Wheelbarrow

A showcase piece, not an example. Procedural wooden wheelbarrow (two shafts that are the handles, each one timber from grip to axle; a flared hopper tray with overlapping floor slats and walls of stacked boards; iron straps; a single flat-tread spoked wheel; rear legs) then the shipped pipeline: unique-cell UVs, Cycles high-to-low normal bake, LOD chain, convex collider, Unity glTF export.

The tray walls seat into the floor rather than sitting on top of it, so a wood bevel cannot open daylight at the joint. The tyre is a flat felloe wrap, not a torus.

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).

Intended size: a 1.56 m long barrow, 0.59 m across the tray top, 0.58 m to the top of the raked front board; the tray is 0.94 m long at its top edge and 0.22 m deep.

Budgets#

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

Hygiene#

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

Joint fit and seat#

Form#

DECIMATE COLLAPSE triangle counts are not identical across series, so LOD2 ratios differ slightly. 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. Construction is closed-form; the only RNG is the seeded per-board wood tone. Bevel inputs are sorted by edge index, so the face order is the same on every run; a Python set of edges handed to the bevel had made it vary. Export byte counts differ on 5.2.1 (glTF serializer), not a gated axis.

Why the frame is two timbers and the tray a hopper#

The second pass found the barrow read as a crate on legs. The frame was three boxes per side: a handle that stopped at the tray's rear corner, a level run hidden exactly under the side wall (both at 0.255 m from the centreline), and a fork that started at the front corner and turned in a 50° V to the hub. From outside, the tray carried everything. The tray itself was a square box, and a spreader between the forks stood inside the wheel, touching neither fork. The iron straps ran 12 mm above the wall tops.

• Shafts. Each shaft is one mitred sweep through the grip, the tray's rear, its front and the axle. It narrows in plan the whole way (0.27 m at the grips, 0.19 and 0.15 m under the tray, 0.038 m at the hub), so it shows under the tray and reads as one piece of wood. The legs and a stretcher hang from it, and a bearer under the tray front, its ends buried in the shafts, replaces the spreader.
• Hopper. The tray is built from a bottom outline and a top outline: sides flared 15°, the front raked 30° out over the wheel for tipping, the low rear leaning back 12°. Each wall is the planar band between its bottom and top edges; the end boards are housed half a board into the sides, and the sides run past the ends' outer faces.
• Measured in the wall's plane. A leaning board's outer face sits lower than its inner one, so AABB tests lie: a Z seam reads negative and a zmin seat reads deep. Seams are measured along each wall's own up direction, and the seat on each wall's lowest board's inner face, both taken from the boards' own faces.
• Straps follow the flared wall and stop 10 mm below its top.

WALL_SEAT went from 12 mm to 8 mm: a raked board's outer bottom edge sits WALL_T·sin(RAKE) below its inner one, and at 12 mm it would have broken through the underside of the 22 mm floor. The handle-join budget is gone, because the handle no longer joins anything: it is the shaft. The shaft-run budget replaces it. Base triangles went from 2536 to 2440, and the collider from 90 to 80.

Walls, wood and iron#

Side and front walls are two boards stacked with a 2 mm seam, held by the straps; the low rear wall stays one board. Wall boards asserts the count and the seam; --wide-seams opens the seams to 8 mm (same boards, same triangles) and exits 17. Faces nobody sees get no bevel: wall bottoms buried in the floor and slat undersides. Wood has a tone and grain per board and shaft; the iron is dark and rusted. The temp .glb is removed after its size is measured.

Falsifiers#

Each violates one named budget. All eleven were run on 4.5.11, 5.1.2 and 5.2.1 and returned the same code on each.

--box-tray stands every wall vertical on the flared tray's top outline, so the envelope and the tray-size budget stay put and only the flare gate can fail.

Run#

blender --background --python wheelbarrow.py --
blender --background --python wheelbarrow.py -- --skip-decimate
blender --background --python wheelbarrow.py -- --stray-vert
blender --background --python wheelbarrow.py -- --lift-z
blender --background --python wheelbarrow.py -- --short-legs
blender --background --python wheelbarrow.py -- --fat-spokes
blender --background --python wheelbarrow.py -- --float-walls
blender --background --python wheelbarrow.py -- --pipe-rim
blender --background --python wheelbarrow.py -- --wide-seams
blender --background --python wheelbarrow.py -- --split-shafts
blender --background --python wheelbarrow.py -- --box-tray
blender --background --python wheelbarrow.py -- --output barrow.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.