Noticed while testing #375, but unrelated to it. I am seeing baseline comparison fails for ocean/planar/manufactured_solution/convergence_both/del4, but only when more MPI tasks are used for the PR tests than were used for the baseline. Turns out this also happens on develop.
I already have a fix. We need to exchange state halos at every RK stage.
Noticed while testing #375, but unrelated to it. I am seeing baseline comparison fails for
ocean/planar/manufactured_solution/convergence_both/del4, but only when more MPI tasks are used for the PR tests than were used for the baseline. Turns out this also happens ondevelop.I already have a fix. We need to exchange state halos at every RK stage.