Chris G Demetriou writes (as do others):

>floating point load average calculations have been gone for a long
>time.

Of course.  My apologies for spreading such misinformation.  (I I
hadn't looked at the loadaverage code for a long, long time, and when
I did, I ran into the problem I described. That' s no excuse for not
checking, though.)

And thank you all for correcting my error.

>... and if there _is_ any FP in the kernel, in the pmax port, other
>than register save/restore and trap handling, the pmax port is screwed
>up.

I certainly haven't added any.  I've even tried hard to avoid using
divides.

To go back to the original subject: I, personally, would prefer to use
X11R6 libraries and clients. Arne has done a great job to put together
X11R5 patches for NetBSD/pmax; but X11R5 is, well, ancient history.  X
is a network protocol, and using X11R6 pl13 clients -- perhaps from a
non-pmax client -- don't alwayas worn with an R5 server.  There
may also portability issues with R5 libraries and newer software that
needs R6.

I would prefer to see the ddi code for DEC framebuffers (pre-Xws) cut
out of R5 and put into an R6 tree, and compiled with the
existing NetBSD config.  I don't see why that wouldn't work.

Personally, I tend to use X11R6 binaries compiled on Ultrix, since I
already had those around.  It mostly works -- except for the Xserver
(which needs the Ultrix Xws interface) and xdm (which cannot read
either NetBSD hashed passwd files, or YP passwd maps via the NetBSD
ypbind.)

--Jonathan