roystgnr

[mesh_smoother_vsmoother.h](https://github.com/libMesh/libmesh/blob/devel/include/mesh/mesh_smoother_vsmoother.h) [mesh_smoother_vsmoother.C](https://github.com/libMesh/libmesh/blob/devel/src/mesh/mesh_smoother_vsmoother.C) Nice fancy mathematics in the underlying thesis+papers, immunity to the failure cases you can see with Laplace, but the code is kind of orphaned; the original author didn't...

Regressed code, probably. We've never had exodiff tests in libMesh and in hindsight we really should have...

> You can run the unit_tests-opt executable directly and pass --re Foo on the command line I kind of like `LIBMESH_OPTIONS='--re Foo' make check -C tests/`, personally. Runs everything in...

Does it fail repeatedly in parallel in the unit test, or intermittently? Does it fail in every parallel CI recipe, or just some subset (e.g. DistributedMesh) of them?

Looks like all your unit test invocations are hardcoding `DistributedMesh`? You should add a `ReplicatedMesh` invocation of each and see if those work (and keep the extra invocations, for better...

It's possible to see differences even without partitioning, depending on how the meshes are generated. For any mesh generation code that adds new unpartitioned objects but doesn't impose an explicit...

> Is anything else needed here? Sorry to vanish on you last week. My one concern is the failures on partitioned meshes. Running on a distributed mesh is kind of...

I fear the odds are nvc++ is "right", at least in the sense that it's still standards-conforming. IIRC the C++ standard doesn't give much in the way of guarantees for...

I'm just being too paranoid and misremembering things. C++ requires full declarations for most templated classes' arguments, but [`unique_ptr` and `shared_ptr` are exceptions](http://howardhinnant.github.io/incomplete.html).

It's definitely a bug, not a missing feature; we do have an error_vector.C case that's intended to handle Exodus filenames. On the other hand, IIRC I wrote that code, and...