Integrating a function that has a codistributed array as a parameter

I am receiving the following error message:

Error using integralCalc/iterateArrayValued (line 159)
    Inputs must be floats, namely single or double.

My code calls

    numf=integral(@fitest,-1,1,'ArrayValued',true,...
                 'RelTol',1e-4,'AbsTol',1e-4);

where fitest is defined as follows

function y=fitest(t)
        ft1=exp(-1i*t.*x);
        ftk = (0.6667)*( (1-t.^2).^3 + (1-t.^2).^3 )...
              -0.1667*( (1-(2*t).^2).^3 + (1-(2*t).^2).^3 );       
        sigUF2 = sigmaUF2*ones(1,n); 
        num = exp(-0.5*(1/h^2)*(t.^2)*(sigUF2+kron(ones(xlen,1),sigma2')));
        den_k = exp(-0.5*(1/h^2)*(t.^2)*sigma2);
        den = sum(den_k, 1);
        y=real((1/(2*pi))*ft1.*ftk.*num./den);        
    end

The input t is a scalar, and the parameter x in the function is a matrix. The code works when run normally, but I am now trying to process larger datasets, and to do I am using the parallel processing command spmd and making x into a codistributed array. I now get the error message above. Can anyone confirm that this is due to the codistributed array parameter? Any suggestions on how to get around this?

[Matt J]

Seems doubtful that codistribution would be the reason. Show us how x is created. Also show us how it finds its way into the workspace of fitest() seeing as fitest only takes t as an input argument.