Issue with horiz_interp

[wfcooke]

flag to show that weight arrays have been allocated may not be set.
optional arguments to bilinear interp_new cannot be used if using horiz_interp_new.

I assume that with FMS2 the desired use of horiz_interp_new is as follows.

call horiz_interp_new ( Interp, T_grid%x, T_grid%y, lon_out, lat_out, interp_method="bilinear")
followed by a
call horiz_interp_del( Interp )

I'm running into an issue where the interpolation point seems to be right on the corner of the grid. When this happens I end up at
call mpp_error(FATAL, "horiz_interp_bilinear_mod: No solution found")

I'll assume this issue cropped up previously as there used to be a workaround for this with the new_search and no_crash_when_not_found arguments.
However I can't send these down through the horiz_interp_new interface.

I used to call
call horiz_interp_bilinear_new (Interp, T_grid%xDEG_TO_RAD, T_grid%yDEG_TO_RAD,&
& lon_out, lat_out, new_search=.true., no_crash_when_not_found=.true.)

If I use this format and call
call horiz_interp_bilinear_del( Interp )
I then run into an issue when it loops around and tries to call horiz_interp_new again. It gives an allocation error.
The reason is that even though
include/horiz_interp_bilinear.inc HORIZ_INTERP_BILINEAR_NEW_2D
allocates
Interp%HI_KIND_TYPE_%wti
etc.
it does not set the value of
Interp% HI_KIND_TYPE_ % is allocated.
That is done in include/horiz_interp.inc HORIZ_INTERP_NEW

But when I call horiz_interp_bilinear_del it checks the value of is_allocated and neither the Real4 or Real8 logical is true. So it doesn't deallocate the weights and when I call the next horiz_interp_new I find they are already allocated.

How to get around this?

1. Add the optional arguments to horiz_interp_new and pass them to horiz_interp_bilinear_new
   (this would allow me to use horz_interp_new/del)
   and
2. Set Interp%XXX%is_allocated to true in each _new routine (bilinear,spherical, conserve)
   ?