ok, i finally had a chance to try the new iisi kernel and it worked!

mac iisi, internal video (b&w), 32-bit mode, 8M

i grabbed that macrom.c and compiled it into the sources from around 04/19
and i had yet another working kernel.  some quick observations: i did need
MRG_DEBUG set, however i was able to build a kernel w/o MRG_DEBUG as long
as i enabled the #ifdef'ed printf's in mrg_aline_super() leading me to 
think that there is some timing problem (possibly explaining someone 
else problems with a over-clocked si)  it also seems that the si kernel,
periodically causes a lvl1dtpanic (leading to code becoming an empyty
routine) about once per second.

anyhow, while designing cards for the mac and the Slots both agree that
the si has an internal video card at slot 0xE, nubus_probe doesnt seem
to be able to see it (checking the last four bytes at the top of the
declaration rom seems to cause a bus error every time--i.e. not valid
bytelanes)  how does Slot's determine its presence?  does it just know
becuase this is a ii si?