Very strange DGECON problem

[GIBBOGLE]

I am using Intel Fortran on Windows, and I have installed the 32-bit BLAS and LAPACK libraries. I am exploring the condition number of some matrices that I generate. A simple example matrix is 18x18. The '1' norm is calculated by DLANGE as 0.773E-01. Factorisation with DGETRF is successful, then DGECON returns condition number = Infinity. So it seems that this matrix is singular. BUT if I insert a statement to write out the array, write(*,'(6e12.3)') A, before the DLANGE statement, the condition number is determined to be 0.658E-03.

I have turned on the usual Fortran Runtime Error Checking, but it shows no problems.

????

Gib