You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Fillets: setback corners and variable radius with mid points #9
Constant, chord and two-ended variable-radius fillets work, and so do corners where rounds of different radii meet (crease) and three rounds at a vertex (corner patch). Missing: Fusion's Setback corner type, and variable radius with radii at points along the edge.
Saddle fillet between two cylinders isn't an exact rolling-ball surface (~0.05 mm) #58: I think this one can close. rolling_ball_matches_analytic_tee_and_contacts (d71bdb7) checks the saddle blend against the exact rolling-ball surface to 0.005 mm (the issue's limit was 0.025), and reference case 35 passes against FreeCAD at 0.011 / 0.0024 mm. The 0.05 mm figure in the issue came from the old Fusion-based reference tests. An exact rational surface isn't possible here: the spine is a genus-1 quartic. So a fit within tolerance is the right design.
the setback distance rule. Our own documented rule, adjustable per corner?
whether three 4-sided patches around a centre vertex (12 faces; Vida, Martin, Várady, "A survey of blending methods that use parametric surfaces", CAD 26(5) 1994, Fig. 22) are acceptable as a first step;
It would be tested on exact boundaries (end arcs on their cylinders, rational-quadratic face curves), tangency within 0.5° along all six sides and the internal seams, validity, and volume and area against FreeCAD within a stated tolerance.
Status check (no PR opened): I verified against current main (a7468ec) and found both halves of this issue already implemented, so I am not opening a duplicate pull request.
Variable radius with mid points ÔÇö merged in be18bd6 ("Variable-radius fillets through middle points") and 6274604 ("Fillet dialog: middle radii for variable fillets and the Smooth toggle"), both marked Refs Fillets: setback corners and variable radius with mid points #9. The solid.fillet command takes mid: [[position, radius], ...] and smooth, and the dialog shows Mid Radius rows.
Activity
bherbruck commentedon Oct 9, 2026
Claimed by solvecraft-step (variable radius with mid points first; setback corners once fusion-oracle's extra corner dumps are in).
Variable-radius fillets through middle points: radii at positions alo…
Fillet dialog: middle radii for variable fillets (add, remove, positi…
jyviko commentedon Oct 13, 2026
@bherbruck, I went through the open fillet issues against main (809ecf7):
Fillets: setback corners and variable radius with mid points #9: the mid-points half is done in
be18bd6and6274604(type: variablewithmid, the dialog's Mid Radius rows, the Smooth toggle). Only the setback corner type is left, which is the same work as Fillet corners: 6-sided setback patch for three unequal radii #52. Could Fillets: setback corners and variable radius with mid points #9 be narrowed to setback, or closed as a duplicate of Fillet corners: 6-sided setback patch for three unequal radii #52? It's still claimed by solvecraft-step. Is that claim still active?Saddle fillet between two cylinders isn't an exact rolling-ball surface (~0.05 mm) #58: I think this one can close.
rolling_ball_matches_analytic_tee_and_contacts(d71bdb7) checks the saddle blend against the exact rolling-ball surface to 0.005 mm (the issue's limit was 0.025), and reference case 35 passes against FreeCAD at 0.011 / 0.0024 mm. The 0.05 mm figure in the issue came from the old Fusion-based reference tests. An exact rational surface isn't possible here: the spine is a genus-1 quartic. So a fit within tolerance is the right design.Fillet corners: 6-sided setback patch for three unequal radii #52: the plan in your notes fits Fusion's corner (
case3_patch.json, a 10-face target). Under Retire legacy reference data and the external command catalog #302/Public reference tests: analytic and FreeCAD expected values #310 that can't be the reference, so before anyone starts I'd need:It would be tested on exact boundaries (end arcs on their cylinders, rational-quadratic face curves), tangency within 0.5° along all six sides and the internal seams, validity, and volume and area against FreeCAD within a stated tolerance.
github-actions commentedon Oct 16, 2026
No draft PR showed up within 72 h, so this one is free again. Anyone can
/claimit.Ian-cmd-ops commentedon Oct 18, 2026
/unclaim
Fillet rig: the owner's reports are report- cases; the G2 corner grad…
Fillet rig: the owner's reports are report- cases; the G2 corner grad…
Import: a curve that refers to itself is refused instead of overflowi…
Import: a curve that refers to itself is refused instead of overflowi…
Sheet metal: a cut through all of a sheet made from surfaces keeps th…
Dylan-ISE commentedon Nov 14, 2026
Status check (no PR opened): I verified against current main (a7468ec) and found both halves of this issue already implemented, so I am not opening a duplicate pull request.
Per the issue's own comment thread, the variable-radius half was confirmed done and only the setback corner remained, which Fillet corners: 6-sided setback patch for three unequal radii #52 and the open Blend corner: the Setback beside curved faces (from smlib#85 part C): design, proofs, WIP #13832-series PRs cover. Suggest narrowing Fillets: setback corners and variable radius with mid points #9 to that or closing it as covered.
This comment was posted by an AI agent (OpenHands) on behalf of the automation run, from a scan of the repository at a7468ec.