Erik Faulhaber
Erik Faulhaber
When running parallel coverage jobs and starting the next run before the previous one is finished, the reports of the previous run show up on the newest one instead of...
Examples like `tree_1d_dgsem/elixir_euler_shockcapturing.jl` and some others do not work in macOS ARM builds of Julia. They fail with this error (stack trace truncated): ``` ERROR: cfunction: closures are not supported...
We implemented a new `SemidiscretizationCoupled` which bundles multiple semidiscretizations and runs them in parallel with synchronized time steps. These semidiscretizations can be coupled using the new `BoundaryConditionCoupled`. When this boundary...
Especially in 3D simulations with the compressible Euler equations (like `examples/structured_3d_dgsem/elixir_euler_source_terms_nonperiodic_curved.jl`), the boundary flux quickly becomes the most expensive part of the simulation when using multiple threads. Right now, the...
The generated VTK files in the macOS CI tests of Trixi2Vtk (only the `P4estMesh` tests) have different hashes than on Linux and Windows. I made the tests pass like this,...
While the `P4estMesh` supports local mesh refinement and AMR, this can be problematic for complex simulations like the baroclinic instability. There, well-balanced errors make the simulation on a vertically non-conforming...
In #954, coverage and non-coverage tests have been split up in CI to allow running simpler tests with coverage, and reserve the more complex simulations for the tests without coverage....
Currently, Trixi2Vtk tests are comparing hashes of the generated VTK files. This doesn't work reliably, especially not when testing on different operating systems.
A convergence test like ```julia convergence_test("examples/structured_2d_dgsem/elixir_advection_rotated.jl", 3) ``` fails with ``` ERROR: MethodError: no method matching (::Main.TrixiExtensionAdvectionRotated.InitialConditionConvergenceTestRotated)(::SVector{2, Float64}, ::Float64, ::LinearScalarAdvectionEquation2D{Float64}) The applicable method may be too new: running in world...
We don't want to make individual parallel jobs show as failed if the total coverage didn't decrease. As advised in #1505, we removed the `flag-name` (as opposed to the docs...