[comp.sys.sun] undump on 4.0 sun3

eggers@cs.buffalo.edu (Bill Eggers) (12/22/88)

Hello,

I recently tried to install the new TeX distribution (2.9) on our suns.
We have a 4/110 serving a 3/50 and a 3/60, all running SunOS 4.0.
Everything compiles ok on both machines, but I have had problems getting
undump to produce a working executable (tex, latex, or slitex) on a sun/3
running 4.0.  Every time I try to run one of those on a sun3, I
immediately get a segmentation fault.  However, it works fine on our sun4.
(I'm used to the reverse being true.)  I am referring to the
undump4.0.shar obtained from the Rice archives, and the undump distributed
with TeX (which are essentially the same).

In fact, this is not just limited to the TeX distribution.  I've tried
using undump with this simple program, and it still bombs for me:

        int j[100000];
main()
{
        int k[100000]; int c;

        printf("Hi There\n"); abort();
}

This problem is not just restricted to the suns here at Chemistry.  I have
tried this on a 3/160, with the same results.  I have tried everything I
could think of to get this to work, and I believe I'm doing the proper
steps to generate all this stuff.  I have tried both -Dsun3 and -Dsun4
options in the makefile, I've even tried using the GENERIC kernel, but I'm
out of ideas.  

I am aware of problems with undump and the fpa. However, I can't believe
that, as far as I know, I'm the only one who has not had this problem with
TeX.  (somebody else in the CS department has the same problem, so I'm not
going crazy.) It can be either one of two things:

1) I'm doing something wrong (most likely), or
2) every other sun3 running 4.0 in existance is still using an old  
   3.X executable to run TeX.  

Forgive me if this solution has been posted, but I try to keep up with
Sun-spots, and only the fpa problem is similar.  Can there be a bug in the
way that a sun3 running 4.0 creates a core dump file? Has anybody had the
same problem?  Has anybody else gotten this to work?  Is it fixed in
4.0.1?  Any help would be greatly appreciated.

[[ Well, there is.  The core.h include file has a "#ifdef FPU" toward the
very end of it which should not be there.  This has been acknowledged as a
bug by Sun.  You need to compile undump with "-DFPU" to get the correct
definition of a struct core.  If you are already doing this then there
must be some other problem.  As a side note, people in the TeX community
are now saying that undumping really doesn't buy you that much anymore.
Loading a .fmt file and the required .tfm files is rather quick under Unix
these days.  They are recommending that TeX installers not bother with
undump, create and maintain .fmt files, and write shell scripts to do the
appropriate incantation for "tex" and "latex".  Unfortunately, .fmt files
are different for different architectures (they are probably different
between the Sun 3 and the Sun 4).  --wnl ]]

		Thanks.

Bill Eggers		Dept. of Chemistry, SUNY at Buffalo
103 Acheson Hall	Buffalo, NY 14214
eggers@cs.buffalo.edu	(716) 831 - 3629

emo@iuvax.cs.indiana.edu (01/04/89)

Greetings,

I too was confused for a while how anyone could install TeX on a Sun 3
running Sun OS 4.0; they probably used a 3.X executable instead.  Funny,
though, the "undump" that comes with the current distribution of C TeX
works on our Sun 4s.  [[ If you got the distribution recently, it is
probably the fixed version.  --wnl ]]

Recently I finished writing a version of "undump" for the new Sun core
file format.  I would be happy to make this available to other people.
However, I would like to do this via ftp (not mail) as I am rather busy
with other things these days (our NeXT machines have just arrived! :-).

I will place a copy of my version of undump in /usr/ftp/pub/undump.c on
iuvax.cs.indiana.edu.  If you have problems picking this up, let me know.

We have also just received our Sun 386i and I am in the process of writing
a version of undump to generate a COFF executable file.  If there is
sufficient interest this can also be made publically available.
Alternatively, has anyone already written this?  I'd prefer not having to
write it at this time.

eric     emo@iuvax.cs.indiana.edu