I'm seeing some strange things with the 1.5Y snapshot and -current on my 
hotrodded G3.

The 1.5Y snapshot boots over ethernet and runs fine in a diskless setup 
on my B&W G3 with the CPU upgraded to a G4.  The USB keyboard still 
doesn't work unless I boot the system with the ADB and plug in the USB 
after the system goes multi-user.  This is the same as I've seen with 
every kernel since 1.5 though.  BTW, comparing log files this appears to 
be a wscons configuration problem.  In the 1.4.3 kernel the keyboard gets 
glued to the display by some callback code that is executed near the end 
of the device configuration process.  In the 1.5+ kernels this never 
happens during boot, but does happen when the USB is plugged in after the 
system goes multi-user.  The callback code in the current version of 
wscons is quite different and the hack in machdep.c that probably invoked 
this doesn't exist anymore. It looks like the equivalent to the 
wscons_glue code isn't executed or exist anymore and as a result the 
keyboard and display don't get glued together anymore.

Anyway, this isn't the real problem I'm seeing.  While trying to test 
some CDROM disk label code for someone I tried to install the cdparanoia 
package.  It installed but when I try to run it I see a PLT error 
attempting to reference getoption_long.  Thinking this was due to a 
package seriously out of date I tried rebuilding the package.  Rebuilding 
causes a kernel panic in nfs_socket at line 929 with the active process 
being "update".  It was suggested that I try building a kernel with 
NEWPMAP as this is more stable.  Unfortunately the kernel doesn't come up 
on my system with or without NEWPMAP.  The sources were sup'ed yesterday. 
The kernel was a plain GENERIC configuration.  It boots over ethernet but 
when it gets loaded and ready to blank the console screen and display the 
BSD banner, it seems to bomb and drops back into MacOS.  The 1.5Y 
snapshot kernel boots and runs fine in the same setup.

Anyone have some ideas here?

Thanks,
-bob