matrix is close to singular or badly scaled, what do you interpret by this error.

I am using MatLab solver 'fmincon' for an optimization problem. Everything seems to be working good except I got one error warning:

 matrix is close to singular or badly scaled. Results may be inaccurate.
 RCOND =  1.722247e-17.

My objective function is following:

 objective = 0.5*sum(log(sigma2(ind))) + 0.5*(y-yhat)*(S\(y-yhat)');

where S is the matrix with singularity problems. My question is that how do I interpret the final solution If I got this warning. Should I discard this solution or keep it? What are probable solution for improving this situation. I have tried the suggestions on other questions such as using backslash instead of 'inv'. But nothing seems to work.

[Matt J]

Why are you using a singular S? I would think that fixing the singularity problems would be the best solution.