Fillet rig: 1668's R7 past the inside corner is refused at 4 + 2 sqrt2

The round's centre (32 - r, 42 - r) leaves the hollow's corner (30, 40)
outside its circle, opening the wall, where sqrt2 (r - 2) = r: the case
expects the refusal naming 6.828. 2852's R30, refused naming 25.24, now
passes too: both come off the known-failing list.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/kernel-cases/fillets/1668-fillet-outer-top-edge-of-thin-wall-r7-past-inside-corner/expected.json

@@ -3,15 +3,9 @@
   "tangency-labs/solidcraft#1668",
   "tangency-labs/solidcraft#1080"
  ],
- "what": "an R7 round of the outer top edge of a 2 mm wall alone: past the sharp inside corner it is cut from the wall as if the hollow were not there, opening a slit along the hollow",
+ "what": "an R7 round of the outer top edge of a 2 mm wall alone: past the sharp inside corner it would cut through the wall into the hollow, opening a slit along it, so it is refused, naming the largest radius that fits",
  "source": "analytic",
- "method": "box 100x60x40 shelled outward 2 mm open at the bottom: 104x64x42 less 100x60x40 = 39552. The round's circle about (y, z) = (25, 35), r 7, takes (1-pi/4) 49 along the edge's 104, less its part over the hollow (y < 30, z < 40), the square [25,30]x[35,40] outside the circle: Int_sqrt24^5 (5 - sqrt(49-u^2)) du = 5(5-sqrt24) - (F(5) - F(sqrt24)), F(t) = (t sqrt(49-t^2) + 49 asin(t/7))/2, = 0.005068, along the hollow's 100 (the end walls x = +-50..52 are solid). V = 38458.895826, exact: the tolerance holds it to 0.04 mm^3",
- "body_count": 1,
- "volume_mm3": 38458.89582646349,
- "tolerance": {
-  "rel": 1e-06
- },
- "continuity": {
-  "order": "g1"
- }
+ "method": "box 100x60x40 shelled outward 2 mm open at the bottom: the wall runs y 30..32, z 40..42 at the top edge (y, z) = (32, 42). The round's centre is at (32 - r, 42 - r); it removes what lies between the edge and its arc. The hollow's corner (30, 40) lies at sqrt2 (r - 2) from the centre: the arc uncovers it, and the section breaks into the hollow along its 100, where sqrt2 (r - 2) = r, r = 4 + 2 sqrt2 = 6.8284. Past that the wall would open (the owner's ruling of 2026-10-30: refused as too large, naming the largest that fits, 6.828).",
+ "outcome": "refused",
+ "refusal": "the largest radius that fits is 6.828 mm"
 }

tests/kernel-cases/fillets/KNOWN_FAILING

@@ -7,9 +7,7 @@
 1080-fillet-sloped-basin-1-mm-rim-r4-2-edges not supported yet: a convex round past its face is not yet built by subtraction
 1080-fillet-sloped-basin-1-mm-rim-r4-4-edges not supported yet: a convex round past its face is not yet built by subtraction
 1207-fillet-bar-top-edge-after-boss-arc-round-r0-5 SMLib's check refuses the round: internal coincident control points in a B-spline curve (an internal pole)
-1668-fillet-outer-top-edge-of-thin-wall-r7-past-inside-corner a slit along the hollow, 0.507 mm^3 off: 38458.389 is V0 less the round's whole section along 104, so the section's part over the hollow (0.005068 mm^2 along 100), which is not material, is taken anyway
 190-fillet-disc-rim-past-its-2mm-wall SMLib status 1001: a convex round past its face is not yet built by subtraction
-2852-fillet-outer-round-r30-past-inner-round-refused refused for another reason: SMLib's result is not a closed solid, not too large naming 25.24
 2893-chamfer-distance-and-angle-past-the-wall-refused refused for another reason: SMLib SM_ERR_INVALID_INPUT, not a check saying the angle runs past the wall
 2893-chamfer-two-distances-elliptical-rim SMLib status 1001 on the elliptical rim's chamfer
 2893-chamfer-two-distances-past-the-bottom SMLib status 1001: SmFilletNx1ClosedCorner cannot compute the closed rim's corner geometry, where the chamfer past the bottom rim should build the frustum