Kernel tests: a law's dip shows its corner ball's sphere, a peak's two rounds cross (the reasons had them swapped)

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

crates/kernel/src/blend_variable.rs

@@ -575,7 +575,7 @@ mod tests {
     }
 
     /// A radius through a middle point, straight pieces 2 → 4 → 2: a corner in the law, where the
-    /// ball there shows between its two sides' touch circles; the balls' union takes 39.541034
+    /// two sides' rounds cross and each ends where they do; the balls' union takes 39.541034
     /// (the halves' great circles would take 40.0590).
     #[test]
     fn variable_radius_with_a_corner_through_a_middle_point() {

crates/smlib/src/ops_tests.rs

@@ -319,8 +319,8 @@ fn a_variable_round_follows_its_law() {
 }
 
 /// A law with a corner (4 → 2 → 4, straight on both halves) on the 20 cube's edge y = z = 20: the
-/// union of the balls, where the two halves' rounds cross and each ends, takes 39.542073 (the
-/// halves' great circles would take 40.0590).
+/// union of the balls, with the corner ball's sphere between the halves' touch circles, takes
+/// 39.542073 (the halves' great circles would take 40.0590).
 #[test]
 fn a_variable_round_with_a_corner_in_its_law_is_its_balls_union() {
     let c = cube(20.0);

tests/KNOWN_FAILING_TESTS

@@ -15,8 +15,8 @@ solidcraft-kernel::blend::corner_tests::a_round_ending_in_an_earlier_round_match
 solidcraft-kernel::blend::corner_tests::corner_blends_hold_at_any_scale at 0.01x the arc case's rounds remove 72.783 (scaled back) against 72.713 at 1x (9.6e-4 off, 1e-4 allowed)
 solidcraft-kernel::blend_corner::tests::issue2422_vertical_after_separate_bottom_rounds SMLib's check refuses the fillet: edge and vertex tolerance 0.97 against the body's 0.01 (#2422)
 solidcraft-kernel::blend_corner::tests::three_rounds_at_a_corner the corner's variable-radius transition between rounds of different radii (its elliptic law, a rail edge the corner interpolates) meets its neighbours at 0.144 degrees once its sections are the balls' touch circles (0 with the great circles); the corner step replaces that transition
-solidcraft-kernel::blend_variable::tests::variable_radius_with_a_corner_through_a_middle_point the two halves' rounds end on their own touch circles at the law's corner and the ball's sphere between them is not built: 39.8302 removed against 39.5410
+solidcraft-kernel::blend_variable::tests::variable_radius_with_a_corner_through_a_middle_point the two halves' rounds are joined from their own end circles at the law's corner, not where they cross: 39.8302 removed against 39.5410
 solidcraft-kernel::corner_g2::tests::a_long_setback_keeps_the_g2_patch_inside_the_box at set backs 11 to 15 the G2 corner patch bulges 2.6e-6 to 1.33e-3 past the box (the explicit corner is in progress)
 solidcraft-kernel::corner_g2::tests::the_box_corner_g2_patch_is_a_closed_solid_curvature_continuous_to_its_neighbours the fitted G2 corner patch bulges 1.7e-5 past the box, which SMLib's exact tight box now sees (the explicit G2 corner builds it inside)
 solidcraft-kernel::tests::fillet_tabbed_plate_outer_loop_and_partial_selection_after_hole the arc and tab rounds together remove 9.20067 against 9.20420 (3.8e-4 off, 1e-5 allowed)
-solidcraft-smlib::ops::tests::a_variable_round_with_a_corner_in_its_law_is_its_balls_union the two halves' rounds end on their own touch circles at the law's corner, not where they cross: 39.6456 removed against 39.5421
+solidcraft-smlib::ops::tests::a_variable_round_with_a_corner_in_its_law_is_its_balls_union the two halves' rounds are joined from their own end circles at the law's corner; the ball's sphere between its two touch circles is not built: 39.6456 removed against 39.5421