Files
Brett Lawson 6e92e25b1e solidcore: retire fillet_overflow_trim_at_creases and fillet_overflow_crease_along_edge (WR-695, WR-717) (#17586)
* solidcore: claim retiring fillet_overflow_trim_at_creases and fillet_overflow_crease_along_edge

Part of #14246.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SdcgEtsWyH5fjgiavRsRdt

* solidcore: retire fillet_overflow_trim_at_creases and fillet_overflow_crease_along_edge

Part of #14246. WR-695: a cliff overflow, a WR-674/WR-706 section-ended
round and an open G1 setback corner are always trimmed at their creases
in place. The subtraction of the region the blends remove is deleted,
together with what only it reached: subtract_removal_region, the
session's application-attribute helpers, and the closing construction in
blend/fillet_cliff_overflow.rs (close_removal_region, lamina_hosts,
lamina_chains, closing_face, close_round_loop, close_band,
close_section_ended_round, covering_surface, lies_on_surface).

WR-717: a crease along an edge of the face it crosses is always that
edge. pieces_in_face_to_ends always returns the along-edge runs, and the
#17695 meet handling is kept.

Fields, SolidCraft preset lines, fixes.json entries, fixture opt-ins and
test kernel settings are removed. The phrase-site pins are lowered to
the refusals that remain (1401 -> 1345). Addenda for WR-695 and WR-717
are included.

The tests that compared the two paths, and the goldens rows that moved,
are in the next commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BUqX4Moqx5bbkSocgEbsSV

* solidcore: expectations for the overflow-crease flag retirement

Part of #14246. Test change: the tests that ran both paths keep the
trimmed path, and two goldens rows that reached the removed subtraction
are re-blessed after an old-vs-new body comparison.

| Test / row | Change | Evidence |
|---|---|---|
| wr695 #17673 variable round (2 tests -> 1) | flag-off comparison replaced by its pinned counts (7,15,10)/(7,14,9) | closed form kept: 2->50 rel 4.87e-10, 50->2 rel 9.26e-10 (<= 1e-7), gap <= 1e-12 (MEASURED) |
| issue_2385 without_the_flag_the_band_is_refused | removed with the flag-off path | crease_inside keeps V(r) closed form to 1e-6 (PROVEN by closed form) |
| issue_2715 subtraction refusal | removed with the subtraction | the trim and kept-feature cases stay |
| goldens big-radius-box_diff_cyl_big_rad-r28-edge12 | result_topology/result_geometry digests | same status, V/E/F 13/20/9; check_brep Level2 clean old and new; vertex sets within 1.5e-14; volume/centroid/inertia to 1e-12 rel, area 2.6e-10 (MEASURED); cone face UV domain [0,1]^2 instead of [0,0.0455]x[0,1] |
| goldens issue_2020_mitred_put_back_d9_full | result_topology/result_geometry digests | same status, V/E/F 8/12/6; check_brep Level2 clean; vertices within 1.4e-14; mass to 1e-12 rel (MEASURED); top planes keep their own UV domains instead of the Boolean's tight ones |

No tolerance changed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BUqX4Moqx5bbkSocgEbsSV

* solidcore: prove the issue_2020 put-back row and update stale overflow docs

Part of #14246. Review round 1 of #17586.

- New test issue_2020_mitred_put_back_trimmed_at_creases.rs proves the
  re-blessed goldens row issue_2020_mitred_put_back_d9_full in closed
  form: 8 vertices (top (3|37, 3|27, 1), bottom (a|40-a, a|30-a, 0),
  a = 3 - sqrt 5) to 4e-8, volume 816 + 116 S1 + 4 S2 to 1e-9 relative,
  V/E/F 8/12/6, manifold and check_brep Level2 clean.
- wr631: the subtraction-refusal test becomes
  a_round_a_hair_past_its_wall_builds_its_closed_form (r 10.01, 10.02
  must build a manifold solid of the closed-form volume to 1e-8). The
  dead refusal branch is dropped. The module doc and the _is_subtracted
  names (wr631, wr635_636) now describe the trim.
- Stale subtraction docs updated in issue_1980, wr635_636 and
  wr636_db4a655.
- Fixture 293: the notes describe the default kernel's refusal by the
  trim; the expectations are unchanged.
- WR-695 addendum: mass rows are stated as moved within the goldens'
  1e-9 rule, with the measured figures, and the default-kernel result
  changes are listed.

No expectation moves.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BUqX4Moqx5bbkSocgEbsSV

* WR-717 addendum: state the results inline

The flag-removal addendum pointed at the PR body for its results. A
permanent record should carry them itself: no goldens row moves because
of this flag (the two moved rows come from the removed subtraction and
are explained in the WR-695 addendum), and the tests that ran with the
flag off pass with it on.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BUqX4Moqx5bbkSocgEbsSV

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-16 21:01:06 +00:00
..