Hi.  I just got NetBSD running yesterday on a 386sx/16; the only point
of which is to have a platform from which to boot strap it onto a 
thinkpad 750.  I suspect I have my work cut out for me; since
available technical information for the tp750 is scant.  I have Linux
running on the thing, but the networking code in linux sux, so I'm
giving NetBSD a try.

Anyway, installation on the thinkpad won't get past the second bootstrap
loader (you can get the boot prompt but it can't load anything after that),
so the first thing I'm trying to do is rebuild the boot code in
/sys/arch/i386/stand.  What I've found so far is:

1. the test for the magic number in boot.c doesn't use the N_GETMAGIC
macro, it tries to compare with the a_magic field of the struct, which
doesn't work anymore.

2. wd.c uses the DELAY() macro, which isn't defined anywhere for the i386
architecture.

3. after fixing the above two things, the linker still complains
about ___main being unreferenced.  I haven't tracked that one down
yet.

Anyway, my suggestion for the next release is for someone to make sure the
boot code still builds from scratch.



Another wierd thing I've noticed:

The kernel keeps complaining about ring buffer overflows on ed0.  Okay, so
our net's a swamp.  Eventually I'll hack the driver and comment out the
message.

But I also notice that such messages don't just go to the console or the log
file; they go to random ptys also.  I occasionally get these when I'm
telnet'ed in from elsewhere, but not all of them -- just a random message or
two.  For example: "last message occurred ## times", without it ever saying
what the last message was.



Is there some discussion list I can monitor so as to be aware of the state of
the kernel, and so I'll know whether someone has already fixed a particular
problem?


thanks,

Keith Moore

------------------------------------------------------------------------------