Solution derivative for fault face kernels has incorrect values

[baagaard-usgs]

Describe the bug

The derivative of the solution, s_x, for the fault face kernels has incorrect values. The offsets of s_x (sOff_x) seem to be correct as they are set to pass just the values for one side of the fault. The values for the derivatives do not seem to be matching up with the offsets.

To Reproduce

PETSc branch: knepley/fix-plex-hybrid-integration
PyLith branch: baagaard/feature-mmstests-dynamic-prescribed-slip

[baagaard-usgs]

@knepley Need to add geometry for neighboring cells to hybrid residual and Jacobian calculation.

• Need to get cell on each side of the fault.
• Make array neighbors and get geometry for neighbors (cell).
• Pass both geometry of faces and neighbors to residual and Jacobian calculations.
• Need to use neighbor geometry for negative and positive sides of the fault calculations (cell Jacobian, etc).

[baagaard-usgs]

@knepley Here is an update on the status of the fault integration for dynamic prescribed slip.

Describe the bug

The derivative of the solution, s_x, for the fault face kernels has incorrect values for 2 faces. The MMS test currently imposes uniform shear strain with zero slip. The xy strain should be -0.5 leading to a y traction of -1. For 2 faces, we get an xy strain of -1 leading to a y traction of -2.

To Reproduce

PETSc branch: knepley/fix-hybrid-geometry
PyLith branch: baagaard/feature-mmstests-dynamic-prescribed-slip

Note: This PETSc branch is not up-to-date with PETSc main, so it reports version 3.19.4. It should be rebased on PETSc main or you will need to update configure.ac to allow use this version of PETSc.

cd tests/mmstests/linearelasticity/faults-2d
make mmstest_linearelasticity_faults2d
make export-data
./mmstest_linearelasticity_faults2d UniformStrainRateExplicit::QuadQ1::testResidual

[baagaard-usgs]

The errors show up in

https://github.com/geodynamics/pylith/blob/baagaard/feature-mmstests-dynamic-prescribed-slip/libsrc/pylith/fekernels/FaultCohesiveKin.hh#L429

and

https://github.com/geodynamics/pylith/blob/baagaard/feature-mmstests-dynamic-prescribed-slip/libsrc/pylith/fekernels/FaultCohesiveKin.hh#L472

[baagaard-usgs]

Matt pushed a change to PETSc branch fix-hybrid-geometry and it is now rebase on PETSc main.