Issues

• 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

• evaluate(..., evalf=True) flags every point as extrapolated

  bug
  underworldcode/underworld3#798 · 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

• test_0053 hang_watchdog: SIGSEGV inside CPython faulthandler dump_traceback during full tier_a runs (state-dependent)

  underworldcode/underworld3#793 · Filter by author lmoresi opened on Sep 26, 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

  underworldcode/underworld3#791 · Filter by author lmoresi opened on Sep 25, 2026

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

  underworldcode/underworld3#788 · Filter by author lmoresi opened on Sep 24, 2026

• 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

  underworldcode/underworld3#784 · Filter by author lmoresi opened on Sep 24, 2026

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

  underworldcode/underworld3#783 · Filter by author lmoresi opened on Sep 24, 2026

• 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

• solve_history loses the Picard stage of a continuation solve: _capture_solve_report reached from 1 of 6 snes.solve sites

  underworldcode/underworld3#778 · Filter by author lmoresi opened on Sep 22, 2026

• A yielding model started from rest divides by zero: strainrate_inv_II_min defaults to 0 while the plastic viscosity is tau_y/(2 edot_II)

  underworldcode/underworld3#776 · Filter by author lmoresi opened on Sep 22, 2026

• Snapshot/restore does not capture expression contents: a rewind returns the fields of one step with the parameters of another

  underworldcode/underworld3#774 · Filter by author lmoresi opened on Sep 21, 2026

• Replay/adjoint linearises about a state the forward run never visited: snapshots restore fields but not parameter values

  underworldcode/underworld3#773 · Filter by author lmoresi opened on Sep 21, 2026

• The transcript fingerprints the symbolic form but never parameter VALUES, so a frozen or shared parameter is invisible

  underworldcode/underworld3#772 · Filter by author lmoresi opened on Sep 21, 2026
  1

• uw.function.evaluate returns SI for any expression carrying a dimensional constant, violating the ND<->units boundary contract

  underworldcode/underworld3#771 · Filter by author lmoresi opened on Sep 21, 2026

• Warm-started solves report CONVERGED_FNORM_RELATIVE without moving: continuation ladders silently produce unconverged states

  underworldcode/underworld3#770 · Filter by author lmoresi opened on Sep 21, 2026
  2

• adjoint_solve inherits the forward solve's INEXACT Newton tolerance: every sensitivity on an iteratively-solved problem is silently short by 15-20%

  underworldcode/underworld3#769 · Filter by author lmoresi opened on Sep 20, 2026