Shape function gradients are incorrectly assumed to be constant within wedge elements

[dseyler]

Description

In Code/Source/solver/nn_elem_props.h, the flag mshType::lshpF is set to indicate whether shape functions are linear (including wedge elements). This flag is then used in several places throughout the code to decide whether the derivative of the shape functions is constant across the element or needs to be computed per Gauss point. This is true for LIN1, TRI3, and TET4 elements but not for WDG elements.

for (int g = 0; g < lM.nG; g++) {
  if (g == 0 || !lM.lShpF) {
    nn::gnn(eNoN, nsd, nsd, Nx_g, xl, Nx, Jac, ksix);   // Nx and Jac
  }
  double w = lM.w(g) * Jac;
  ...
}

This bug appears 35 times in the codebase, and there are currently no wedge element test cases to catch it.

file	line number
post.cpp	263, 433, 581, 678, 762, 921, 1775, 2060
fluid.cpp	634, 642, 706, 715
fsi.cpp	181, 189, 281, 290
all_fun.cpp	390, 589, 656
stokes.cpp	137, 169, 178
ustruct.cpp	330, 363, 371
cep.cpp	490
cmm.cpp	876
eq_assem.cpp	248
heatf.cpp	112
heats.cpp	89
Integrator.cpp	1171
l_elas.cpp	112
mesh.cpp	114
set_bc.cpp	1416
sv_struct.cpp	297

Reproduction

There are no wedge element test cases to reproduce this bug.

Expected behavior

Either a new flag should be added to indicate that an element has a constant shape function gradient or these locations in the code should be gated on element type rather than lshpF

Additional context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct and Contributing Guidelines

[ktbolt]

@dseyler This is a C++ conversion error made by yours truly; the Fortran has lM%lShpF = .FALSE.

[ktbolt]

Note that we should be using wedge elements for boundary layer meshing.

[dseyler]

@ktbolt Good to know! Sounds like the solution is just a one line change setting lShpF to false for wedge elements, then?

[ktbolt]

@dseyler Just the one line change; adding a test will perhaps uncover other bugs.