fix(showcase): wooden ladder and hitching post visual quality pass

[TMHSDigital]

Summary

Visual quality pass on wooden-ladder and hitching-post. Generators only. Inspection sheets stayed in .scratch/ and were not committed.

Hero stills re-rendered on Blender 5.2.1 LTS (Cycles, CPU). check_framing was called with no deviation=. Ladder fill max 0.856, hitch fill max 0.800, both margins above 0.02. framing_ok.

wooden-ladder

Inspection rounds: 3.

Defects (round 1, old mesh)

1. Shoe cubes were world-axis while the stiles are raked 12°, so each stile stabbed through its shoe. ortho_left, ortho_right, ortho_top, ortho_bottom.
2. Top caps were the same small chrome boxes, not ferrules. ortho_front, ortho_top.
3. All six rungs were the same dowel. ortho_front, ortho_top, clay.
4. Iron read as blown-out plastic. ortho_front, ortho_top.
5. Wood was one flat color. threeq_a.
6. Stiles were featureless boxes; the clay silhouette was two sticks. clay, ortho_left.
7. Coplanar-pair count was not asserted.

Round 3 (sole height 32 mm, sole chamfer, stronger barrel variation): front, side, top, clay, and the shoe close-up show a ladder. The sleeve follows the rake and sits in a level plate. The ground close-up shows wood, sleeve, and sole stacked with no daylight.

Generator

• Sleeve is built in the stile frame, then the tread is a separate world-axis plate from Z=0 after the rake. The plate height is the station that covers the tilted sleeve, not a Z nudge.
• Rungs are one shell: tenon inside the stile, barrel only in the clear span, so clearance stays on the tenon when the barrel changes.
• Cap plate overlaps the sleeve by 2 mm so the two metal tops are not coplanar.
• Bands are sized to the post-bevel section (grip = bevel offset + 1.5 mm) and added after the stile bevel.

Triangles 936 → 2536. The increase is turned rungs and ferrules, not hidden-face removal.

Hygiene (4.5.11 LTS / 5.1.2 / 5.2.1 LTS)

Default exit 0 on all three. Tris 2536, AABB (0.4897, 0.3760, 1.4870), zmin 0, hygiene and zfight 0, clearance 0.008, engage 0.01399, breakout 0.03780, shoe bite 0.020, cover 0.038, rail zmin 0.02037, sole zmin 0. LOD 0.5000 / 0.2192. Collider 24. Export 187528 / 187528 / 187504.

Falsifier	Exit	Budget
--skip-decimate	9	LOD1 ratio 1.0
--stray-vert	15	loose_v=1
--twin-sole	15	zfight=6
--lift-z	16	zmin 0.05
--fat-rungs	17	clearance 0
--short-stile	18	cover −0.012

hitching-post

Inspection rounds: 2 on the rebuilt generator, after one look at the old mesh.

Defects (old mesh)

1. Rings were faceted tori clipped into the arm, not hung through an eye. threeq_b, ortho_left, joint close-up (overlap 26 with the wood).
2. The eye sat on the face with no shank into the timber. ortho_left, threeq.
3. The arm was three boxes (tenon plus two stubs). ortho_front.
4. The shoe was an open band; the post end grain was the ground contact and shared Z=0 with the band. ortho_bottom, ortho_left.
5. The pyramid cap was beveled with the post, leaving a seam. ortho_left.
6. XY recenter walked the post off the origin because the rings stuck out in −Y.
7. Iron was flat black.
8. Coplanar-pair count was not asserted.

Round 1 of the rebuild still ran the shank through the ring (joint close-up of the eye and ring; BVH overlap 9). Round 2 stops the shank in the top of the eye tube. Front, clay, and both three-quarter views read as a hitching post: eaved cap, one rail, two hung rings, cup shoe.

Generator

• Ring center is the eye center plus (0, 0, −RING_MAJOR), so the eye lies on the centerline. The eye hangs under the rail; a face-mounted ring swings into the timber.
• The shank station is the top of the eye tube, not the eye center.
• One rail through the post, beveled with the post. Cap is added after the bevel and embedded by CAP_EMBED.
• Sole is a closed plate at Z=0; the shoe wall bites both the post and the sole. Z snap only, no XY recenter.

Triangles 1598 → 1680.

Hygiene (4.5.11 LTS / 5.1.2 / 5.2.1 LTS)

Default exit 0 on all three. Tris 1680, AABB (0.5000, 0.1490, 1.2460), zmin 0, hygiene and zfight 0, post zmin 0.010, post XY 0, arm cos 1, ring error 0, ring–wood 0, eye–wood 0, shank–ring 0, shank–eye overlap 40, sole–shoe overlap 8, wood–metal gap 0.00075. LOD 0.5000 / 0.2190. Collider 80. Export 127428 / 127428 / 127412.

Falsifier	Exit	Budget
--skip-decimate	9	LOD1 ratio 1.0
--stray-vert	15	loose_v=1
--twin-sole	15	zfight=6
--lift-z	16	zmin 0.05
--clip-ring	18	ring_err 0.024 (also overlaps the rail; centerline fires first)
--short-post	19	post zmin 0.030

Conventions

showcase/README.md now requires, for every future piece:

• Level sole on a raked leg (exit 18)
• Hung ring whose eye sits on the centerline (exit 18), with --short-post for the cup seat (exit 19)
• --twin-sole as the coplanar-face falsifier (exit 15)

Existing hygiene, named supports, joint-fit, and plumb budgets stay.

Test plan

• Default and every falsifier on 4.5.11 LTS, 5.1.2, and 5.2.1 LTS
• Framing gate on both stills, 5.2.1 LTS, no deviation=
• CI smoke on 5.2, 4.5, and the labeled 5.1 job

Made with Cursor