Date: Thu, 1 Jun 1995 07:11:46 -0400
   From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>

   >  -- NetBSD mounts `/dev/sd0a' on `/', regardless of what drive it was
   >     booted from.  I expected, and would have preferred, the SunOS
   >     behavior of mounting the root file system from the `a' partition
   >     on the boot device.

   1.0 did this.  -current attempts to do what you want, at least if your
   kernel was built with "swap generic"; of course, nothing will help if
   the kernel was explicitly configured with "root on sd0a".  I think the
   -current code to do this is broken, though.

The binary snapshot of -current I just downloaded still mounts `/dev/sd0a',
even though configured with `swap generic'.

However, I'm a lot farther along now.  My X11-R5 binaries work, except of
course for `xload' and a little hack I use called `perfmon' -- I attempted to
kludge-rebuild the latter, but without success -- I kept getting "undefined
symbol: _iob".  Anyone know what that is?

Let's see, some more things I ran into:

 -- (Minor) /dev/MAKEDEV is missing two dots in the line that sets `unit' for
    the `cgeight' devices, which were apparently just added.  (Random Q: what
    is a CG8, anyway?  I know, a new video card, but what does it do,
    resolution-wise and such?  I'm looking to upgrade my CG3 soon...)

 -- One of my disks is an old Maxtor LXT-213S that apparently doesn't
    understand the LUN concept.  The -current kernel thinks it's eight drives,
    one at each LUN.  Fortunately, this didn't get in my way too badly -- it
    did push one of my other drives up to `sd10', after which I was not able
    to mount it (`mount', oddly, complained that the directory I was trying to
    mount on did not exist, when it clearly did) -- but there was nothing on
    that drive I needed immediately.  However this could screw somebody else
    trying to install this kernel... seems like another reason why the
    distributed kernel should have hardwired SunOS SCSI mappings.

 -- I don't know if this is of any particular consequence, but the new kernel
    claims my Imprimis Wren VII and Micropolis 1924 are SCSI-1.  I'm 75% sure
    that the Wren-VII is SCSI-2, and 98% sure that the 1924 is.  Is it
    possible that the kernel is wrong about this?  It does correctly identify
    my Seagate ST112550N as SCSI-2.

 -- `mount' doesn't seem to know about the new clean flags -- first `fsck'
    says that a filesystem doesn't need to be checked, and then `mount' says
    that it does.

 -- Another requirement for running many SunOS binaries is to symlink
    `/etc/termcap' and `/usr/lib/termcap' to `/usr/share/misc/termcap' (I
    don't really know that both links are necessary; I just made them both.)

 -- The SunOS `tcsh' sorta half-worked, but had big trouble with signals.  I
    succeeded in rebuilding it, but to do so I had to bring over the SunOS
    `vfork.h' -- no such file seems to exist in NetBSD.  Shouldn't it be
    there?

 -- I had to manually turn on `clocal' on the serial ports before I could use
    `tip'.  How is this normally handled?  I.e., should I just add the
    appropriate `stty' commands to `/etc/rc.local', or is there a better way?

 -- Is there documentation on setting up SLIP?  I brought over my SunOS CSLIP
    2.7 configuration files and `tip' executable, and that almost worked but
    not quite.

 -- I can't untar the X11-R6 binary (`X11R6pl11.tar.gz') that I downloaded
    from Gatekeeper.  Yes, I'm using GNU tar, and yes, the file checksums
    correctly.  Tar gives messages like it gives when a gzipped tar file has
    been FTPed in text mode -- maybe that happened before the checksum was
    computed?

-- Scott