[Linux 64 bit] MEX seg fault on simple FORTRAN file

Hi,

I've been following the documentation on mex routines and am trying to compile a simple timestwo.F example, located here: matlabroot/extern/examples/refbook/.

This gateway routine is listed here:

===================== begin timestwo.F =============================

#include "fintrf.h"
C
#if 0
C     
C     timestwo.F
C     .F file needs to be preprocessed to generate .for equivalent
C     
#endif
C     
C     timestwo.f
C
C     Computational function that takes a scalar and doubles it.

C     This is a MEX-file for MATLAB.
C     Copyright 1984-2009 The MathWorks, Inc.
C     $Revision: 1.12.2.7 $

        subroutine mexFunction(nlhs, plhs, nrhs, prhs)
  C-----------------------------------------------------------------------
  C
        mwpointer plhs(*), prhs(*)
        mwpointer mxGetPr, mxCreateDoubleMatrix
        mwpointer x_pr, y_pr
  C-----------------------------------------------------------------------
  C

        integer nlhs, nrhs
        integer mxIsNumeric
        mwsize mxGetM, mxGetN
        mwsize m, n, size
        real*8  x, y

C     Check for proper number of arguments. 
      if(nrhs .ne. 1) then
         call mexErrMsgTxt('One input required.')
      elseif(nlhs .gt. 1) then
         call mexErrMsgTxt('Too many output arguments.')
      endif

C     Get the size of the input array.
      m = mxGetM(prhs(1))
      n = mxGetN(prhs(1))

        size = m*n

C     Check to insure the input is a number.
      if(mxIsNumeric(prhs(1)) .eq. 0) then
         call mexErrMsgTxt('Input must be a number.')
      endif

C     Create matrix for the return argument.
      plhs(1) = mxCreateDoubleMatrix(m,n,0)
      x_pr = mxGetPr(prhs(1))
      y_pr = mxGetPr(plhs(1))
      call mxCopyPtrToReal8(x_pr,x,size)

C     Call the computational subroutine.
      call timestwo(y, x)

C     Load the data into y_pr, which is the output to MATLAB
      call mxCopyReal8ToPtr(y,y_pr,size)     

        return
        end

===================== end timestwo.F ===============================

here's the (verbose) output when I compile:

$ mex -fortran -v timestwo.F timestwo_sub.for

===================== begin compilation output ====================

-> mexopts.sh sourced from directory (DIR = .)
   FILE = /mex/examples/mexopts.sh
----------------------------------------------------------------
->    MATLAB                = /usr/local/bin/Matlab
->    CC                    = gcc-4.2
->    CC flags:
         CFLAGS             = -fPIC -fno-omit-frame-pointer -ansi -D_GNU_SOURCE -pthread -fexceptions
         CDEBUGFLAGS        = -g
         COPTIMFLAGS        = -O -DNDEBUG
         CLIBS              = -Wl,-rpath-link,/usr/local/bin/Matlab/bin/glnxa64 -L/usr/local/bin/Matlab/bin/glnxa64 -lmx -lmex -lmat -lm -lstdc++
         arguments          =  -DMX_COMPAT_32
->    CXX                   = g++
->    CXX flags:
         CXXFLAGS           = -fPIC -fno-omit-frame-pointer -ansi -D_GNU_SOURCE -pthread 
         CXXDEBUGFLAGS      = -g
         CXXOPTIMFLAGS      = -O -DNDEBUG
         CXXLIBS            = -Wl,-rpath-link,/usr/local/bin/Matlab/bin/glnxa64 -L/usr/local/bin/Matlab/bin/glnxa64 -lmx -lmex -lmat -lm
         arguments          =  -DMX_COMPAT_32
->    FC                    = gfortran
->    FC flags:
         FFLAGS             = -fPIC -fno-omit-frame-pointer -fexceptions -fno-automatic
         FDEBUGFLAGS        = -g
         FOPTIMFLAGS        = -O
         FLIBS              = -Wl,-rpath-link,/usr/local/bin/Matlab/bin/glnxa64 -L/usr/local/bin/Matlab/bin/glnxa64 -lmx -lmex -lmat -lm
         arguments          =  -DMX_COMPAT_32
->    LD                    = gfortran
->    Link flags:
         LDFLAGS            = -pthread -shared -Wl,--version-script,/usr/local/bin/Matlab/extern/lib/glnxa64/fexport.map -Wl,--no-undefined
         LDDEBUGFLAGS       = -g
         LDOPTIMFLAGS       = -O
         LDEXTENSION        = .mexa64
         arguments          = 
->    LDCXX                 = 
->    Link flags:
         LDCXXFLAGS         = 
         LDCXXDEBUGFLAGS    = 
         LDCXXOPTIMFLAGS    = 
         LDCXXEXTENSION     = 
         arguments          = 
----------------------------------------------------------------

-> gfortran -c  -I/usr/local/bin/Matlab/extern/include -I/usr/local/bin/Matlab/simulink/include -fPIC -fno-omit-frame-pointer -fexceptions -fno-automatic  -DMX_COMPAT_32 -O  "timestwo.F"

-> gfortran -c  -I/usr/local/bin/Matlab/extern/include -I/usr/local/bin/Matlab/simulink/include -fPIC -fno-omit-frame-pointer -fexceptions -fno-automatic  -DMX_COMPAT_32 -O  "timestwo_sub.for"

-> gfortran -O -pthread -shared -Wl,--version-script,/usr/local/bin/Matlab/extern/lib/glnxa64/fexport.map -Wl,--no-undefined -o  "timestwo.mexa64"  timestwo.o timestwo_sub.o  -Wl,-rpath-link,/usr/local/bin/Matlab/bin/glnxa64 -L/usr/local/bin/Matlab/bin/glnxa64 -lmx -lmex -lmat -lm

===================== end compilation output ======================

my compiler options were as follows:

            FC='gfortran'
            FFLAGS='-fPIC -fno-omit-frame-pointer -fexceptions -fno-automatic'
            FLIBS="$RPATH $MLIBS -lm"
            FOPTIMFLAGS='-O'
            FDEBUGFLAGS='-g'

when I start up MATLAB and run the routine, it seg faults. This does not happen with the g95 compiler, and I can't figure out why this is the case. Does anyone have any suggestions?

Thanks,

Tim

Answer 1

[James Tursa]

The timestwo.f example only works for scalar inputs. How are you calling it? (Side Note: size is a Fortran intrinsic function name, so I would pick a different variable name in the code)

[Tim D]

I'm using a scalar input to drive the function. Here's what I use:

x = 2;

y = timestwo(x);

[Tim D]

Also, I should add that I have tried this on two separate 64 bit machines (both running Linux) and both have seg faulted. If I compile on a 32 bit machine and take away the "-fPIC" option, then it runs correctly in MATLAB.

Thanks again,

Tim