Issues

• Every viscoelastic Stokes solve prints a PETSc error banner from _check_velocity_preconditioner

  underworldcode/underworld3#827 · Filter by author lmoresi opened on Oct 7, 2026

• test_1063 [ti]: constrained free-slip topography 1.8% off its recorded anchor on the anchor's own mesh

  underworldcode/underworld3#826 · Filter by author lmoresi opened on Oct 7, 2026

• uw.unwrap passes keep_constants into depth: expands one level, and keep_constants=False expands nothing

  bug
  underworldcode/underworld3#824 · Filter by author lmoresi opened on Oct 7, 2026

• unwrap_expression(symbolic_keep_constants) is exponential on nested constitutive models: 563 s for a 10k-node tree; JIT setup 10-40+ min on the notch

  underworldcode/underworld3#823 · Filter by author lmoresi opened on Oct 6, 2026

• Native velocity FMG fails with err62 (injection) for P3 velocity on a refined Plex (Stokes/NavierStokes)

  underworldcode/underworld3#821 · Filter by author lmoresi opened on Oct 6, 2026

• Automatic cold Picard warm-up diverges a LINEAR constrained Stokes solve (Stokes_Constrained + multiplier free-slip)

  underworldcode/underworld3#817 · Filter by author lmoresi opened on Oct 5, 2026
  1

• Forward integration-point history: the per-cell fit is unstable where the flow empties a cell

  underworldcode/underworld3#811 · Filter by author lmoresi opened on Oct 4, 2026

• Passive tracers that linger at a no-slip wall cross the facet and are deleted

  underworldcode/underworld3#810 · Filter by author lmoresi opened on Oct 4, 2026

• A refined mesh reloaded from its checkpoint has cells joined to the wrong vertices

  underworldcode/underworld3#809 · Filter by author lmoresi opened on Oct 4, 2026

• release/v3.2.0: RHEL recipes pin NumPy <2 despite Pixi requiring NumPy 2

  underworldcode/underworld3#808 · Filter by author gthyagi opened on Oct 4, 2026
  

• test_adapt_with_repair_runs_in_parallel: FAILED then hangs at np=2

  bug
  underworldcode/underworld3#807 · Filter by author jcgraciosa opened on Sep 30, 2026

• Solver introspection does not report the numerical configuration a solve ran with

  underworldcode/underworld3#806 · Filter by author lmoresi opened on Sep 30, 2026

• solver.gradient() takes the adjoint of an unconverged forward solve and returns a wrong-signed number

  underworldcode/underworld3#804 · Filter by author lmoresi opened on Sep 28, 2026

• test_1063_constrained_freeslip_parallel[ti]: topography moved from its recorded value (37.75 vs 37.08) at np 4

  bug
  underworldcode/underworld3#801 · Filter by author lmoresi opened on Sep 27, 2026
  1

• evaluate() is off by up to 4e-5 for a field the P2 space holds exactly, at points near a node

  bug
  underworldcode/underworld3#799 · Filter by author lmoresi opened on Sep 27, 2026

• Particle (Lagrangian) stress history differs from serial by 1.2e-5 at np >= 4

  bug
  underworldcode/underworld3#797 · Filter by author lmoresi opened on Sep 27, 2026

• The automatic cold-start 'Picard' warm-up (and solve(picard=N)) runs nrichardson with no linear solve — the design specifies a frozen-coefficient Picard step

  fixed-in-PR
  underworldcode/underworld3#791 · Filter by author lmoresi opened on Sep 25, 2026
  1

• Viscoelastic stress history in a units model: the particle store takes dimensional magnitudes, the nodal flavour crashes, and no history store carries stress units

  fixed-in-PR
  underworldcode/underworld3#788 · Filter by author lmoresi opened on Sep 24, 2026
  1

• Adjoint field dual is silently zero for a field that enters through a Dirichlet datum

  underworldcode/underworld3#787 · Filter by author lmoresi opened on Sep 24, 2026

• uw.function.evaluate silently returns values for points far outside the mesh domain

  underworldcode/underworld3#786 · Filter by author lmoresi opened on Sep 24, 2026

• Swarm.repopulate: after a removal the coordinate rows no longer match the value rows, so every refill in that call reads unrelated particles

  fixed-in-PR
  underworldcode/underworld3#784 · Filter by author lmoresi opened on Sep 24, 2026
  1

• Particle Lagrangian stress history: an inflow cell is refilled from stale neighbours, not from inflow_value

  fixed-in-PR
  underworldcode/underworld3#783 · Filter by author lmoresi opened on Sep 24, 2026
  1

• Selecting yield_smoother='powermean' silently sets yield_softness=1.0 — the harmonic mean — so a model reads as Drucker-Prager while running a strong smoothing

  underworldcode/underworld3#782 · Filter by author lmoresi opened on Sep 23, 2026

• adjoint_solve(target=...) does not validate the target spaces: a mismatch is a bare numpy broadcast error

  underworldcode/underworld3#781 · Filter by author lmoresi opened on Sep 23, 2026
  1

• A cold solve has no absolute convergence test (atol=0), so CONVERGED_FNORM_RELATIVE is its only possible success reason

  underworldcode/underworld3#779 · Filter by author lmoresi opened on Sep 22, 2026