Ask an un-set-up field split for its blocks quietly (#827)

[lmoresi]

Fixes #827.

A field split has block solvers only after PCSetUp. Asking an un-set-up split for them is a PETSc error, and PETSc prints its ~38-line "Object is in wrong state / Must call KSPSetUp() or PCSetUp() before calling PCFieldSplitGetSubKSP()" banner before the exception reaches Python, so except could not hide it and healthy solves looked broken.

Three paths reached it:

Path	Site
viscoelastic solve: the velocity-preconditioner check ran before solving (#827 reproducer)	SNES_Stokes._check_velocity_preconditioner
plain warm solve with no linear solve, after a consistent_jacobian change rebuilt the solver	same
custom-multigrid "still installed?" test before each solve, on a solver rebuilt by a consistent_jacobian change; every notch run with a non-nested gmsh FMG tail printed it	custom_mg._pcmg_still_installed

Change. _fieldsplit_sub_ksps(pc) in utilities/_petsc_tools.py returns the blocks, or [] when there are none yet, with the banner silenced for that one call: the "return" error handler still raises, it only stops the printing, and it is popped on every path. (The "ignore" handler that broke 23 tests in an earlier attempt turns errors into success; this does not.) Both sites use it, and the viscoelastic check moves to after the solve, as on the plain path.

Behaviour change, by design. Before, the viscoelastic check always returned early, so a viscoelastic solve on a mesh without a hierarchy never reported that its velocity block was running GAMG. It now warns and records velocity.fell_back_from_fmg, as a plain solve already did (once per process under the default warning filter).

Tests. tests/test_1077_quiet_unsetup_fieldsplit.py captures fd-level stderr. The three path tests fail on the unfixed code on the banner assertion and pass with the fix; a fourth checks the helper ([] before set-up, two blocks after). Related suites (multigrid, custom-MG hierarchy, rotated free-slip, MG option bundle, restore/continuation, viscoelastic): 77 passed, no banner. A real notch driver run with a gmsh tail: 1 banner before, 0 after.

Adversarial review. Record in docs/reviews/2026-10/subksp-banner-827-adversarial-review.md. Findings: the plain-path site was untested (fixed, test added); the helper leaked into the uw namespace via a star import (fixed, now private); docstring (fixed); the viscoelastic warning is a behaviour change (kept, stated above). Failed attacks: handler stack restored on every path including KeyboardInterrupt; np 2 and np 4 with no banner and no hang (set-up state follows PCSetUp, run on every rank); _pcmg_still_installed semantics unchanged; the other getFieldSplitSubKSP call sites are all after a setUp, inside a solve, or already guarded; no import cycle.

Not in this PR.

• petsc_generic_snes_solvers.pyx (_discard_preconditioner_setup) carries an identical copy of the silencing pattern; left alone so the solver core is untouched in a bugfix. A follow-up can route it through the helper.
• Pre-existing: the fallback warning's stacklevel=3 resolves to memprobe.py, so every solve shares one warning location.
• Pre-existing, unrelated: test_0017_describe_and_render.py::test_every_family_describes_itself_at_the_class_level fails when test_1057_ddt_transport_plugin.py ran earlier in the same process. test_1057 defines a solver subclass TensorTransport with no summary, and the description test enumerates every subclass.

Checklist (UW3 Style Charter — docs/developer/UW3_STYLE_CHARTER.md)

• I have read the Style Charter and followed it over the surrounding code (S2)
• No drive-by refactors, renames, or "while I was here" cleanups (S9)
• Bug fixes ship the regression test, written first, with level_* and tier_* markers (S8)
• No hedging names (maybe_ / try_ / do_) and no commented-out code (S3, S4)
• Every exception swallow states its sanctioned failure mode in a comment (S4)
• New data access uses .array / mesh.X.coords — no with ....access(...), no mesh.data (S7)
• Parallel safety considered — np2/np4 checked where swarm/mesh/solver dispatch is touched (S11)
• No pixi.toml / pixi.lock or other dependency changes riding in an unrelated PR
• AI-assisted work carries the attribution line below

Underworld development team with AI support from Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[github-actions]

🔍 Architectural Review Checklist

Thank you for submitting an architectural review! Reviewers should validate:

Design & Architecture

• Design rationale is clear and well-justified
• Trade-offs are documented with alternatives considered
• System architecture is comprehensible
• Integration points are clearly identified

Implementation

• Implementation matches documented design
• Code quality meets project standards
• Breaking changes are identified and justified
• Backward compatibility is properly addressed

Testing & Validation

• Testing strategy is adequate for the changes
• Test coverage is sufficient
• Edge cases are properly covered
• Performance impact has been assessed

Documentation

• Known limitations are clearly documented
• Benefits are quantified with metrics
• User-facing changes are documented
• Migration guide provided (if needed)

---

Review Process: See CODE-REVIEW-PROCESS.md

Approval: This PR merge = Review formally approved

[github-actions]

✅ Test Suite: success