the usage of a std::unique_ptr in mpi_gather_test results in init after finalize

#include "dca/function/domains/local_domain.hpp"

#include "dca/testing/minimalist_printer.hpp"

std::unique_ptr<dca::parallel::MPIConcurrency> concurrency;

//See below, life cycle issue with MPI

//std::unique_ptr<dca::parallel::MPIConcurrency> concurrency;

dca::parallel::MPIConcurrency* concurrency_ptr;

TEST(MPIGatherTest, GatherLocalDmn) {

  using Dmn1 = dca::func::dmn<4>;

    val2[i] = i;

  Dmn2::set_elements(val2);

  dca::parallel::MPIGang gang(*concurrency, 4);

  dca::parallel::MPIGang gang(*concurrency_ptr, 4);

  LocalDmn::initialize(gang);

      local_f(i1, i2) = Dmn1::get_elements()[i1] * LocalDmn::get_elements()[i2];

  concurrency->gather(local_f, f, gang);

  concurrency_ptr->gather(local_f, f, gang);

  for (int i2 = 0; i2 < Dmn2::get_size(); ++i2)

    for (int i1 = 0; i1 < Dmn1::get_size(); ++i1)

int main(int argc, char** argv) {

  int result = 0;

  concurrency = std::make_unique<dca::parallel::MPIConcurrency>(argc, argv);

  // This results in a copy constructor beging called at somepoint,  resulting in an MPI_INIT after the finalize.

  // concurrency = std::make_unique<dca::parallel::MPIConcurrency>(argc, argv);

  concurrency_ptr = new dca::parallel::MPIConcurrency(argc, argv);

  ::testing::InitGoogleTest(&argc, argv);

  ::testing::TestEventListeners& listeners = ::testing::UnitTest::GetInstance()->listeners();

  if (concurrency->id() != 0) {

  if (concurrency_ptr->id() != 0) {

    delete listeners.Release(listeners.default_result_printer());

    listeners.Append(new dca::testing::MinimalistPrinter);

  }

  result = RUN_ALL_TESTS();

  delete concurrency_ptr;

  return result;

}