Re: Problem with mex file

[frank wang]

Hi, John,

 

It is strange. The problem codes are:

  if (nRhs > 1)
  {
   px = mxGetPr(pRhs[1]);
   for (n=0; n<mxGetM(pRhs[1])*mxGetN(pRhs[1]); n++)
    pScl[n] = (unsigned)px[n];
   px = mxGetPr(pRhs[2]);
   pScl[5] = (unsigned)px[0];
   px = mxGetPr(pRhs[3]);
   pScl[6] = (unsigned)px[0];
  }
where pScl is initialized to {0 0 0 0 0 0}.

 

The following commands work in Matlab:

y=fft_mx1(rand(1024,1));
y=fft_mx1(rand(1024,1),0,0,0);
y=fft_mx1(rand(1024,1),[0 0]);
y=fft_mx1(rand(1024,1),[0 0,0]);
y=fft_mx1(rand(1024,1),[0 0,0],0);
y=fft_mx1(rand(1024,1),[0 0,0],0,0);
 

However, the following two does not:

 

y=fft_mx1(rand(1024,1),0);
y=fft_mx1(rand(1024,1),0,0);

 

I just want to let you know this and to make Octave even better.

 

Thanks

 

Frank
 

On 9/8/06, John W. Eaton <address@hidden> wrote:

On  8-Sep-2006, frank wang wrote:

| I use the mkoctfile --mex to compile the c code and it works great. Thanks.
|
| However, when I run the code with fewer input, it will cause segment fault.
| It is caused by function mxGetPr. For example, if in the c code, I have
| px=mxGetPr(pRhs[3]). But I only passed 2 parameters, this will cause segment
| fault. I fix this by changing my c code. But I just want to point it out
| since in Matlab it works.

Why should it work?  If you pass two parameters, then the values are
in prhs[0] and prhs[1].  Anything beyond that is going to be garbage
anyway.  Maybe it is just dumb luck that it works in Matlab?

jwe