🚀 go-pugleaf

	We have some SUNs (4/670 and SS2) that I have recently installed
SUN's 2.0.1 version C and fortran on.  The SUNs use SunOS 4.1.3 and MATLAB
4.0A.

	We are having a terrible time gettting the MEX utility to work.
[It had worked before installing the new compilers.]  The file, .mexrc.sh
in the matlab/bin subdirectory has a bug that prevents proper linking.
[BUG 1].  I customized a user's .mexrc.sh file as follows to try to
overcome that problem.  This is the line I changed:

            LD_FCLIBS="/usr/lang/SC2.0.1/libF77.a /usr/lang/SC2.0.1/libm.a 
                       /usr/lib/libc.a"

	    It was: LD_FCLIBS=

This allows a sample mex (fortran) file to be built with fmex, but when
we try to run in MATLAB, we get:

Undefined symbol '___builtin_saveregs' in file /home/updike/usr6/nixon/stdfilstuff/testformex.mex4.
Undefined symbol '___f_clos' in file /home/updike/usr6/nixon/stdfilstuff/testformex.mex4.
Undefined symbol '_flush_' in file /home/updike/usr6/nixon/stdfilstuff/testformex.mex4.
Undefined symbol '_f_exit' in file /home/updike/usr6/nixon/stdfilstuff/testformex.mex4.
Undefined symbol '___base_conversion_write_only_do' in file /home/updike/usr6/nixon/stdfilstuff/testformex.mex4.
??? Invalid MEX-file



___builtin_saveregs: This ugly beast is apparently only defined in a gcc
library.  Arrrggh!

	A second BUG is that I changed the CC= line in .mexrc.sh to
CC=acc hoping to use the ANSI compiler.  This line is not picked up
by fmex (or cmex).

	I have a call into The Mathworks, several hours old, but so far
no response.  [Frustration......]  Is there a policy at the Mathworks to 
return calls within a specified time limit?

      Has anyone seen this problem before?  If so, could you comment on how you
solved it?

	Thanks in advance for all responses.

ric steinberger
ric@updike.sri.com