Closed ndellingwood closed 2 years ago
CI checks in #1405 show this too.
@e10harvey yeah, looks like all of the kokkos-kernels nightlies were impacted as well
Was there a merge in between #1399 i.e. SHA https://github.com/kokkos/kokkos-kernels/commit/2ced7df26e780cef8599151a2cf5a3a6cb1d7cfd and its target SHA https://github.com/kokkos/kokkos-kernels/commit/7ce1b8fef757e91919d543a0b278888f31ef91c5 (reported in the set of passing CI build information) that introduced the disruption? Trying to understand why it passed CI but we're seeing issues outside of that testing
@ndellingwood @e10harvey Sorry about that. I just found the problem - I copy-pasted the SpAdd files to do the graph color ETI, and forgot to change two include guard names. So as soon as both PRs were combined, those collided and the SpAdd eti_spec_avail stopped getting defined. That explains why PR testing didn't catch it. Will open a PR for the fix in a minute.
@brian-kelley thanks for the update and fast fix!
Nightly builds (first detected with Trilinos builds with kokkos and kokkos-kernels develop branches) are failing with linking errors following merge of #1399. Snip below is output from cuda/9.2 + gcc/7.2.0 build, though this is occurring across other nightly builds so far
@brian-kelley can you investigate?