Sean Davidson writes:

[description of dynamically-linked /sbin and /bin, and patches, snipped]

Sean,

Thanks for sending the patches. I've incorporated the patches, or
equivalents, for some of them. A more-or-less identical crt0.c had just
gone into the tree.  The sys/arch/pmax/stand changes are committed
too.   if pppd is really broken, you should send a PR.

I'm doing a `make build' with a bsd.own.mk with patches based on yours
(but with AINC+= -KPIC also) to check that it really gets everything
built PIC in a way that keeps the linker happy.

I can't apply the patches for shared binaries in /sbin and /bin as
given: that would affect all NetBSD ports. Perhaps doing it inside #if
(arch == "mips") would work.  I'm also thinking about what else needs
to be added to your list:  root's shell, if it's not sh.


>The only things missing form me in NetBSD/pmax is the dlopen, dlcose, dlsym
>routines for program control of loading shared libraries, R4000 support so
>I can put my R4000 cpu in my DECstation 5000/125, and and X11R6 X server.

dlopen, dlclose, and dlsym are dependent on ld.so.  I think the right
long-term fix is to get the NetbSD/alpha ld.elf_so working on mips
boxes.  I've added definitions of the mips-specific ELF relocs, but I
need some assistance from either cgd or Matt Thomas to figure out how
to implement them in the NetBSD/alpha elf ld.so.

Work on the r4000 support is in progress; Michael Hitch has a 5000/50
booting to single-user mode and running programs, albeit with the
caches disabled.  There is still some locore work to do to eliminate
the double-mapped kernel stack.

I have half of the Ultrix ws interface used by X11R6 implemented.  The
ws ioctls, framebuffer mapping, and colour-register access mapping are
implemented and appear to work on a cfb.  This is reverse-engineered
from looking at the X11R6.1 X server source and from tracing an
X11R6.1  Xserver compiled for Ultrix.

I haven't committed the ws changes because
	* the code is still prototype quality,
	* an X server that doesn't handle input is useless.
	* I did it as  an Ultrix-emulation hook, rather than
	  as  native NetbSD/pmax framebuffer ioctls.

Are you interested in reworking the in-kernel input-queue (keyboard and
mouse events) to use the ws format, rather than the 4.3bsd qvss/qdss
format?

>I'm including my changes if anyone is interested.

Yes, with thanks...