Wrap assert include with extern C++

[waynemitchell]

Addresses #1191 and #1192

[waynemitchell]

It seems that Intel's oneAPI 2025 adds a C++ templated version of assert.h, which is causing compile issues due to assert.h being included within an extern "C" block in _hypre_utilities.h. I'm not that familiar with how extern blocks should be used, so I want to make sure that the fix here is not going to have unintended consequences.

[ivikiwi-1111]

It did help to remove those extern C linkage errors mentioned in #1192, but compilation still fails with following error:

icpx  -O3 -fsycl -fsycl-unnamed-lambda  -DHAVE_CONFIG_H -I.. -I. -I./.. -I./../utilities    -qmkl -I/include -DMKL_ILP64 -I/opt/intel/oneapi/mkl/2025.0/include       -c csr_matop_device.c -o csr_matop_device.obj
icpx: warning: treating 'c' input as 'c++' when -fsycl is used [-Wexpected-file-type]
csr_matop_device.c:186:101: error: too few arguments to function call, expected at least 2, have 1
  186 |       HYPRE_ONEMKL_CALL( oneapi::mkl::sparse::release_matrix_handle(&hypre_GpuMatDataMatHandle(data)) );
      |                          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                 ^
./../utilities/_hypre_utilities.hpp:817:7: note: expanded from macro 'HYPRE_ONEMKL_CALL'
  817 |       call;                                                                                  \
      |       ^~~~
/opt/intel/oneapi/mkl/2025.0/include/oneapi/mkl/spblas/sparse_structures.hpp:41:1: note: 'release_matrix_handle' declared here
   41 | release_matrix_handle(sycl::queue &queue,
      | ^                     ~~~~~~~~~~~~~~~~~~~
   42 |                       matrix_handle_t *p_spMat,
      |                       ~~~~~~~~~~~~~~~~~~~~~~~~~
   43 |                       const std::vector<sycl::event> &dependencies = {});
      |                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
csr_matop_device.c:3274:44: warning: 'trsv' is deprecated: Use oneapi::mkl::sparse::trsv(queue, ... , diag_val, /* alpha */ 1.0, ...) instead. [-Wdeprecated-declarations]
 3274 |    HYPRE_ONEMKL_CALL( oneapi::mkl::sparse::trsv( *hypre_HandleComputeStream(hypre_handle()),
      |                                            ^
/opt/intel/oneapi/mkl/2025.0/include/oneapi/mkl/spblas/sparse_operations.hpp:232:1: note: 'trsv' has been explicitly marked deprecated here
  232 | ONEMKL_DECLARE_DEPRECATED_SPARSE_TRSV(double);
      | ^
/opt/intel/oneapi/mkl/2025.0/include/oneapi/mkl/spblas/sparse_operations.hpp:221:7: note: expanded from macro 'ONEMKL_DECLARE_DEPRECATED_SPARSE_TRSV'
  221 |     [[deprecated("Use oneapi::mkl::sparse::trsv(queue, ... , diag_val, /* alpha */ 1.0, ...) instead.")]] \
      |       ^
1 warning and 1 error generated.
make[2]: *** [../config/Makefile.config:66: csr_matop_device.obj] Error 1
make[2]: Leaving directory '/home/test/hypre/src/seq_mv'
make[1]: *** [Makefile:92: all] Error 2
make[1]: Leaving directory '/home/test/hypre/src/seq_mv'
make: *** [Makefile:91: all] Error 1

[waynemitchell]

@ivikiwi-1111 , hmm... it seems there are many breaking changes in oneAPI 2025... rather than continuing to go back and forth with you on fixes, let me try to get oneAPI 2025 installed on a system that I have access to and try to get it compiling myself.

[victorapm]

It seems that Intel's oneAPI 2025 adds a C++ templated version of assert.h

Hi Wayne, your fix looks correct to me. Thanks!