Fix regression introduced by PR #1422

[wanbinchen-hnc]

This regression referred in issue #1442 occurs when we pass a struct component for an assumed-shape formal argument, e.g. geom%freq in

alpha = alpha_dyn_qho(alpha_0, omega, geom%freq, dalpha, grad_request_t())

where the third formal argument is declared as

type(quad_pt_t), intent(in) :: freq(0:)

The root cause is that we have lost the parent geom in the expression for the descriptor of geom%freq when using the function get_extent.

We fix this by manually constructing the subscript expression for the extent of geom%freq.

[bryanpkc]

LGTM. This solves the problem in #1442 but the libmbd build still fails, this time with a different error; I'll create a new issue. I'll give @shivaramaarao a chance to review before merging this.

[shivaramaarao]

looks good to me.

[pawosm-arm]

Hi @bryanpkc this commit fixes the building issue for me, yet bear in mind that I'm building libmbd from the 0.10.3 tag, not the top of the trunk. Also, I'm doing plain regular build, an attempted MPI build has failed like this:

[ 10%] Building Fortran object src/CMakeFiles/mbd.dir/mbd_mpi.F90.o
cd src && mpifort -DWITH_MPI -DWITH_SCALAPACK -Dmbd_EXPORTS -Isrc  -DWITH_MPI -g -fPIC   -c src/mbd_mpi.F90 -o CMakeFiles/mbd.dir/mbd_mpi.F90.o
F90-S-0000-Internal compiler error. incomplete interface file, missing AST    1038  (src/mbd_mpi.F90: 6)
F90-F-0004-incomplete IPA file, missing AST   (src/mbd_mpi.F90: 6)
F90/aarch64 Linux FlangArm F90  - 1.5 2017-05-01: compilation aborted

I don't know how often MPI build of this library is needed, for me the plain build is sufficient.

[bryanpkc]

Also, I'm doing plain regular build, an attempted MPI build has failed like this

Thanks for the heads-up. Let me try to look at this as well.