it's because the ite driver is getting
confused. I would imagine that setting the console to serial console should
do the trick for truly headless machines (even with no actual terminal on
the serial port).

As you pointed out this is different from the problem on IIsi/IIci's and other
machines. That actually has to do with memory allocation and monitor sensing.
I'm working on the problem as part of intvidtest2.

Hopefully by 1.4 we'll be using wscons/fb (with intvid code) instead of
the ite/grf combo we're using now. All of these strange problems will go
away.