OMPI build fails when cuda is enabled

[AboorvaDevarajan]

Background information

What version of Open MPI are you using? (e.g., v3.0.5, v4.0.2, git branch name and hash, etc.)

ompi - master

Describe how Open MPI was installed (e.g., from a source/distribution tarball, from a git clone, from an operating system distribution package, etc.)

git clone

If you are building/installing from a git clone, please copy-n-paste the output from git submodule status.

ompi]$ git submodule status
 074b437c236852a898009573a2883dd9526207f1 3rd-party/openpmix (v1.1.3-2908-g074b437)
 eba2595e56b559ba3144f6e5e13453eca6642c24 3rd-party/prrte (dev-31073-geba2595)

Please describe the system on which you are running

* Operating system/version: RH8.3
* Computer hardware:  ppc64le

---

Details of the problem

$ git clone --recursive https://github.com/open-mpi/ompi.git
$ cd ompi
$ ./autogen.pl 
$ ./configure --disable-man-pages --enable-mca-no-build=btl-uct --enable-mpi1-compatibility     --with-cuda=/usr/local/cuda 
$ make -j 20

  CCLD     libopen-pal.la
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o): In function `mca_common_cuda_free':
common_cuda.c:(.text+0x300): multiple definition of `mca_common_cuda_free'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):common_cuda.c:(.text+0x300): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o): In function `opal_cuda_memcpy':
common_cuda.c:(.text+0x3a0): multiple definition of `opal_cuda_memcpy'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):common_cuda.c:(.text+0x3a0): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2ec): multiple definition of `opal_cuda_verbose'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2ec): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o): In function `mca_common_cuda_malloc':
common_cuda.c:(.text+0x8b0): multiple definition of `mca_common_cuda_malloc'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):common_cuda.c:(.text+0x8b0): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x328): multiple definition of `mca_common_cuda_enabled'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x328): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.data+0x40): multiple definition of `cuda_event_max'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.data+0x40): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x320): multiple definition of `cuda_event_ipc_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x320): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x308): multiple definition of `cuda_event_ipc_frag_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x308): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x318): multiple definition of `cuda_event_dtoh_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x318): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x300): multiple definition of `cuda_event_dtoh_frag_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x300): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x310): multiple definition of `cuda_event_htod_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x310): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2f8): multiple definition of `cuda_event_htod_frag_array'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2f8): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o): In function `mca_common_cuda_register_mca_variables':
common_cuda.c:(.text+0x15e0): multiple definition of `mca_common_cuda_register_mca_variables'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):common_cuda.c:(.text+0x15e0): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2f0): multiple definition of `libcuda_handle'
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o):(.bss+0x2f0): first defined here
mca/common/cuda/.libs/libmca_common_cuda_noinst.a(common_cuda.o): In function `mca_common_cuda_fini':
common_cuda.c:(.text+0x18b0): multiple definition of `mca_common_cuda_fini'

CUDA builds are frequently broken, I guess we can add cuda build checks in CI to avoid CUDA build issues.

[awlauria]

Thanks @AboorvaDevarajan . Closing as a dup of #8656.