Address compile issues in ExodusProblemGenerator GCC10

[gahansen]

I ended up having some additional compilation errors using GCC10 on RHEL7 after issue #589 was resolved. Here is a quick PR that gets the code to compile for me. I made two changes in this PR:

1. Address incorrect number of arguments in ArborX::minMax() and sendAcrossNetwork() calls.

2. Address compilation error from using Kokkos unmanaged views (hitting static_assert on line 538 of ArborX_DetailsUtils.hpp

I'm not certain that I did the correct thing here, but these changes do point at the issues that I saw so please feel free to give me feedback or do something different here.

All of the tests run for me except for:

    DataTransferKit_Interpolation_MPI_4

Here are the errors I see in that test:

Running unit tests ...

0. Interpolation_DeviceTypeKokkos_Compat_KokkosSerialWrapperNode_one_topo_one_fe_three_dim_UnitTest ... [Passed] (0.00168 sec)
1. Interpolation_DeviceTypeKokkos_Compat_KokkosSerialWrapperNode_two_topo_two_dim_UnitTest ... [Passed] (0.00067 sec)
2. Interpolation_DeviceTypeKokkos_Compat_KokkosSerialWrapperNode_one_topo_one_fe_three_dim_hdiv_UnitTest ...
 Y.extent( 0 ) = 5 == ref_size = 5 : passed

 Check: rel_err(ref_sol[i] + dim * j, Y_host( i, j ))
        = rel_err(0, 0) = 0
          <= tol = 1e-06 : passed

Check: rel_err(ref_sol[i] + dim * j, Y_host( i, j ))
        = rel_err(-0.125, -0.5) = 0.75
          <= tol = 1e-06 : FAILED

 Check: rel_err(ref_sol[i] + dim * j, Y_host( i, j ))
        = rel_err(-0.2, -0.8) = 0.75
          <= tol = 1e-06 : FAILED

 Check: rel_err(ref_sol[i] + dim * j, Y_host( i, j ))
        = rel_err(-0.25, -1) = 0.75
          <= tol = 1e-06 : FAILED

 Check: rel_err(ref_sol[i] + dim * j, Y_host( i, j ))
        = rel_err(-0.45, -1.8) = 0.75
          <= tol = 1e-06 : FAILED
NOTE: Unit test failed on processes = {0, 1}
 (rerun with --output-to-root-rank-only=<procID> to see output
 from individual processes where the unit test is failing!)
 [FAILED]  (0.000648 sec) Interpolation_DeviceTypeKokkos_Compat_KokkosSerialWrapperNode_one_topo_one_fe_three_dim_hdiv_UnitTest
 Location: /projects/albany/src/Trilinos/DataTransferKit/packages/Discretization/test/tstInterpolation.cpp:371

[gahansen]

Just force-pushed some formatting changes suggested by Travis.

[gahansen]

Sounds great @aprokop ! Once the commit goes in that addresses https://github.com/arborx/ArborX/issues/419, I'll update this PR. One question, should I cancel this one and re-submit the ExecutionSpace{} mods at that time, or should I just update this PR (#592) given those changes at that time?

[aprokop]

Once the commit goes in that addresses arborx/ArborX#419, I'll update this PR

We'll first need to merge not only the PR in ArborX, but also to update ArborX in DTK.

One question, should I cancel this one and re-submit the ExecutionSpace{} mods at that time, or should I just update this PR (#592) given those changes at that time?

Just update this PR.

[gahansen]

Just updated PR to reflect @aprokop comments above.

[aprokop]

@gahansen Can you please rebase this PR on top of master and see if it fixes the problem?

[gahansen]

@aprokop - I rebased the PR and made several changes to address minor changes in ArborX (DistributedTree --> DistributedSearchTree). Still see the static_assert compile error coming from the use of unmanaged views (error log attached).

CompileError_11032020.txt

[aprokop]

@gahansen If you had to do a change to fix the DistributedSearchTree, this means that when you updated you did not pull the latest submodules. Can you please make sure your submodule in packages/Search is up to date?

[gahansen]

Sorry @aprokop - submodules are new to me but I think I have done this properly now. I rebased, updated the submodules, and pushed this commit to my fork for the automated tester. These changes did fix the compile errors I was seeing on the GCC10 machine. All tests are clean with the exception of "DataTransferKit_Interpolation_MPI_4" that was noted at the beginning of the PR.

[aprokop]

@gahansen The CI is failing now, while it was passing before. It seems that together with ArborX submodule you have also updated the TriBITS submodule. I wonder if the errors are coming from that upgrade. Could you please revert it so that you patch only contains changes to the files in DTK?

[gahansen]

Thanks @aprokop ! I think the third time is the charm. It looks like the checks passed this time. Thanks for all your help with this, I'll know how to do it next time!

[aprokop]

@gahansen Thanks!