fix: restage the game-prop pair heroes from deployed-page review

README.md

@@ -605,8 +605,8 @@ The compound-collision contract prop pipelines (engines generally,
 FiveM/GTA-style prop workflows specifically) ingest: one convex hull piece
 per part group, built with `bmesh.ops.convex_hull` from a coarse
 `sec(π/n)`-inflated cage — never the dense render mesh, whose hull measures
-368 faces, over the 255-face per-piece engine budget. Closed-form plane
-tests prove containment (5.9e-08), convexity, watertightness, outward
+380 faces, over the 255-face per-piece engine budget. Closed-form plane
+tests prove containment (4.4e-08), convexity, watertightness, outward
 winding, and Euler characteristic 2 per piece. Proud details cost cage rows;
 concave grooves are free.
 

ROADMAP.md

@@ -112,7 +112,7 @@ Not committed; target list for the next content version. (v0.3.0 shipped the smo
 - Degenerate-bevel weld hazard (snippet or rule): bevel width ≥ half a box dimension creates zero-area faces whose loops weld on glTF export (found authoring `gltf-export-roundtrip`, where the count check caught a 36-vertex weld)
 - ~~GAMMA_CROSS blend-curve witness~~ **SHIPPED** as `examples/vse-gamma-cross/` — the cross blends in a gamma-0.5 space: `((1-t)·√A + t·√B)²` with `t = (frame − start)/duration`, never 1 inside the effect; mid-cross dips 0.115 below the sRGB lerp from crimson/teal (closed form (0.341, 0.349, 0.463) confirmed per frame); AgX-default sampling poisons the fit (0.146 red-channel error, `view_transform='Standard'` mandatory); deleting a consumed input orphans-and-deletes the effect — follow-up to `vse-cut-list`
 - Falsy `bpy_prop_collection` trap snippet: an empty collection is falsy, so `editor.strips or editor.sequences` silently falls through to the legacy accessor on an empty timeline — always branch on `hasattr`; likely generalizes across the API (found authoring `vse-cut-list`)
-- ~~Collision compound witness~~ **SHIPPED** as `examples/collision-hull-proxy/` — game-prop collision as a compound of convex pieces, each a `bmesh.ops.convex_hull` of a coarse `sec(π/n)`-inflated cage (containment 5.9e-08, watertight, positive signed volume, Euler 2, per-piece 255-face budget: body 70, caps 60×3, compound 250); a hull of the dense render mesh measures 374 faces — over budget — which is why pipelines hull cages; proud details cost cage rows, concave grooves are free; byte-identical on 4.5.11 and 5.1.2
+- ~~Collision compound witness~~ **SHIPPED** as `examples/collision-hull-proxy/` — game-prop collision as a compound of convex pieces, each a `bmesh.ops.convex_hull` of a coarse `sec(π/n)`-inflated cage (containment 4.4e-08, watertight, positive signed volume, Euler 2, per-piece 255-face budget: body 70, caps 60×3, compound 250); a hull of the dense render mesh measures 380 faces — over budget — which is why pipelines hull cages; proud details cost cage rows, concave grooves are free; byte-identical on 4.5.11 and 5.1.2
 - ~~Custom-normals / shade-by-angle witness~~ **SHIPPED** as `examples/custom-normals-shade/` — the post-4.1 shading contract: `use_auto_smooth`/`use_custom_normals`/`calc_normals` are AttributeError on BOTH 4.5.11 and 5.1.2; `set_sharp_from_angle` sharp sets match an independent dihedral recompute exactly (188/388 edges, 3 meshes); evaluated normal welds/splits exact (0.0 dev); custom split normals survive depsgraph evaluation within int16 quantization (1.407e-04, not float-exact); **divergence**: the legacy `shade_auto_smooth` operator CANCELS headless on 4.5 ("Asset loading is unfinished", mesh untouched, no exception) while 5.1 FINISHES with the Smooth-by-Angle NODES modifier — the data API is the portable path
 - prop-origin-transform witness — origin to base center, `transform_apply` through the data API, delta transforms, `matrix_parent_inverse` so parented children do not teleport; closed forms: post-apply scale exactly (1,1,1), local bbox min Z == 0, world bbox unchanged (builds on `parent-inverse-orrery`, does not duplicate it)
 - mesh-hygiene-audit witness — the engine-ingest checklist as executable contract: no ngons, no loose vertices, no non-manifold edges, no zero-area faces, consistent outward winding; closed forms via Euler characteristic and exact edge-face incidence counts

docs/gallery/custom-normals-shade/index.html

@@ -659,7 +659,7 @@ <h2>Source</h2>
     <span class="c"># semi-gloss painted steel: rough enough to read as paint, glossy enough</span>
     <span class="c"># that a highlight exposes every normal discontinuity — matte would hide</span>
     <span class="c"># the very shading differences this render exists to show</span>
-    steel = make_material(<span class="s">&quot;OliveDrab&quot;</span>, (<span class="n">0.31</span>, <span class="n">0.35</span>, <span class="n">0.15</span>), <span class="n">0.8</span>, <span class="n">0.26</span>)
+    steel = make_material(<span class="s">&quot;OliveDrab&quot;</span>, (<span class="n">0.30</span>, <span class="n">0.345</span>, <span class="n">0.13</span>), <span class="n">0.8</span>, <span class="n">0.26</span>)
     accent = make_material(<span class="s">&quot;FuelMarker&quot;</span>, (<span class="n">0.55</span>, <span class="n">0.06</span>, <span class="n">0.03</span>), <span class="n">0.3</span>, <span class="n">0.30</span>)
 
     <span class="c"># the checked base can IS the by-angle variant; the two failure modes are</span>
@@ -712,16 +712,16 @@ <h2>Source</h2>
 
     <span class="c"># key/fill/rim/wedge per docs/VISUAL-STYLE.md</span>
     light(<span class="s">&quot;Key&quot;</span>, (-<span class="n">4.0</span>, -<span class="n">5.0</span>, <span class="n">6.0</span>), <span class="n">520.0</span>, <span class="n">4.5</span>, (<span class="n">1.0</span>, <span class="n">0.96</span>, <span class="n">0.9</span>), (<span class="n">50</span>, <span class="n">0</span>, -<span class="n">38</span>))
-    light(<span class="s">&quot;Fill&quot;</span>, (<span class="n">5.5</span>, -<span class="n">3.5</span>, <span class="n">2.5</span>), <span class="n">120.0</span>, <span class="n">9.0</span>, (<span class="n">0.75</span>, <span class="n">0.85</span>, <span class="n">1.0</span>), (<span class="n">65</span>, <span class="n">0</span>, <span class="n">55</span>))
+    light(<span class="s">&quot;Fill&quot;</span>, (<span class="n">5.5</span>, -<span class="n">3.5</span>, <span class="n">2.5</span>), <span class="n">150.0</span>, <span class="n">9.0</span>, (<span class="n">0.75</span>, <span class="n">0.85</span>, <span class="n">1.0</span>), (<span class="n">65</span>, <span class="n">0</span>, <span class="n">55</span>))
     light(<span class="s">&quot;Rim&quot;</span>, (<span class="n">1.5</span>, <span class="n">4.5</span>, <span class="n">4.0</span>), <span class="n">340.0</span>, <span class="n">3.0</span>, (<span class="n">0.6</span>, <span class="n">0.78</span>, <span class="n">1.0</span>), (-<span class="n">58</span>, <span class="n">0</span>, <span class="n">170</span>))
     light(<span class="s">&quot;Wedge&quot;</span>, (<span class="n">2.5</span>, <span class="n">5.5</span>, <span class="n">4.0</span>), <span class="n">400.0</span>, <span class="n">6.0</span>, (<span class="n">1.0</span>, <span class="n">0.76</span>, <span class="n">0.5</span>), (-<span class="n">68</span>, <span class="n">0</span>, <span class="n">190</span>))
     <span class="c"># the shading audit light: a tall strip whose reflection runs down each</span>
     <span class="c"># can face — it stair-steps on flat shading, warps at the rim on</span>
     <span class="c"># smooth-everything, and stays straight with crisp edges on by-angle</span>
     ld = bpy.data.lights.new(<span class="s">&quot;Strip&quot;</span>, <span class="s">&#x27;AREA&#x27;</span>)
-    ld.energy = <span class="n">460.0</span>
+    ld.energy = <span class="n">400.0</span>
     ld.shape = <span class="s">&#x27;RECTANGLE&#x27;</span>
-    ld.size = <span class="n">1.0</span>
+    ld.size = <span class="n">1.35</span>
     ld.size_y = <span class="n">8.0</span>
     ld.color = (<span class="n">0.9</span>, <span class="n">0.95</span>, <span class="n">1.0</span>)
     strip = bpy.data.objects.new(<span class="s">&quot;Strip&quot;</span>, ld)
@@ -743,7 +743,7 @@ <h2>Source</h2>
 
     scene.render.engine = <span class="s">&#x27;CYCLES&#x27;</span> <span class="k">if</span> engine == <span class="s">&#x27;cycles&#x27;</span> <span class="k">else</span> eevee_engine_id()
     <span class="k">if</span> engine == <span class="s">&#x27;cycles&#x27;</span>:
-        scene.cycles.samples = <span class="n">48</span>
+        scene.cycles.samples = <span class="n">64</span>
     <span class="k">else</span>:
         <span class="k">try</span>:
             scene.eevee.taa_render_samples = <span class="n">64</span>

docs/gallery/index.html

@@ -541,8 +541,8 @@ <h2><a href="light-link-studio/">light-link-studio</a></h2>
         </a>
         <div class="card-body">
           <h2><a href="collision-hull-proxy/">collision-hull-proxy</a></h2>
-          <p class="teaches">A fire hydrant street prop inside its compound collision shell: four convex pieces hulled by bmesh.ops.convex_hull from a coarse inflated cage. The dense render mesh is never hulled - its hull would measure 368 faces, over the 255-face per-piece engine budget. Closed-form plane tests prove containment, convexity, watertightness, outward winding, and Euler characteristic 2 per piece.</p>
-          <p class="witnesses"><span class="tag">witnesses</span> A hull of the dense render mesh measures 368 faces, over the 255-face per-piece budget - the cage is a coarser lathe with rings inflated by sec(pi/n) so each cage ring circumscribes its render ring exactly (containment 5.9e-08). Proud details cost cage rows; concave grooves are free under the hull. Piece budgets: body 70, caps 60+60+60, compound 250.</p>
+          <p class="teaches">A fire hydrant street prop inside its compound collision shell: four convex pieces hulled by bmesh.ops.convex_hull from a coarse inflated cage. The dense render mesh is never hulled - its hull would measure 380 faces, over the 255-face per-piece engine budget. Closed-form plane tests prove containment, convexity, watertightness, outward winding, and Euler characteristic 2 per piece.</p>
+          <p class="witnesses"><span class="tag">witnesses</span> A hull of the dense render mesh measures 380 faces, over the 255-face per-piece budget - the cage is a coarser lathe with rings inflated by sec(pi/n) so each cage ring circumscribes its render ring exactly (containment 4.4e-08). Proud details cost cage rows; concave grooves are free under the hull. Piece budgets: body 70, caps 60+60+60, compound 250.</p>
           <a class="card-link" href="collision-hull-proxy/">View example <span aria-hidden="true">&rarr;</span></a>
         </div>
       </article>

examples/collision-hull-proxy/README.md

@@ -24,27 +24,27 @@ per-piece face limit (255 is the common cap). The example builds that
 structure and derives every property from closed forms, per piece:
 
 - **Containment.** Every render-mesh vertex lies on the inner side of every
-  face plane of its piece's hull — max excursion **5.9e-08** (tol 2e-4). This
+  face plane of its piece's hull — max excursion **4.4e-08** (tol 2e-4). This
   holds by construction, not luck: the cage is a coarse lathe whose rings are
   inflated by `sec(π/n)`, so each n-gon cage ring *circumscribes* its render
   ring exactly. Proud details (a rim, a nut) must be paid for in cage rows;
   concave details (the grooves) are free — the hull bridges over them. That
   trade-off is the collision-authoring lesson.
 - **Convexity.** The same plane test restricted to the hull's own vertices
-  (5.0e-08) — what a convex solver assumes when it uploads the piece.
+  (4.8e-08) — what a convex solver assumes when it uploads the piece.
 - **Watertight + manifold.** Every hull edge borders exactly two faces, and
   V − E + F = 2: a convex hull is a topological sphere.
 - **Outward winding.** Positive signed volume by the divergence theorem
-  (body piece 0.968858) — inverted collision is a physics no-op.
+  (body piece 1.022381) — inverted collision is a physics no-op.
 - **Budget.** Every piece ≤ 255 faces: body 70, pumper 60, side caps 60+60;
   the compound totals 250. The naive approach fails this — a hull of the
-  *dense render mesh* measures **368 faces**, over budget, which is why real
+  *dense render mesh* measures **380 faces**, over budget, which is why real
   pipelines hull a cage, never the render mesh.
 
 **What each check catches on failure:** an artist editing the render mesh
 after the hull was generated (probe: pumper lug stretched 2×, exit 3,
 measured escape 0.730000); inverted hull winding (probe: polygons flipped,
-exit 5, signed volume −0.968858, the exact negation); a non-watertight piece
+exit 5, signed volume −1.022381, the exact negation); a non-watertight piece
 (probe: one face deleted, exit 4, three border edges).
 
 **Version witness:** output is byte-identical on Blender 4.5.11 LTS and

examples/collision-hull-proxy/collision_hull_proxy.py

@@ -62,15 +62,17 @@
     (0.37, 0.205),   # neck
     (0.35, 0.320),   # barrel
     (0.346, 0.545),
-    (0.334, 0.555),  # barrel groove (concave)
-    (0.334, 0.575),
+    (0.328, 0.555),  # barrel groove (concave)
+    (0.328, 0.575),
     (0.346, 0.590),
     (0.345, 0.880),  # barrel top
     (0.348, 0.930),
     (0.336, 0.950),  # upper groove (concave)
     (0.336, 1.000),
     (0.348, 1.030),
     (0.36, 1.090),   # shoulder under bonnet
+    (0.388, 1.110),  # bonnet brim (proud — paid for with a cage row)
+    (0.362, 1.132),  # under-brim tuck (concave)
     (0.35, 1.150),   # bonnet dome (concave curve — cage keeps every row)
     (0.30, 1.220),
     (0.21, 1.280),
@@ -87,14 +89,15 @@
     (0.37, 0.205),
     (0.345, 0.880),
     (0.36, 1.090),
+    (0.388, 1.110),  # the brim is proud: it must appear in the cage
     (0.35, 1.150),
     (0.30, 1.220),
     (0.21, 1.280),
     (0.10, 1.315),
     (0.00, 1.325),
 ]
 
-NUT_PROFILE = [(0.09, 1.325), (0.072, 1.40), (0.0, 1.40)]
+NUT_PROFILE = [(0.115, 1.325), (0.095, 1.44), (0.0, 1.44)]
 
 # Side caps at z=0.76, pumper (bigger) at z=0.66. A cap's outlet direction
 # equals its azimuth (degrees from +X); the camera sits near azimuth 292, so
@@ -198,6 +201,7 @@ def build_hydrant():
     red = make_material("EnamelRed", (0.48, 0.035, 0.022), 0.15, 0.30)
     yellow = make_material("EnamelYellow", (0.82, 0.53, 0.05), 0.15, 0.32)
     iron = make_material("CastIron", (0.05, 0.052, 0.06), 0.9, 0.48)
+    iron_light = make_material("CastIronLight", (0.14, 0.145, 0.16), 0.85, 0.42)
 
     inflate_body = 1.0 / math.cos(math.pi / CAGE_SEG_BODY)
     inflate_cap = 1.0 / math.cos(math.pi / CAGE_SEG_CAP)
@@ -211,7 +215,7 @@ def build_hydrant():
         if p.center.z > 1.09:
             p.material_index = 1
     nut = lathe_object("Nut", NUT_PROFILE, 5)
-    nut.data.materials.append(iron)
+    nut.data.materials.append(iron_light)
     body_cage = lathe_object("BodyCage", CAGE_BODY_PROFILE, CAGE_SEG_BODY, inflate_body)
     nut_cage = lathe_object("NutCage", NUT_PROFILE, CAGE_SEG_NUT, inflate_nut)
     groups = [{"name": "body", "render": [body, nut], "cage": [body_cage, nut_cage]}]
@@ -439,7 +443,7 @@ def light(name, loc, energy, size, col, rot):
 
     # key/fill/rim/wedge per docs/VISUAL-STYLE.md
     light("Key", (-3.5, -4.5, 5.5), 500.0, 4.5, (1.0, 0.96, 0.9), (48, 0, -35))
-    light("Fill", (5.0, -3.5, 2.5), 110.0, 9.0, (0.75, 0.85, 1.0), (65, 0, 50))
+    light("Fill", (5.0, -3.5, 2.5), 130.0, 9.0, (0.75, 0.85, 1.0), (65, 0, 50))
     light("Rim", (1.5, 4.5, 3.5), 300.0, 3.0, (0.6, 0.78, 1.0), (-55, 0, 170))
     light("Wedge", (2.5, 5.5, 4.0), 380.0, 6.0, (1.0, 0.76, 0.5), (-68, 0, 190))
     # small warm glint off the camera-left shoulder: lifts the pumper cap face
@@ -459,7 +463,7 @@ def light(name, loc, energy, size, col, rot):
 
     scene.render.engine = 'CYCLES' if engine == 'cycles' else eevee_engine_id()
     if engine == 'cycles':
-        scene.cycles.samples = 48
+        scene.cycles.samples = 64
     else:
         try:
             scene.eevee.taa_render_samples = 64

examples/custom-normals-shade/custom_normals_shade.py

@@ -451,7 +451,7 @@ def render_still(can, path, engine):
     # semi-gloss painted steel: rough enough to read as paint, glossy enough
     # that a highlight exposes every normal discontinuity — matte would hide
     # the very shading differences this render exists to show
-    steel = make_material("OliveDrab", (0.31, 0.35, 0.15), 0.8, 0.26)
+    steel = make_material("OliveDrab", (0.30, 0.345, 0.13), 0.8, 0.26)
     accent = make_material("FuelMarker", (0.55, 0.06, 0.03), 0.3, 0.30)
 
     # the checked base can IS the by-angle variant; the two failure modes are
@@ -504,16 +504,16 @@ def light(name, loc, energy, size, col, rot):
 
     # key/fill/rim/wedge per docs/VISUAL-STYLE.md
     light("Key", (-4.0, -5.0, 6.0), 520.0, 4.5, (1.0, 0.96, 0.9), (50, 0, -38))
-    light("Fill", (5.5, -3.5, 2.5), 120.0, 9.0, (0.75, 0.85, 1.0), (65, 0, 55))
+    light("Fill", (5.5, -3.5, 2.5), 150.0, 9.0, (0.75, 0.85, 1.0), (65, 0, 55))
     light("Rim", (1.5, 4.5, 4.0), 340.0, 3.0, (0.6, 0.78, 1.0), (-58, 0, 170))
     light("Wedge", (2.5, 5.5, 4.0), 400.0, 6.0, (1.0, 0.76, 0.5), (-68, 0, 190))
     # the shading audit light: a tall strip whose reflection runs down each
     # can face — it stair-steps on flat shading, warps at the rim on
     # smooth-everything, and stays straight with crisp edges on by-angle
     ld = bpy.data.lights.new("Strip", 'AREA')
-    ld.energy = 460.0
+    ld.energy = 400.0
     ld.shape = 'RECTANGLE'
-    ld.size = 1.0
+    ld.size = 1.35
     ld.size_y = 8.0
     ld.color = (0.9, 0.95, 1.0)
     strip = bpy.data.objects.new("Strip", ld)
@@ -535,7 +535,7 @@ def light(name, loc, energy, size, col, rot):
 
     scene.render.engine = 'CYCLES' if engine == 'cycles' else eevee_engine_id()
     if engine == 'cycles':
-        scene.cycles.samples = 48
+        scene.cycles.samples = 64
     else:
         try:
             scene.eevee.taa_render_samples = 64

examples/gallery.json

@@ -373,8 +373,8 @@
     {
       "name": "collision-hull-proxy",
       "dir": "examples/collision-hull-proxy",
-      "teaches": "A fire hydrant street prop inside its compound collision shell: four convex pieces hulled by bmesh.ops.convex_hull from a coarse inflated cage. The dense render mesh is never hulled - its hull would measure 368 faces, over the 255-face per-piece engine budget. Closed-form plane tests prove containment, convexity, watertightness, outward winding, and Euler characteristic 2 per piece.",
-      "witnessesFix": "A hull of the dense render mesh measures 368 faces, over the 255-face per-piece budget - the cage is a coarser lathe with rings inflated by sec(pi/n) so each cage ring circumscribes its render ring exactly (containment 5.9e-08). Proud details cost cage rows; concave grooves are free under the hull. Piece budgets: body 70, caps 60+60+60, compound 250.",
+      "teaches": "A fire hydrant street prop inside its compound collision shell: four convex pieces hulled by bmesh.ops.convex_hull from a coarse inflated cage. The dense render mesh is never hulled - its hull would measure 380 faces, over the 255-face per-piece engine budget. Closed-form plane tests prove containment, convexity, watertightness, outward winding, and Euler characteristic 2 per piece.",
+      "witnessesFix": "A hull of the dense render mesh measures 380 faces, over the 255-face per-piece budget - the cage is a coarser lathe with rings inflated by sec(pi/n) so each cage ring circumscribes its render ring exactly (containment 4.4e-08). Proud details cost cage rows; concave grooves are free under the hull. Piece budgets: body 70, caps 60+60+60, compound 250.",
       "hero": "docs/gallery/assets/collision-hull-proxy-hero.webp",
       "preview": "examples/collision-hull-proxy/preview.webp",
       "tags": [