Daniel Arndt
A build with modules shows ``` real 3m11.304s user 46m50.805s sys 3m1.869s ``` and without modules ``` real 3m25.149s user 52m40.656s sys 2m57.070s ```
I guess you are referring to https://github.com/kokkos/kokkos/blob/9642f8763f4639b9932eb960c2e3e5fe25cf0156/core/src/Kokkos_CopyViews.hpp#L314-L316.
I can reproduce with g++-15 but not with clang when adding the warning flag explicitly. We are currently enforcing - Wall - Wextra - Wextra-semi - Wunused-parameter - Wshadow -...
Apparently, the `Kokkos::Array` constructor is invoked and aggregate initialization complains with gcc (still) even if there is only one member when we omit a second pair of braces. Ultimately, the...
https://godbolt.org/z/Tx41c9Ps1 shows that the C-style array constructor is used but we still get warnings for the std::array one. Adding a `std::initializer_list` constructor would avoid the warning.
Would you mind rebasing now that we have https://github.com/kokkos/kokkos/pull/8732 merged?
> **CUDA-11.7-NVCC**: [issues with mdspan](https://cloud1.cees.ornl.gov/jenkins-ci/job/Kokkos/job/PR-8728/5/pipeline-overview/?selected-node=460) I expected https://github.com/kokkos/kokkos/pull/8193 to take care of that warning (which should really be taken care of on the `mdspan`-side).
> in the SYCL builds on HPSF CI That was present in https://github.com/kokkos/kokkos/pull/8732 (https://gitlab.spack.io/kokkos/kokkos/-/pipelines/1337010) already.
> Could it be related to the fact that this CI run uses Makefile? Ah, yes, that would do it.