Skip to content

Independent tessellation: zoo fixtures and sample-exact UV fill - #61

Draft
jmossymoss wants to merge 7 commits into
cursor/independent-tessellate-b00cfrom
cursor/independent-fixture-visual-46d0
Draft

jmossymoss wants to merge 7 commits into
cursor/independent-tessellate-b00cfrom
cursor/independent-fixture-visual-46d0

Conversation

@jmossymoss

@jmossymoss jmossymoss commented Aug 21, 2026

Copy link
Copy Markdown
Owner

Independent tessellation on zoo fixtures, then MP9.

Tried OCCT’s tessellator as the leftover-face fill. On cylinders it only refines around (zero curvature along the generator), so you get height-spanning slivers and lost rim samples. Slotted went from 0 opens to 432. Leftover faces stay on the sample-exact UV fill; OCCT IncrementalMesh remains last resort.

Structured fills (planar n-gon, drum loft, 4-sided Coons) still own the zoo primitives.

Open in Web Open in Cursor 

cursoragent and others added 3 commits August 21, 2026 20:08
Full-period tori were ear-clipped in periodic UV and collapsed to a fan.
Iso-band UV grids ignored exact rim samples so fillets and tubes leaked.
Failed UV fills fell back to a triangle fan that covered holes.

Use a UV lattice on doubly-periodic charts, loft closed drums from shared
rim samples, blend 4-sided patches in 3D, and send failed fills to OCCT.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
Closed-cylinder winding follows the face-edge sample order so caps weld
without a global flip heuristic. A full torus uses both closed-edge
counts, so the tube is not a 4-sided washer.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
Trimmed drums (MP9-class cylinders with many hole wires) were dumping to
OCCT after ear-clip failed, which dropped shared rim samples and opened
thousands of cracks. UV fill now Delaunay-webs the sample loops in the
parameter plane, bridges holes without crossing, and only force-clips
unholed rings. Loft and Coons refuse unequal opposite counts instead of
resampling them away.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
@cursor cursor Bot changed the title Independent tessellation: fix torus, drums, and folded 4-sided patches Independent tessellation: zoo fixtures and sample-exact UV fill Aug 21, 2026
cursoragent and others added 4 commits August 21, 2026 21:12
Drum-loft plus OCCT fallback had opened ~8k cracks on mp9_Edited; the
trimmed-drum class is back near the prior UV-fill baseline with fewer
folds. MP9 remains T4 evidence, not a T0 gate.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
Unholed faces were ear-clipped into sliver fans; the OCCT planar mesher
was only invoked when holes existed. Run that CDT on every sample loop
first, drop the force-clip fan, and keep interior Steiner points on the
CAD surface.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
A UV-polygon CDT still fans: the parameter plane has no curvature, so
OCCT never inserts interior points. Run IncrementalMesh on the CAD face
first (snap rims to shared samples) and keep the UV CDT only if that
emits nothing.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
OCCT IncrementalMesh on drums only splits around: generators are straight,
so it emits height-spanning slivers and misses shared rim samples (slotted
opened 432 edges). Use it as last resort, not the default fill.

Co-authored-by: jmossymoss <jmossymoss@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants