Subject: A new user's comments
To: None <port-sparc@NetBSD.ORG>
From: Scott L. Burson <gyro@zeta-soft.com>
List: port-sparc
Date: 06/01/1995 11:51:47
Wow!  Thanks for all the replies.

I sent that message twice, the second time adding the postscript -- sorry if I
neglected to stop the first copy from going out, as I thought afterwards I
might have.

Now, let's see...

   Date: Thu, 01 Jun 1995 18:08:38 +1000
   From: matthew green <mrg@mame.mu.OZ.AU>

       -- 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.

   this is most likely because the kernel was configured with:

   config netbsd root on sd0 swap on sd0

   if

   config netbsd swap generic

   worked, then you wouldn't have this problem.  (it *will* work soon,
   or maybe even now, but i don't think so).

Ah.  Okay.

       -- I have to say that I don't at all understand the purpose of numbering the
	  SCSI disks the way NetBSD does.  The prospect of either changing the
	  jumpers on my other drives, or globally editing /etc/fstab, whenever I add
	  or remove a disk other than the one with the highest SCSI ID doesn't
	  appeal to me.  I know, I can turn it off if I want, but what I'm getting
	  at is that I think it should be off in the distributed installation
	  kernel, which should have, hardwired, either the standard SunOS mappings
	  (0->3, 1->1, 2->2, 3->0) or direct mappings.  As it was, I used the
	  `id3_scsi' kernel and wound up with 0->3, 1->0, 2->1, 3->2 which was
	  unexpected to say the least.

   this is related to the above;  it "works" under sunos because the
   kernel is generally compiled with hardwired mappings from 3120 to
   0123.  put this in your config file:

   sd0     at scsibus? target 3 drive ?
   sd1     at scsibus? target 1 drive ?
   sd2     at scsibus? target 2 drive ?
   sd3     at scsibus? target 0 drive ?

Right, I understood that I can reconfigure this, but I was suggesting that
(particularly as people will be migrating to NetBSD *from SunOS*) the
distributed kernel should be set up this way.  Once we get NetBSD booted, we
can set this however we like, or use the NetBSD consecutive-numbering feature.

   From: Jason Thorpe <thorpej@nas.nasa.gov>
   Date: Thu, 01 Jun 1995 08:07:25 -0700

    >     (0->3, 1->1, 2->2, 3->0) or direct mappings.  As it was, I used the
    >     `id3_scsi' kernel and wound up with 0->3, 1->0, 2->1, 3->2 which was
    >     unexpected to say the least.

   I thought this was mentioned somehow/somewhere in the installation 
   notes...

Oops, gack, you're right, it is specifically mentioned... I missed it :-)

       -- The installation instructions say nothing about `pwd_mkdb'!  I had to do
	  some sleuthing to find this sucker.  The instructions should really say
	  specifically how to install one's SunOS `passwd' file:

      [ ... ]
	    [Perhaps a `sed' script or something could be supplied to do this?]

   that would be nice;  do you want to write it ?

I knew you would say that :-)  I guess it's about time for me to learn `sed'
after all these years...

	       # pwd_mkdb passwd.tmp
	       # pwd_mkdb -p master.passwd

	    The first execution of `pwd_mkdb' updates `/etc/spwd.db'; the second
	    regenerates the SunOS-style password file (minus the encrypted
	    passwords) into `/etc/passwd', where SunOS executables may expect to
	    find it.

   you should be using vipw(8) for editing the password database.

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

   >  -- The installation instructions say nothing about `pwd_mkdb'!  I
   >     had to do some sleuthing to find this sucker.

   As mrg says, why weren't you using vipw?

Oh.  Believe it or not, I didn't know about it.  Directly editing
`/etc/passwd' has worked fine for me ever since the days of V7 -- since I've
never used YP, er, NIS -- and I just never learned the modern way.  I guess if
I'm the only one in the world who doesn't know about this then there's no
reason to mention it in the instructions... but I don't know, there may be
more folks like me out here...

       -- It would be nice if NetBSD came with a precompiled version of `tcsh',
	  since the SunOS executable doesn't work (yet?).

   it doesn't ?  i might look at this..

Well, not in 1.0.  I will try a more recent version... (see below)

   >  -- To run many SunOS executables, it is necessary to
   >       # ld -s /usr/libexec/ld.so /usr/lib
   >     At least, I guess this is the right thing to do -- it seems to work.

   [der Mouse]
   Do you mean "ln -s"?  I can't see how "ld -s /usr/libexec/ld.so /usr/lib"
   would do anything useful.

Yes, of course, oops :-)

   try:  man compat_sunos

Okay, thanks.

   > After I got the system running, it became evident that the SunOS
   > binary compatibility is not quite where I need it to be to
   > permanently change over to NetBSD.  I tried to start my SunOS X11R5,
   > but `xinit' got a bad system call error.

   I think this is probably because you're using the NetBSD ld.so and
   libraries.  Try with the SunOS ones:

	   # mkdir /sunos-compat-hacks
	   # mount -r sunos_host:/usr/lib /sunos-compat-hacks
	   # mount -t union -r -o -b /sunos-compat-hacks /usr/lib

Yow!  So *that's* what union filesystems are.  Actually, I only have one
machine -- I presume it will work to just symlink `lib*{.so*,.a}' from the
NetBSD `/usr/lib' to the SunOS one?

       -- Q: What is the command, corresponding to `pstat -s', to see how much swap
	  space is available?

   um, pstat -s ?

   splode ~> pstat -s
   Device      1024-blocks     Used    Avail Capacity  Type
   /dev/sd0b         50400    23556    26844    47%    Interleaved
   /dev/sd1b             0  *** not available for swapping ***

Huh?... oh, I guess I didn't have `/usr/sbin' on my path.

   there is a pre-built X11R6 on the netbsd ftp sites;  it might only work
   under -current, however, as i only have one machine and i tend to keep
   at -current most of the time.  if you get X11R6 patchlevel 11, and apply
   the small patch i've made available with the binary, you should be able
   to build it out-of-box, modulo a small bug in xdm where it doesn't call
   setlogin(), and thus getlogin() returns "root" always..

      [Oh -- I just discovered the X11R6 binary on Gatekeeper, which by the way is
      where I got everything.  Okay, I'll use that.]

   oops.  that will teach me to answer before reading on  :-)

   if you use xdm, i would still fix that anyway.  if you can't find the
   patch, email me and i'll send it to you.

I don't use `xdm'.  Someone said they were running their SunOS R5 binaries
under -current, so I'll try that.

   > What is the degree of commitment to the SunOS compatibility?  Is 100%
   > compatibility a goal, or even possible?

   [der Mouse]
   Not possible.  There are too many little details that are different.
   For example, it strikes me as extremely unlikely that anyone will ever
   make it sufficiently compatible that SunOS trace(1) will work, and it
   could never work for tracing native NetBSD executables.

Well, okay, I don't care about running SunOS executables for which a NetBSD
equivalent exists, like the various system programs.  If X11R5 works in
-current, and the Lisp I use works or can be made to, and I don't have to
rebuild more than a handful of the GNU and other random utilities I use, and I
can run things like Netscape, I'll be happy.

   Date: Thu, 01 Jun 1995 18:11:43 +1000
   From: matthew green <mrg@mame.mu.OZ.AU>

	 NetBSD-current's fsck does the former.  Every version of fsck ever
	 shipped with NetBSD does/did the latter.

      ??  I didn't notice it doing so... maybe I just missed it.

   didn't notice which ?  netbsd 1.0 does not have a "clean filesystem"
   flag implimented, but -current does.  you won't see this in 1.0 because
   it's not there.

I didn't notice it checking the filesystems in parallel.  (And it's something
one would think I would notice, with three filesystems in the 900 - 1300MB
range.)  Oh well, no matter, I'll try -current.

Thanks again, everyone, for all the replies.

-- Scott