bug in Exodus_for with gfortran 10.2

[ecoon]

I was building SEACAS on a new Catalina OSX machine, and ran into the following issue when trying to build Exodus_for.

/Users/uec/code/ats/amanzi-tpls/build-ecoon-memory_management/relwithdebinfo/openmpi-4.0.4_1-clang-default/seacas/seacas-173a1e6-source/packages/seacas/libraries/exodus_for/src/addrwrap.F:1267:36:

 1140 |         CALL EXGSP4 (IDEXO4, IDESS, NSESS, NDESS, IERR4)
      |                                    2
......
 1267 |         CALL EXGSP4 (IDEXO4, IDESS, NSESS4, NDESS4, IERR4)
      |                                    1
Error: Type mismatch between actual argument at (1) and actual argument at (2) (INTEGER(4)/INTEGER(8)).
/Users/uec/code/ats/amanzi-tpls/build-ecoon-memory_management/relwithdebinfo/openmpi-4.0.4_1-clang-default/seacas/seacas-173a1e6-source/packages/seacas/libraries/exodus_for/src/addrwrap.F:3328:28:

  309 |         CALL EXGCOR4 (IDEXO4, XN, YN, ZN, IERR4)
      |                              2
......
 3328 |       call exgcor4 (idexo4, start_node_num, num_nodes, x_coor,
      |                            1
Error: Type mismatch between actual argument at (1) and actual argument at (2) (INTEGER(8)/REAL(8)).

There are a few others like this, so you'll see the full error log (seacas-build-error.log) attached below. It wasn't clear to me whether this was a generated file or not, but looking at the calls (to exgcor4 especially) it is clear that one uses real and one uses int. I believe that gfortran 10 has gotten more strict about standards -- this might compile with gfortran -std=legacy (I've had to deal with this in mpich already) so I'm willing to try that if it would be useful.

Note there is no urgency on this -- we don't actually use Exodus_for so I just turned it off -- but figured I would let you know. Let me know if I can help testing a fix or anything.

Compilers are:

gfortran 10.2.0 (Homebrew's default gcc)
clang-1103.0.32.62 (system default)
openmpi 4.0.4_1 (Homebrew's default openmpi)

Included logs of configure and build for completeness:

seacas-configure-out.log
seacas-configure-err.log
seacas-build-out.log
seacas-build-err.log

[ecoon]

For what it is worth, it looks like this should fix it (adding -fallow-argument-mismatch to FFLAGS and FCFLAGS).

Unidata/netcdf-fortran#212

[gsjaardema]

It looks like you are using an older version of this file. Many gfortran-10 errors/warnings were fixed a few months ago when that version was released.

Is your source code from a current clone of the seacas repository?

[ecoon]

Oh, good, my mistake. We are using a slightly older version. Will update, thanks!