2.8bsd on 11/23

utzoo!decvax!harpo!npois!cbosg!cincy!jreuter utzoo!decvax!harpo!npois!cbosg!cincy!jreuter
Sat Apr 10 14:33:23 AEST 1982


Item 2) As for the difference between 1kboot and 40boot, You will find
that 1kboot is actually LARGER than 40boot.  They both use the same
standalone *.c code and differ only in the very small assembler code.
I found this out when I had the misfortune of having to single step
through the code.  It must have something to do with I&D vs. nonI&D
machines or something.  If you are daring, try to go right from bootup
to 40boot.  Maybe it will work.

Item 3) You are supposed to use 512restor to read the Bell tape for all
the unchanged Bell V7 files, but 512restor never worked right here.
Chris Kent had a friend send us a tar of all the Bell sources.  We
still may be missing a few things (like libU77.a).  Has anyone gotten
512restor to fly?

As for ovadb and that f2 stuff, that is because ovadb is a real kluge.
I understand they have fixed that recently.  Also, if you are using
small disks, you might consider moving all the /usr/ucb stuff to
/usr/bin.  On our 11/23, we have a Winchester that looks like 4 rk's.
One is entirely /usr/bin (with /usr/ucb in there).  This was necessary
because the /usr disk is one rl01.  (/usr/lib is another rk.)  Also,
csh builds the hash table faster with fewer directories.  This will
cause problems with pathname assumptions in Mail, vsh, etc.

Item 4) I moved EVERYTHING to what would be considered a "usual"
directory.  In other words, there is no /usr/ucb, /kernel, or
whatever.  It is pretty much a standard v7 hierarchy.  This made things
much nicer, except that in /usr/src/cmd we now have Berkeley's makefile
and Bells cmake and whatever else there was.

Item 5) Libraries:  You might not know that /usr/src/cmd/fstab.c really
belongs in /usr/src/lib/libc/sys with an appropriate entry in the
compalls and mklibs.  This thing is used by the Berkeley df and the
4.1BSD mount and umount.  Dump also seems to use it.

Item 8)  Don't waste you life on Ingres.  It is huge and probably will
never run on a nonI&D machine without tons of work.  Also, one part of
Pascal won't fit.  I heard in the wind that someone overlayed it, but I
don't know for sure.  I asked Carl but ucbvax seems to be a black hole
for all of my mail.  I haven't gotten replies in quite a while.  I put
in a Pascal that was from Tektronix on a USENIX tape (Delaware tape, I
think).  True, vi runs right away, but you may find that it doesn't
know where the error message file is.  There was a naming problem there
( do a strings on vi to find what the real name should be.)  Anyway,
that is an old version of vi and version 3.6 can be compiled and run
once you fix the makefile.  There were two bug reports to that effect
at one time.

Odds 'n ends)  Because of the funcky UCB stdio package, several things
need to be sprinkled with flushes.  The programs that come to mind
are:  write, passwd, user(8), leave, clock, chessclock

The kernel option UCB_PGRP doesn't work.  It limits ok, but process
groups are never properly established, as far as I can tell.  I found
this out when *I* ran out of processes while somebody else was running
spell!

The kernel on our 11/60 crashes horribly, but this seems to be entirely
because of some problem in combining a dz with a second-source rp.  Our
23 has run for as long as 9 days under production use (I took it down
after that for some testing).  It has hung up a few times--it just
seems to grind to a halt, and that only happens when vi is running.
Anybody have any ideas?
	Jim Reuter



More information about the Comp.bugs.2bsd mailing list