Nightly test failures, builds with gcc/8.3.0 as host compiler: cc1plus: error with KokkosSparse::Impl::Sequential::TrsvWrap<...>::divide

[ndellingwood]

Nightly cuda/11.4 builds are failing to compile with error:

06:59:06 In member function 'void KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::CommonBlocked::divide(RangeMultiVectorType, const UBlock&, KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::lno_t, KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::lno_t) [with bool IsLower = false; bool Transpose = false; CrsMatrixType = KokkosSparse::Experimental::BsrMatrix<const Kokkos::complex<double>, const int, Kokkos::Device<Kokkos::Cuda, Kokkos::CudaSpace>, Kokkos::MemoryTraits<1>, const int>; DomainMultiVectorType = Kokkos::View<const Kokkos::complex<double>**, Kokkos::LayoutLeft, Kokkos::Device<Kokkos::Cuda, Kokkos::CudaSpace>, Kokkos::MemoryTraits<3> >; RangeMultiVectorType = Kokkos::View<Kokkos::complex<double>**, Kokkos::LayoutLeft, Kokkos::Device<Kokkos::Cuda, Kokkos::CudaSpace>, Kokkos::MemoryTraits<1> >]':
06:59:06 cc1plus: error: 'void* __builtin_memcpy(void*, const void*, long unsigned int)' source argument is the same as destination [-Werror=restrict]
06:59:06 [ 67%] Building CXX object CMakeFiles/kokkoskernels.dir/sparse/eti/generated_specializations_cpp/gauss_seidel_symbolic/Sparse_gauss_seidel_symbolic_eti_COMPLEX_DOUBLE_ORDINAL_INT_OFFSET_INT_LAYOUTLEFT_EXECSPACE_CUDA_MEMSPACE_CUDASPACE.cpp.o
06:59:06 [ 67%] Building CXX object CMakeFiles/kokkoskernels.dir/sparse/eti/generated_specializations_cpp/gauss_seidel_symbolic/Sparse_gauss_seidel_symbolic_eti_COMPLEX_DOUBLE_ORDINAL_INT_OFFSET_INT_LAYOUTLEFT_EXECSPACE_SERIAL_MEMSPACE_HOSTSPACE.cpp.o
06:59:06 [ 68%] Building CXX object CMakeFiles/kokkoskernels.dir/sparse/eti/generated_specializations_cpp/gauss_seidel_symbolic/Sparse_gauss_seidel_symbolic_eti_COMPLEX_DOUBLE_ORDINAL_INT_OFFSET_SIZE_T_LAYOUTLEFT_EXECSPACE_CUDA_MEMSPACE_CUDASPACE.cpp.o
06:59:06 In member function 'void KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::CommonBlocked::divide(RangeMultiVectorType, const UBlock&, KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::lno_t, KokkosSparse::Impl::Sequential::TrsvWrap<CrsMatrixType, DomainMultiVectorType, RangeMultiVectorType>::lno_t) [with bool IsLower = false; bool Transpose = false; CrsMatrixType = KokkosSparse::Experimental::BsrMatrix<const Kokkos::complex<double>, const int, Kokkos::Device<Kokkos::Serial, Kokkos::HostSpace>, Kokkos::MemoryTraits<1>, const int>; DomainMultiVectorType = Kokkos::View<const Kokkos::complex<double>**, Kokkos::LayoutLeft, Kokkos::Device<Kokkos::Serial, Kokkos::HostSpace>, Kokkos::MemoryTraits<3> >; RangeMultiVectorType = Kokkos::View<Kokkos::complex<double>**, Kokkos::LayoutLeft, Kokkos::Device<Kokkos::Serial, Kokkos::HostSpace>, Kokkos::MemoryTraits<1> >]':
06:59:06 cc1plus: error: 'void* __builtin_memcpy(void*, const void*, long unsigned int)' source argument is the same as destination [-Werror=restrict]
06:59:06 cc1plus: all warnings being treated as errors
06:59:06 make[2]: *** [CMakeFiles/kokkoskernels.dir/sparse/eti/generated_specializations_cpp/trsv/Sparse_trsv_eti_COMPLEX_DOUBLE_ORDINAL_INT_OFFSET_INT_LAYOUTLEFT_EXECSPACE_CUDA_MEMSPACE_CUDASPACE.cpp.o] Error 1

This follows merge of commits

Changes:
Git (git https://github.com/kokkos/kokkos-kernels.git)

Add support for BSR matrices to some trsv routines (#2104) ([detail](https://jenkins-son.sandia.gov/job/KokkosKernels_semsrhel7gpu01_cuda114_gcc830_cpp17/214/changes#detail))
Lapack - SVD: adding quick return when cuSOLVER is skipped (#2107) ([detail](https://jenkins-son.sandia.gov/job/KokkosKernels_semsrhel7gpu01_cuda114_gcc830_cpp17/214/changes#detail))

The failure in Trsv indicates this is related with the #2104 PR, @jgfouca can you take a look?

Reproducer (kokkos-dev-2):

module load sems-cmake sems-gcc/8.3.0 sems-cuda/11.4.2

$KOKKOSKERNELS_PATH/cm_generate_makefile.bash --with-devices=Cuda,Serial --arch=Volta70 --compiler=$KOKKOS_PATH/bin/nvcc_wrapper --cxxflags="-O3 -Wall -Wunused-parameter -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized " --cxxstandard="17" --kokkos-path=$KOKKOS_PATH --kokkoskernels-path=$KOKKOSKERNELS_PATH --with-scalars='double,complex_double' --with-ordinals=int --with-offsets=int,size_t --with-layouts=LayoutLeft --with-cuda-options=enable_lambda   --no-examples

make -j16

[ndellingwood]

@jgfouca The build failed with cuda/11.1+gcc/8.3.0 and in just a gcc/8.3.0 build as well, so it looks like gcc/8.3.0 may be the commonality

[ndellingwood]

Also, it is the -Werror that is tripping this as an error ([-Werror=restrict])

[ndellingwood]

@jgfouca thanks for fast fix in #2111 , nightlies are looking much better!