Degenerate Bevel Weld

A runnable example isolating the half-dimension bevel collapse: author a bevel width against one box and reuse it on a thinner one, and at offset == min_dimension / 2 the bevel band pinches into zero-area faces — and those degenerate triangles cross the export boundary. Re-parsed from the shipped GLB with stdlib only, they are sitting in the file, where an engine-side merge-by-distance welds their loops. Found authoring gltf-export-roundtrip (its count check caught a 36-vertex weld on a thin crate part); this example isolates the threshold.

Pipeline arc neighbors: round-trip fidelity in gltf-export-roundtrip, topology gates in mesh-hygiene-audit, watertight parametric solids in bmesh-gear.

What it witnesses: a 1.6 × 0.4 × 1.0 slab (min dim 0.4, half = 0.2) with every edge beveled at 3 segments. All closed form or independently re-derived:

• Threshold. Offset 0.10 (< 0.2) yields zero zero-area faces; offset 0.20 (== min/2) yields exactly 12 == 4 min-axis edges × 3 segments. min_area collapses 3.2e5× (5.8e-04 → 1.8e-09).
• Collapse witness. Coincident-position verts == 16 == 4 edges × (segments + 1), re-derived by 6-decimal position grouping — the loops a merge-by-distance welds.
• Export crossing. A stdlib re-parse of the GLB recomputes every triangle area from the raw POSITION + indices buffers: 32 degenerate triangles ship (the 12 collapsed faces, triangulated — a MEASURED regression constant like curve-bevel-arc's EXPECT_VERTS; re-measure recipe in check()), versus 0 for the safe mesh. The exporter ships every loop (positions == 384 loops) — nothing warns you.

What each check catches on failure: safe offset pushed to the threshold (exit 3); degenerate offset backed below it (exit 4); coincident closed form corrupted (exit 6); GLB degenerate-tri expectation off (exit 8, e.g. if a future Blender changes triangulation — re-measure and update deliberately).

Version witness: output is byte-identical on Blender 4.5.11 LTS and 5.1.2 — same counts, same min_area values, same GLB triangle census.

Render as proof: two rugged hard-shell equipment cases whose shells are the check's two meshes — beveled_box(DIMS, 0.10) and beveled_box(DIMS, 0.20) — with every fitting (handle, steel draw latches with wire bails, rubber corner guards, molded frame, purge valve, ID plate, feet) kept inside the flat front land that survives both offsets, so the bevel is the only difference. The cases turn their end panels to the camera: the left keeps a flat end framed by clean chamfer bands; on the right the end and top lands are gone and the band rolls into a knife ridge at mid-depth. That collapsed seam is traced as a thin red-orange ink line (low emission, no neon halo) from live mesh data (every edge of every face thinner than 1e-6), with a small amber pin on each of the 12 zero-area faces the check counts — change the offset and the overlay moves or disappears. The clean case is the designed asset and passes the asset-quality floors (exit 11).

Run#

blender --background --python degenerate_bevel_weld.py --
blender --background --python degenerate_bevel_weld.py -- --both-safe
blender --background --python degenerate_bevel_weld.py -- --output bevel.png
blender --background --python degenerate_bevel_weld.py -- --output bevel.png --engine cycles

Exit codes#

Per-script sequential checks. 9 is a valid check code; there is no rule against it. 10 is the shared framing helper, 11 the shared asset-quality helper.

The blender-smoke workflow runs the check on Blender 5.2 LTS and 4.5 LTS (5.1 on the weekly cron, the needs-5.1 PR label, or manual dispatch). Smoke does not pass --output. Its catalog falsifier is --both-safe (expects exit 4).