Water Trough

A showcase piece, not an example. Procedural staved water trough on a timber stand (watertight U-hull, board-built ends with a drain bung, contained water with a rippled surface and meniscus, iron straps through-bolted and clipped over the rim, pegged trestle legs) then the shipped pipeline: unique-cell UVs, Cycles high-to-low normal bake, LOD chain, convex collider, Unity glTF export.

The hull, ends, straps and water all sample the same YZ arc. Each end is three vertical boards whose bottoms follow the hull's outer arc and whose tops are level, and the staves tenon into them. It is neither a bounding-box slab around the U (--box-ends) nor an open U-band (--open-ends). Straps sit at STRAP_X, offset from LEG_X, so the trestle does not punch through the iron. Stretchers span the inner faces of the legs. --round-band is not used; --float-strap is the seat falsifier.

It asserts budget conformance of the generated result. It does not witness an API contract. "It rendered without error" is not a check.

Composes skills mesh-editing-and-bmesh, bake-high-to-low, depsgraph-and-evaluated-data, engine-export-presets, and snippets bake_normal_high_to_low.py, setup_bake_target_image.py, lod_chain.py / decimate_to_budget.py, convex_hull_collider.py, export_preset_unity.py (helpers copied, not imported as a package). Hygiene combinatorics match examples/mesh-hygiene-audit (copied, not imported).

Intended size: 1.08 m tray length, 0.44 m across the U, 0.51 m overall height; outer AABB 1.111 × 0.552 × 0.512 m (the bung adds 18 mm in X, the rim clips 7 mm in Z).

Budgets#

Declared as named constants; every gate recomputes from the mesh, materials, UVs, evaluated LOD, collider, or export file.

Base triangles rose from 2484 to 3700 in the first quality pass: box end slabs and a single extruded U became ten jittered staves, U end-caps, continuous straps offset from the legs, and stretchers that meet the inner faces. The second pass took them to 3124: the open U-band ends became solid end boards, which close the section with fewer faces. The third pass took them to 3316: each end became three boards. The fifth pass took them to 4792: the water became a lofted, rippled volume (~900 triangles; it was one flat quad on a U prism), and the straps gained ten bolt heads and four rim clips, the trestles six pegs, and the right end a bung. The band moved from 2800–3500 to 4300–5200, the same ~±10 % slack around the measured count. The collider hulls the trough without the sub-2 cm hardware (bolts, pegs, bung): detail that small carries no collision meaning. The rim clips are 47 mm wide and stay in it. The 5.1.2 column was dropped with this change: the new model was measured on 4.5.11 and 5.2.1 only.

DECIMATE COLLAPSE triangle counts are not identical across series — the gate is a ratio band, not an exact count. Bake pixels are stochastic; the gate is has_data plus operator FINISHED, not byte-identity. Stave-width jitter uses closed-form sin(i); plank tone is the only RNG, seeded (TONE_SEED). Export byte counts differ by 16 B on 5.2.1 (glTF serializer), not a gated axis.

The hull-size budget (exit 19) runs before the outer AABB budget (exit 8) so --narrow-hull is not swallowed by a moved bounding box.

Hygiene#

Recomputed from the generated mesh, not asserted about the script.

Joint fit and seat#

Why the ends are solid boards#

The first pass made each end a U-shaped band the staves tenoned into. That closed the rim and nothing else: the water's own flat end face sat exposed where a board belongs, visible in the end orthos and the end-cap close-ups. Water in that trough would pour straight out, and with the ends open the piece read as a hammock or sling. The strap and water gap budgets passed throughout, because the water did touch the hull; nothing asked whether anything held it in at the ends.

containment_audit asks. From each water end face it casts rays outward along X, from the face's perimeter vertices and from points drawn 50% and 90% of the way in from the centroid, so the middle of the section is tested and not only the rim a band would cover. Every ray must hit timber within 1.5 × END_T. --open-ends restores the U-bands: 119 of the 120 rays miss, and the piece exits 18.

Why each end is three boards#

The third pass found each end was one D-shaped slab: a single board from rim to rim, one tone, no joint, the thing the showcase's "a head or bottom is boards, not a slab" rule exists to prevent (end orthos, end-cap close-ups, the hero). Each end is now three vertical boards of uneven width (END_BOARD_FRACS), each with its own tone and grain, its bottom following the hull arc under its own span. Neighbours lap 1 mm into each other (END_BOARD_LAP), so the joint is tight: a through seam would let the water out and a containment ray through, and the bevel on each board's face edges shows the joint as a groove. The rim span is now read on the union of an end's boards, so the inner edge of a board at a joint is not taken for the end's rim. Each board follows the arc with three samples (END_BOARD_SAMPLES); four cost 3556 triangles, over the 3500 ceiling. --slab-ends restores the one-piece board and exits 20.

Surface#

Staves, boards and legs came out of one flat material, so every board was the same board. paint_planks writes a seeded PlankTone and each shell's own long axis as GrainDir, as face attributes. The wood shader stretches its grain along that axis. The water was a 0.08-roughness mirror that went near-white under the key. It is now a dark body with a facing-ratio sky tint toward grazing angles and a faint ripple bump. A light placed to reflect in the surface was tried first and rejected: it either mirrored as a white sheet or showed nothing, and its spill lifted the dark stage. The fill dropped from 68% to 50%, so the inner stave walls show above the waterline; at 68% the surface sat nearly flush with the rim and read as felt in a frame. Straps widened from 28 to 40 mm. The third pass darkened the grazing tint from a pale grey-blue (0.20, 0.27, 0.29) to a deep teal (0.055, 0.105, 0.115): the hero looks across the surface near grazing, where that tint wins, and it read as a painted slab.

The fifth pass made the surface geometry. The water is lofted along X, WATER_NX stations with WATER_NY strips across each, under two crossing standing waves of RIPPLE_AMP that damp to zero at the staves, and a MENISCUS_H meniscus that climbs the wall over MENISCUS_W. The water is the one smooth-shaded material, so the key light runs along the crests. The straps were invisible from the hero camera, which looks over the rim while the straps wrap the hull's underside. Each strap now ends in a clip hooked over the rim and carries five domed carriage-bolt heads. The trestle legs are pegged through their stretcher tenons, and a drain bung sits low in the right end.

The fourth pass made the surface read as liquid rather than as a teal panel: larger ripples (noise scale 22 to 7) under a stronger bump (0.08 to 0.30) at roughness 0.03 catch the key as glints, and the render path turns on EEVEE screen-space raytracing (4.2+) so the staves mirror in the water. The iron gets vertically stretched rust runs that dull its metallic response. A lower camera was tried to show the hull straps and rejected: the water surface dropped out of view.

Falsifiers#

Each violates one named budget. After the fifth pass all ten were re-run on 4.5.11 and 5.2.1 and returned the same code on each.

--short-legs lifts the shoes and stretches each leg's foot down to the floor, so AABB zmin stays 0 (the legs still plant) and the named-support gate is what fires. Lifting the shoes alone used to let the whole piece re-ground on the legs. That shrank the AABB by ~10 mm, 0.4 mm inside the bounding-box tolerance, and a 1.5 mm model change in this pass tipped it to exit 8. With the feet stretched, the envelope moves 1.6 mm.

--float-strap lifted the straps 0.040 m until the fifth pass. With bolt heads and rim clips riding on the straps, that widened the AABB past its tolerance and the probe hit exit 8 before the strap-seat budget it exists for. It now lifts them STRAP_FLOAT (0.012 m), still a 0.0117 m gap against the 0.008 m budget.

Run#

blender --background --python water_trough.py --
blender --background --python water_trough.py -- --skip-decimate
blender --background --python water_trough.py -- --stray-vert
blender --background --python water_trough.py -- --lift-z
blender --background --python water_trough.py -- --short-legs
blender --background --python water_trough.py -- --box-ends
blender --background --python water_trough.py -- --float-strap
blender --background --python water_trough.py -- --narrow-hull
blender --background --python water_trough.py -- --open-ends
blender --background --python water_trough.py -- --slab-ends
blender --background --python water_trough.py -- --float-bolts
blender --background --python water_trough.py -- --output trough.png

Smoke passes no flags.

Exit codes#

File-local. 9 is a valid check code. 10 is reserved for gallery_framing.check_framing on the --output path. 15–19 are the hygiene and joint-fit family; 20–21 are file-local.