[bluecmd]

Follow-up to #257 (lesson: wire back-edges when the owner lands). topology.brep_p7 has been on main since #246, so heal::stitch::p1's SmBrep::GlueEdgesGeneral wrapper (UNPORTED(topology.brep_p7) on main, heal/stitch/p1.rs:2013) now calls the real routine.

What it does

• glue_edges_general (the stitch module's wrapper) takes the kernel and calls topology::brep::p7::glue_edges_general(k, b, Some(keep), orientation, dist, do_region_nesting, Some(delete), inside_faceuses, &mut env). Upstream notifies the B-rep's own SmTopology objects through its context; the heal path keeps no records, so env is a default record per arena object (brep_p2::ArenaNotifyState::new(b), the same seeding brep p5/p6 use) over the kernel's notification context (Context::with_notification_context): SmTopology::Notify's propagation finds both edges, and callbacks installed on the kernel see the events (Reviewer A pre-review must-fixes 1 and 2; an empty record map made the MergeInBrep notify fail silently, and context: None dropped the callbacks).
• All three call sites pass k: DoSimpleFaceStitch (cpp:1005/1013, status propagated) and StitchEdgesOfVertices (cpp:4145, success tested) in heal::stitch::p1, and DoStitching's edge gluing at a vertex (cpp:2151/2183) in heal::stitch::p2, which imports the same wrapper. Arguments unchanged: pInsideFaceuses stays NULL.
• The header's unported list and the modules.toml note are updated. Since port: topology.brep_p2 #149 (batch batch: land 1 reviewed-clean contributor PR #315) part 7's notifies are live, so the wrapper no longer stops at a stub of its own: a wire edge to delete stops at UNPORTED(topology.brep_p4): SmBrep::DeleteEdge, and face-bounded edges run the glue up to part 7's own stubs. origin/main (e5f3e44, with port: topology.brep_p2 #149) is merged in; the merge conflicts were this module's header list and the modules.toml note.

Coverage, plainly

• No heal.stitch_p1 or heal.stitch_p2 difftest case reaches the glue today: all call sites sit behind earlier deferrals (topology.face_p2 SmFace::GetTolerance, topology.face_p4, query.brep, topology.brep_p1), so both reports are unchanged from main. The wiring is pinned by two unit tests with a recording notify callback on the kernel: two wire edges between the same vertices see the B-rep PreEdit and then return Unported { upstream: "SmBrep::DeleteEdge", unit: "topology.brep_p4" } from inside the part 7 routine, with the callback restored on the kernel afterwards (the same edge twice is the upstream no-op); two face-bounded edges reach the MergeInBrep notify, which the seeded records let run.

Gate (sc-builder-0, on c338dbd)

• fmt OK, clippy (plain and cpp) 0 errors, cargo test -p solidcore 1528 passed, wasm OK.
• difftest heal.stitch_p1 / heal.stitch_p2: both byte-identical to main's committed reports (post-port: topology.brep_p2 #149):

  PARTIAL heal.stitch_p1 cases=300 passed=232 failed=0 deferred=68 max_rel_err=0e0 with_values=232 no_values=0 blocked_by=harness:analytic-geometry-export(1),query.brep(10),topology.brep_p1(5),topology.brep_p4(1),topology.face_p2(49),topology.face_p4(2)
  PARTIAL heal.stitch_p2 cases=556 passed=332 failed=0 deferred=224 max_rel_err=0e0 with_values=331 no_values=1 blocked_by=intersect.curve_class_p2(11),query.brep(121),topology.brep_p1(92)
  

• --green: 234 units, failed=0.

🤖 Generated with Claude Code

https://claude.ai/code/session_01BfiWnWawYvdkvRXZtfC7zK

Status (updated 2026-10-21 10:23 UTC)

• Next: merge round for batch batch: land 2 reviewed-clean contributor PRs #364 done at 5a3b9413 (merge only). stitch_p1 300/234/0/66 and stitch_p2 556/332/0/224, identical to main's reports. Waiting on maintainer review.