TRI7 center node adjustment and TRI6/7 degeneracy checks after midpoint relocation

[miaoyinb]

closes #4480

[miaoyinb]

As this PR moves TRI7 center node, MOOSE's meshgenerators/xy_delaunay_generator.xydelaunay/quadratic_tri7 test needs to be regolded.

[moosebuild]

Job Coverage, step Generate coverage on 0c245e1 wanted to post the following:

Coverage

	304e97	#4481 0c245e
	Total	Total	+/-	New
Rate	65.61%	65.62%	+0.02%	100.00%
Hits	78653	78711	+58	58
Misses	41232	41232	-	0

Diff coverage report

Full coverage report

This comment will be updated on new commits.

[roystgnr]

As a general rule "midpoint goes at the centroid" isn't always right, since you can do some clever things with boundary layers by using slightly-off-center higher order nodes in the grading, but in this context I think this is exactly the right thing to do.

If you don't want to create a big "have to get the submodule updated to regold" vs "can't merge the PR without breaking a ton of recipes" chicken-and-egg problem, the general way to do that is to add new behavior as an option before we change behavior by default. Default the new behavior to "off", then after MOOSE is updated turn it on there and regold, then change the default to "on".

[miaoyinb]

Good point, I will make it optional.

[moosebuild]

Job Distributed make check sweep (odd) on 4659afd : invalidated by @miaoyinb

[miaoyinb]

Added the option and made the original approach default.

Are the two failed tests related? Thanks @roystgnr

[roystgnr]

Sorry, I've been out on vacation for a couple weeks.

The failed tests are definitely related. I can reproduce the failures with a configure --enable-distmesh build of libMesh, with LIBMESH_RUN='mpirun -np 3' LIBMESH_OPTIONS='--re Poly2TriEdge' make -j 65 check -C tests/ (substitute your own -j argument there). I haven't dug more deeply than identifying the failing test, but it looks like rank 0 thinks it's passing but one of the other two ranks is returning a non-0 exit code, probably failing a CppUnit assertion.

[miaoyinb]

Sorry, I've been out on vacation for a couple weeks.

The failed tests are definitely related. I can reproduce the failures with a configure --enable-distmesh build of libMesh, with LIBMESH_RUN='mpirun -np 3' LIBMESH_OPTIONS='--re Poly2TriEdge' make -j 65 check -C tests/ (substitute your own -j argument there). I haven't dug more deeply than identifying the failing test, but it looks like rank 0 thinks it's passing but one of the other two ranks is returning a non-0 exit code, probably failing a CppUnit assertion.

Thanks, let me look into this.