singular matrix in fmincon

I get an warning for singularity with calling fmincon. I would like to know what is causing this, however the error occurs in a p-file making it hard to investigate:

Warning: Matrix is close to singular or badly scaled. Results may be
inaccurate. RCOND =  1.553382e-17. 
> In backsolveSys (line 17)
  In solveKKTsystem (line 15)
  In computeTrialStep (line 66)
  In barrier (line 357)
  In fmincon (line 796)
  In Call_Optimize (line 115)

What is computed in backsolveSys?

Answer 1

[Matt J]

It looks like it is solving an ill-condition linear system of equations, similar to the following

     >> b=[1; 1e-16];   diag(b)\b
    Warning: Matrix is close to singular or badly scaled. Results may be inaccurate.
    RCOND =  1.000000e-16. 
    ans =
         1
         1