Path: utzoo!mnetor!uunet!husc6!hao!gatech!purdue!i.cc.purdue.edu!j.cc.purdue.edu!pur-ee!uiucdcs!uxc.cso.uiuc.edu!uicsrd.csrd.uiuc.edu!hoefling
From: hoefling@uicsrd.csrd.uiuc.edu
Newsgroups: comp.sys.apollo
Subject: Fooling Fortran
Message-ID: <11400004@uicsrd.csrd.uiuc.edu>
Date: 15 Jan 88 18:04:00 GMT
Lines: 44
Nf-ID: #N:uicsrd.csrd.uiuc.edu:11400004:000:1607
Nf-From: uicsrd.csrd.uiuc.edu!hoefling    Jan 15 12:04:00 1988


All I wanted to do was pass a value to the csh status variable from 
an Apollo Fortran program.  Like a C program would do for "exit(n);".
The manuals had nothing on the subject.  The (800) 2APOLLO help line
was baffled (actually they suggested I write a file containing the 
status code, then cat it once I was back in the shell, but I didn't
care for that solution).

So, in playing around, we came up with a still-pretty-cruddy solution,
but one which hid the cruddiness inside the program, so I don't mind it
so much.

For those interested, here's the Fortran code:

      integer x,y,z
      pointer /z/ x
      equivalence (z,y)
      y = 8
      call exit(x)
      end

This returns the value 8 to the status variable in csh, which you can
see by "echo $status".  Apparently, "call exit" passes the argument by
reference (puts the arg address on the stack), while the exit routine
is expecting the *value* to be on the stack.  What results if you don't
play the above tricks is that the status variable gets what is apparently
the address of the argument.  So that's why we had to trick Fortran into
believing that the address of the variable x was actually the status 
return code value (8 in this case).  

If anyone knows of a cleaner way of doing this, I would love to hear of
it.

Jay Hoeflinger

University of Illinois at Urbana-Champaign
Center for Supercomputing Research and Development

    UUCP:    {ihnp4,uunet,convex}!uiucuxc!uicsrd!hoefling
    ARPANET: hoefling%uicsrd@uxc.cso.uiuc.edu
    CSNET:   hoefling%uicsrd@uiuc.csnet
    BITNET:  hoefling@uicsrd.csrd.uiuc.edu

(217) 244-0288