Hi; I've also been playing with Takashi Hamada's kernel
	(PMTEST.961214.120.sbc) for the last few days, and I'm happy to
	report that, and had some additional information.
	My configuration is pb520c, 12M ram, 100M unix partition, booted
	with all extensions off.

	For starters both the internal _and_ external monitors work in both
	1 and 8 bit mode (yay!).  I tried the external monitor in 640x480
	and 832x6?? (whatever the number is for Applescan 14" monitors).  

Luca Bonci (sp?) wrote:
> 
> Hello, I installed the Hamada sbc Kernel on my PB520 (12M Ram, 60M Unix 
> Partition) and I am really satisfied because finally I could boot
> NetBSD on my system.
> 
> I found the following problems:
> 
> sometimes, using common command like 'ls' or 'man' or 'vi',
> a 'segmentation fault' error is generated and a ls.core (or vi.core ...)
> file is dumped. Anyway the system does not crash and if you repeat the 
> command it works.
> 
	I had similar behavior as luca regarding random crashes (just keep
	typing the command in until it doesn't crash.)  I thought this might
	indicate some problem with the 68LC040's cache (since when macs
	with 040's started appearing, apple provided a way to turn the
	cache off.)  So I dragged out a copy of cache switch, and shut it
	off.  The speed difference in the MacOS is noticable, btw.  Anyway,
	the machine wouldn't boot into netbsd anymore.  It got the the
	point where it asked which disk was the root partition (which it
	occasionally does), and the keyboard didn't seem to work.  Is there
	code in the adb driver that requires/expects the cache to be on?
	So anyway, for the time being, it seems that the cache needs to be
	on (I didn't try a serial console).
	
> adduser does not work: the system answer 'command not found'
	Didn't try it.

> passwd does not work: the system reply is 'password file is busy'
	This happened to me when vi crashed while I was editing the
	passwd file.  I just deleted /etc/ptmp and vipw and passwd
	started working again.
	
> 
> I tried also to use the C compiler with a very short and stupid
> 'test' program. The compiler seems to work but the system does not
> recognize the executable, i.e., the answer is 'command not found'.
> 
	I think this is just because root doesn't have '.' in it's path
	by default.  I tried this too, using ./commandname and it worked
	fine.

	I did play around a bit with compiling.  Hello world worked for
	me, anything involving floating point math returned -0.0 as the
	result of any calculation, and any double precision math caused
	a bus error, and occasionally hung the machine.  I tried both
	with and without the -msoft-float option on compile - no apparent
	difference.  I was using base12 binarys and the non-fpu math libs.
	I'm installing the current binaries (as of November 30th) from 
	wormey's page to see if that changes anything.

> I hope this report can help in fixing the bugs,
> 
> regards,
> 
> luca
> 
	Me too,
	Blaise