feat(showcase): add butter-churn

[TMHSDigital]

Adds showcase/butter-churn: a plunge churn with a maple cross dasher whose handle rises through the lid. The body is sixteen tapered oak staves that narrow toward the top, bound by three iron hoops. It has a three-board bottom set in a croze and a two-board lid with a hole for the handle.

Selection: the first entry in the pool was the ox-cart tongue I added as a restock in #255. I moved it behind the bookshelf rather than building it now:

• it is a hitch pair, meaningful only beside the yoke;
• a bare tongue or plough cannot stand on its own;
• as a restock it should never have been placed ahead of the older candidates.

The churn is the next unclaimed standalone prop. There were no open PRs, and the only feat/showcase-* ref is already-merged work.

Binaries (all local, .scratch/):

• blender-4.5.11-windows-x64 reports Blender 4.5.11 LTS
• blender-5.1.2-windows-x64 reports Blender 5.1.2
• blender-5.2.1-windows-x64 reports Blender 5.2.1 LTS

The budget that matters: plunge stroke (exit 20)

The body narrows toward the top, so a plunger that clears the staves at rest can jam partway up. It still sits inside the body without touching it, the handle still clears the lid and the envelope does not move.

The piece reads the inner wall off the staves ring by ring and the plunger's reach off the slats. It then finds how far the plunger can rise before it comes within 10 mm of the wall, or reaches the lid, and asserts a 400 mm floor.

	reach	stroke
default	100.6 mm	513.8 mm
--wide-dasher	125 mm slats	140.4 mm → exit 20

Under --wide-dasher the plunger still clears the staves at rest, and every other budget passes. live-run-proven

Falsifiers: exit codes per binary (live-run-proven)

	default	--skip-decimate	--stray-vert	--lift-z	--short-stave	--one-piece-bottom	--short-handle	--float-hoops	--float-lid	--tight-hole	--wide-dasher	--output
4.5.11	0	9	15	16	16	17	17	18	18	18	20	—
5.1.2	0	9	15	16	16	17	17	18	18	18	20	—
5.2.1	0	9	15	16	16	17	17	18	18	18	20	0

Flag	Target budget	Measured	Exit
--skip-decimate	LOD1 ratio	1.0000	9
--stray-vert	hygiene	loose_v 1	15
--lift-z	grounded zmin	0.050	16
--short-stave	named staves	one foot 6 mm up	16
--one-piece-bottom	bottom boards	1 of 3	17
--short-handle	handle bite	−5.0 mm	17
--float-hoops	hoop seat	−3.5 mm	18
--float-lid	lid seat	−3.6 to −3.0 mm	18
--tight-hole	handle clearance	−2.0 mm	18
--wide-dasher	plunge stroke	140.4 mm (floor 400)	20

Envelope: every falsifier leaves the AABB unchanged. live-run-proven

Checks run locally

• tests/check_falsifier_targets.py: the static check passes. --run <BLENDER> --only butter-churn passes all 10 falsifiers on each of the three binaries. live-run-proven

• tests/smoke/run_example.py --forbid-skip: [PASS] butter-churn on 4.5.11, 5.1.2 and 5.2.1. live-run-proven

• Other local checks:

  • check_exit_code_readme, check_import_export_rules, check_engine_id, check_gallery_images, test_harness
  • the inline validate-manifest, validate-counts and referenced-paths steps, plus the gallery-drift logic, all in a clean worktree (34 showcase pieces, no drift)

  All pass. live-run-proven

• Determinism: every budget value is byte-identical across the three binaries. The only difference is the glTF byte count: 239796 on 5.2.1 against 239804 on the others, which is exporter metadata. There is no RNG. Check-only wall-clock is about 1.8 s. live-run-proven

• Remote-only checks: the ecosystem drift check and the 4.5 / 5.1 / 5.2 smoke jobs are CI-proven on this PR.

Fix cycles (live-run-proven unless marked)

• 240 coplanar cross-shell pairs. Neighbouring stave feet and tops, the two lid boards, and the bottom boards (faces and a rim chord the seams split) each shared a plane. The fixes:

  • stave ends are canted inward, doubled on alternate staves (at one cant, neighbours 5.6° apart still matched: 32 pairs remained);
  • one lid board stands 0.6 mm proud;
  • the middle bottom board steps up and in by 0.6 mm.

  The count is now 0.

• Collider steal. --float-hoops first exited 11 (collider 242 against 240) instead of 18. The hoops add nothing to collide with, so they came out of the body hull and the ceiling was not raised. The collider is now 178 triangles, and the falsifier exits 18.

• Framing (inspection-only for the look, measured by the gate). The first render filled 0.944 tall, over the gate's 0.90 cap, and exited 10. The camera was pulled back 12%, and the fill is now 0.839.

Visual gates

Contact sheet: docs/gallery/contact-sheets/butter-churn-contact-sheet.webp

criterion	butter-churn	calibration	verdict
mean luminance	0.2327	0.151–0.302	in band
stage darkness	0.2031	0.145–0.236	in band
wedge warmth	+0.149	−0.43 … +0.37	warm pool on the floor behind the churn
saturation	0.2001	0.144–0.231	in band
subject fill	0.839 tall	0.70–0.90	passes the framing gate, min margin 0.072
thumbnail legibility	—	—	a hooped staved churn with a handle reads at tile size

Asset sheet: docs/gallery/asset-sheets/butter-churn.webp, beside the four re-rendered reference panels. Verdict: passes.

• Alone under neutral light, it shows per-stave tone, dark iron hoops, a lid overhanging the rim and a turned handle.
• It sits at the level of the rocket and arm, and is not the least-designed object.

inspection-only

Close-ups (lit and clay: lid, hoop, foot, underside, side, handle, rim):

• the lid step and the hole's clearance are visible;
• three bottom boards show from below;
• the hoops sit on the staves with no speckle;
• one soft dithered contact shadow shows under a hoop edge at 800 px, which is EEVEE shadow sampling, not geometry.

inspection-only

The hero was rendered on 5.2.1 through scripts/render_hero.py, and build_gallery.py was run last. I read the generated <img alt> and the witnesses callout in docs/gallery/index.html and docs/gallery/butter-churn/index.html. inspection-only

needs-5.1 is applied because the piece runs a normal bake.

🤖 Generated with Claude Code