On Thu, 2003-06-26 at 12:39, D'Arcy J.M. Cain wrote:
> Odd question I guess but here is the situation.  I thought that it was time to 
> bump the home system (i386) up to -current from -current 1.6Q Mar 22/03.  I 
> built and installed a new kernel with today's sources and got the following 
> message (as read from the screen by my wife over the phone to me.)
> 
> syslogd:unknown property name 'local0'
> kernel page fault trap code = 0
> stop at NetBSD:syn_cache_get+0x308:movl0(%edx),%ecx
> db>
> 
> Now, I wonder if I should be concerned about this or will it be fixed if I 
> build the userland programs.  Oddly enough, this version is running elsewhere 
> with no problems on a system that we couldn't keep up for 30 minutes under 
> 1.6.1.  We updated the kernel there so that we could use a different SCSI 
> controller so we suspect that there was a hardware issue on that one.,
> 

I think you should update userland when updating the kernel when using
current. I thought that was documented.

A userlang program that does a system call incorectly should not crash
the kernel unless it bypasses kernel security by accessing /dev/mem
directly. I think X does that i dont know about other probrams.

> I'm not sure what to do here.  I understood that it was generally safe to 
> update the kernel without updating the rest of the system.  Hopefully that's 
> still true and this is an anomaly.
> 
> Thanks.