Issues

• A solve on a moved refined mesh runs on the pre-move coordinates: the move skips dm_hierarchy[-1]

  bug
  underworldcode/underworld3#1037 · Filter by author lmoresi opened on Nov 10, 2026

• NavierStokesSLCN with no bodyforce set fails at setup: ShapeError (2, 1) + (1, 2) in F0

  bug
  underworldcode/underworld3#1034 · Filter by author lmoresi opened on Nov 10, 2026

• Timestep setters: a UWexpression is frozen, a non-time quantity is accepted, a 0-d array rewires every step (TransientDarcy, Richards, NavierStokesSLCN)

  bug
  underworldcode/underworld3#1031 · Filter by author lmoresi opened on Nov 10, 2026

• θ has no effect in Diffusion, TransientDarcy and Richards: the symbolic flux history is the current flux

  bug
  underworldcode/underworld3#1029 · Filter by author lmoresi opened on Nov 10, 2026

• TransientDarcy: a forced setup changes the answer by 7-9 %

  bugfixed-in-development
  underworldcode/underworld3#1027 · Filter by author lmoresi opened on Nov 9, 2026
  1

• AdvDiffusionSLCN leaks a SNES on every step when the timestep changes

  bugfixed-in-development
  underworldcode/underworld3#1026 · Filter by author lmoresi opened on Nov 9, 2026
  2

• SLCN estimate_dt limits the step by the cell diffusion time, which shrinks it as h² on fine meshes

  underworldcode/underworld3#1025 · Filter by author lmoresi opened on Nov 9, 2026
  1

• Stokes adjoint, matrix route: the velocity multigrid is built from J's block, not J^T's

  bugfixed-in-development
  underworldcode/underworld3#1021 · Filter by author lmoresi opened on Nov 8, 2026
  2

• Custom-P reinstall on the standard path leaks deferred PETSc objects at np > 1: +28 entries and 8 MB per reinstall on Stokes (#1007 pattern)

  bugfixed-in-PR
  underworldcode/underworld3#1020 · Filter by author lmoresi opened on Nov 8, 2026
  2

• preconditioner = "gamg" on an adapt child with a FAC level fails with PETSc error 73: GAMG reads the leftover FAC composite key

  bugfixed-in-PR
  underworldcode/underworld3#1019 · Filter by author lmoresi opened on Nov 8, 2026
  1

• After mesh.deform the FAC patch is declined: the custom-P coarse levels do not move with the fine mesh

  bugfixed-in-PR
  underworldcode/underworld3#1017 · Filter by author lmoresi opened on Nov 8, 2026
  3

• Stokes adjoint_solve drops the custom-FMG velocity hierarchy and its FAC patch: one-level mg, 29 iterations against the forward's 4

  bugfixed-in-development
  underworldcode/underworld3#1016 · Filter by author lmoresi opened on Nov 8, 2026
  1

• Custom FMG with a FAC patch: re-solve after mesh.deform fails with PETSc error 73 (empty composite); the hierarchy is not re-installed

  bugfixed-in-development
  underworldcode/underworld3#1015 · Filter by author lmoresi opened on Nov 8, 2026
  1

• Stokes: reassigning bodyforce leaks about 23 MB per solve (28 MB per rank at np 2); Poisson and an updated uw.expression do not

  bugfixed-in-PR
  underworldcode/underworld3#1012 · Filter by author lmoresi opened on Nov 8, 2026
  3

• A weakly imposed nonzero velocity makes the cold-solve target follow the penalty: 1e-3 to 3e-2 off at tolerance 1e-4

  bug
  underworldcode/underworld3#1011 · Filter by author lmoresi opened on Nov 8, 2026

• Track feature/shear-band-length-scale: coupled volume fields and micromorphic plasticity, awaiting rebase

  underworldcode/underworld3#1010 · Filter by author lmoresi opened on Nov 8, 2026
  1

• Nodal stress history on Waters and King: a bias and stalled convergence on the #735 port branch

  underworldcode/underworld3#1009 · Filter by author lmoresi opened on Nov 8, 2026
  1

• Rotated free-slip: reassigning an expression each step leaks ~14 MB per solve at np 2 through the prolongation rebuild

  fixed-in-development
  underworldcode/underworld3#1007 · Filter by author lmoresi opened on Nov 8, 2026
  1

• The storage-layout loop is written out in seven places; one helper would stop them disagreeing

  underworldcode/underworld3#1006 · Filter by author lmoresi opened on Nov 8, 2026

• A MATRIX swarm variable reads back transposed

  fixed-in-PR
  underworldcode/underworld3#1005 · Filter by author lmoresi opened on Nov 8, 2026
  1

• The RBF route returns a matrix-valued derivative as (N,1,1), broadcast through global_evaluate

  underworldcode/underworld3#1004 · Filter by author lmoresi opened on Nov 8, 2026

• reset_default_model leaves the non-dimensional scaling flag on, so it leaks between tests

  underworldcode/underworld3#1001 · Filter by author lmoresi opened on Nov 7, 2026
  1

• monotone clamp on one component of a multi-component variable raises cannot reshape

  underworldcode/underworld3#1000 · Filter by author lmoresi opened on Nov 7, 2026

• global_evaluate of a derivative raises at np 2 and up when a point is outside the domain

  underworldcode/underworld3#999 · Filter by author lmoresi opened on Nov 7, 2026
  1

• FreeSurface fails in a units model: ring coordinates are physical, the plex coordinates are model

  underworldcode/underworld3#998 · Filter by author lmoresi opened on Nov 7, 2026
  1