2.8bsd on 11/23 with RLs

[utzoo!decvax!duke!unc!lhn]

-------- Long one.  About 70 lines -------------
I hope that this account of my troubles with 2.8BSD will be useful
to someone.  Also this account contains a few questions which I
myself would like to find the answer to.

To set the stage, let me tell you that my system uses Rl01 and RL02
and consists of an 11/23.  My account will no doubt contain inaccurate
facts.  Please forgive me.


I started with the January 1982 version of the 2.8bsd tape.
1) Had problems with the unix system provided on tape.  It would die when
   booted from tape.  Finally Carl Smith @ Berkeley told us to do the following:
    Boot from tape again.  When it starts to print the section lengths,
     halt the cpu.  Then use the console to patch locations 160, 162.  They
should contain whatever 320,322 (or is it 330, 332) contain.  Then start the cpu
      right where it halted previously.  You should get a
    mem = xxxx      message and a # prompt.
    Yea!

2) Of course I had to fix the rlboot.s code and wrote it out to the boot block.
    I fixed it to look for the 16 bytes headers and to echo input.
   Right now I do not know why 1kboot has to be run first (probably because
    of the sheer size of /unix, even overlaid).  No source of /1kboot is to
   be found, however.  And what is /40boot's purpose?

3) Copied the rest of the tape using tar with no problems.  Followed the
    instructions to install libraries such as /usr/ucb...
   When time came to regenerate the system, I made sure that l40.s had 160
   as the vector address of Rls.  Changed the *.h files as little as possible.
   By the way, I never ran the ovadb portion of the makefile.  What is the
   ...f2... reference about?  I am mystified by that one.  Never had time to
   find out by sitting down and looking at ovadb source.

4) When time came to generate other binaries, I quickly found out that the
   include files were not found in their usual place.  I quickly copied the
   include files used during system regeneration to their 'expected' place
   and things worked.
   The annoying thing was to discover that ranlib was not there.  Also useful
   binaries such as grep, find and nroff were not there.  I happened to have
   a copy of the DEC V7m tape and copies those binaries over, and they worked.

5) In the course of compiling things, I was getting 'Local symbol botch'
   messages.  Finally remembered Chris Kent's account of his experiences and
   remade /lib/libc.a as well as /lib/libovc.a and it worked!
   Watch out for the compall script when assembling *.s files.  It uses cc
   to do so while the following would work better, resolving the symbol botch:

 as -o ???.o  /include/sys/sys.s  ???.s     or something similar.

Actually the last compall entry for a .s file is right.  Just copy that format
for all .s files.

6) Hunted around for ovld.  Finally got a hunch and looked at ld.c, which then
   told me to
   cc -DMENLO_OVLY ld.c      to get ovld!!!
   ovas2 was in the wrong place.  Took a little work (thanks to find) to find it.

7) f77!!!    Finally got f77, ovf77 and f77pass1 binaries.  Finally got 2
   simple programs to compile and run (boy, do they take space!) and discovered
   to my disappointement that built-in fcts such as csqrt, tan were unresolved!
   Which means again that ...libU77.a had to be redone for sure.  Probably even
   ...libF77 and libI77 for good measure.  However, NO source on the 2.8bsd tape!
   This meant going all the way back to the Bell tape, at which point I ran out of
   time and energy.

8) Haven't even looked at Ingres or Pascal yet.  Not even uucp.  At least vi
   worked right away on my 11/23 without recompiling...
   Final question:  has anyone gotten 2.8bsd to work on 11/44 with RLs?
   ours crashes every 20 to 40 minutes in single-user mode...  Haven't dared to go
   multi-user yet!


End of story.  Your comments, help , suggestions are welcome at this point.

Le H. Nguyen