fix: bug fix for obs_sort_module.F90.

[greent12]

This is a proposed fix for Issue #11

A conditional statement was added on line 3158 to check wether the pointer 'next' is associated or not before going into the following block of code. This bug fix was stated by the user "Alpha_su" in the obsgrid forum: https://forum.mmm.ucar.edu/threads/obsgrid-exe-fails-after-a-few-seconds.13474/

[brianreen]

Thanks for submitting this code to address issue #11. The original code inadvertently relies on short-circuiting of Boolean expressions, but since the Fortran standard does not require this, the code only works with certain compilers. While the fix you submitted would avoid the issue, it seems to me that it would be more straightforward to not add the ASSOCIATED check but rather to break the existing IF statement into two IF statements so that next is evaluated only when is_sounding is FALSE. Specifically, I would suggest:

        IF(.NOT. is_sounding) THEN
          IF(.NOT. eps_equal(next%meas%height%data, obs(i)%info%elevation, 1.)) THEN
            obs(i)%info%discard = .TRUE.
            PRINT '(A,A,A,A,F12.2,A,F12.2,A,A,A,A)','WARNING: Ob indicates is_sounding=.FALSE. ',&
             'but the ob height does ',&
             'not match the station elevation.  This ob will be omitted since Obsgrid assumes ',&
             'is_sounding=.FALSE. indicates a surface ob. Height = ', next%meas%height%data, &
             ', Elevation = ', obs(i)%info%elevation,', Ob ID = ',TRIM(ADJUSTL(obs(i)%location%id)),&
             ', Ob Time = ', obs(i)%valid_time%date_char
          ENDIF
        ENDIF