SMLib makes a fillet or chamfer face with no tag, so history::blends gives
each Generated face FaceOrigin::Edge for every blended edge whose two faces
both lie next to it and no longer meet directly. For the fillet shim's
history once it lands; a corner patch between blends keeps no origin.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
examples/smlib_booleans runs truck and SMLib on the same cases (named pairs
whose results are known by hand, then fuzz::boolean_shapes for N seeds),
each as union, intersection and cut, and compares volume, face and edge
counts and the shape of the history (how many faces Kept, Modified, Split,
Merged, Generated; edges from an operand edge, new, or without origin). It
writes a JSON and a Markdown report and never changes the default kernel.
SMLib runs in a child process per case, so a crash inside it is reported as
that case's result. Its primitives and booleans arrive with the operation
shims; until a call exists the case is reported as not run on SMLib.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brep::tagged gives a copy whose faces and edges are tagged as one operand's;
after an operation on tagged copies, Brep::history reads back each result
face's origins and change (Kept, Modified, Split, Merged, Generated) and each
edge's faces and origin edges, as SMLib itself carried them through the
operation's splits, merges, copies and coincidences. Raw::history turns that
into solidcraft_geom's History, renumbered into canonical order by each
body's Order, with a new edge recorded as where its two faces meet.
The FFI declarations are in src/ffi/history.rs, beside ffi.rs. Tests cover
the C layout, hostile records, the meeting edges, renumbering and bad
orders, and (with SMLib built) a tagged box.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new crate, solidcraft-step (L1, no modelling backend), holds the Part 21
parser (moved from the kernel, which now reads through it) and STEP geometry
as plain data: lines, conics, B-splines (rational and implied knots),
polylines, trimmed curves, surface curves with their pcurves, 3D offset
curves; planes, cylinders, cones, spheres, tori (degenerate too), B-spline
surfaces, revolutions, linear extrusions, offset and rectangular trimmed
surfaces. Each keeps STEP's own definition and parameterisation, in mm and
radians, with pcurves scaled by their surface's parameter units. Shared
entities stay shared. A writer emits the same entities back, deduplicated.
Tests per entity check values against analytic points, units, round trips
and hostile input; an ignored test reads and round-trips every edge curve
and face surface of the tangency-labs/models STEP files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ffi::line was a safe public function reading any raw pointer. The log
callback now lives in ffi.rs, reads the C string there, and hands the
crate a String through a safe sink, so no safe function takes a pointer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A new crate (L1, only the kernel uses it) builds tangency-labs/smlib's SMLib,
SM_API and C API with the cc crate when its `native` feature is on, from the
commit in crates/smlib/SMLIB_REV (scripts/fetch-smlib.sh fetches just those
sources). The kernel's `smlib` feature turns it on. Without it nothing C++
is built and every call returns NotBuilt, so workspace builds need no C++
toolchain.
unsafe is confined to src/ffi.rs. The wrapper owns handles, maps statuses to
errors, checks the C API version, and sends SMLib's per-call log to the log
crate. Crash reports note the native kernel call in flight.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SMLib makes narrowing cones with a negative half angle; schema 4 now takes
them as they are (r(v) = r0 + v tan a already covers them), with the
natural v range below the apex. The check that no region lists a shell
twice was quadratic per region (about 1e9 comparisons at the caps); it now
remembers the last region that listed each shell.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The adapter now saves each edge's curve record and interval when the edge first appears and
computes every face's pcurves from those records and the surface record it writes, with the
shared plain-record routine; the truck-face version in native/pcurve.rs is deleted. Loops go
in as the record keeps them (relative to the surface), so a reversed face no longer has its
wires reversed and back again.
The native tests are ported (adapter::face_alone gives one face's record for them). The cut
body's intersection pcurves now fit at 1e-6 mm too: the routine follows the converged
intersection, and truck's own points are within its 1e-6 mm of it, which the lift check
allows for on intersection edges. docs/native-brep.md says where pcurves come from and what
a surface intersection's point at t is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
solidcraft_geom::brep::pcurve::pcurves(surface, loops, tolerance) computes every edge use's
pcurve from our own records: the surface record and, per loop, each use's curve record,
interval and direction. No backend's parameter space is involved, so the truck adapter and an
SMLib conversion can share it (owner decision 2026-10-21).
brep::eval evaluates every schema 4 kind: planes, splines and revolutions through the display
evaluator (EvalSurface, EvalCurve, tess::uv), cylinders, cones, spheres and tori in closed
form in angle and height, and a surface intersection as the point of both surfaces in the
leader's normal plane (what truck's subs converges to), by Newton's method.
Exact in our parameters: lines, polylines and whole splines on planes; a circle about the axis
of a cylinder, cone, sphere or torus (a u-line); a line along a cylinder's or cone's ruling (a
v-line). Each exact pcurve is checked by lifting it; everything else is the cubic B-spline fit
the kernel had, with its span and round caps. Curve2::eval moves here from the kernel.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The parameter box, the projection of a point onto an evaluator surface (closed form on a
plane, damped Gauss-Newton elsewhere), unwrapping and the pole test work on geom's
EvalSurface alone. They move from the kernel's tessellator to solidcraft_geom::tess::uv so a
pcurve routine on plain B-rep records can use them without truck; the kernel re-exports them
and its tessellation is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A test over every design the repository keeps (the action matrix's
corpus: old design files, examples, regressions, demo and typical
designs, the oracle recipes; plus the sample designs, the engine's
design fixtures and the older designs of the hole matrix, taper and
projection tests: 154 designs). Each is rebuilt and recorded in
tests/naming/<group>/<design>.txt: per body the names of its faces and
edges with whether each follows exact history, and every stored name
(feature face picks, fillet/chamfer/bend edges, sheet metal edges, the
face a sketch is attached to) with the face or edge it finds now. A
difference fails with the design and the changed lines; record an
intended change with NAMING_CORPUS_BLESS=1, as FEATURE_WALK_BLESS does.
Face and edge indices are left out on purpose. They are the kernel's
order: truck already gives a fillet's or a boolean's faces in another
order from one process to the next (oracle 07 and 37 changed on every
run), and SMLib will order them differently again, while what must not
change is the names and what references find. Names are listed sorted.
One design's names themselves change between runs on truck
(closed_sweep_torus_cut: level wall halves numbered #1/#2 in kernel
face order); it is listed in UNSTABLE with the reason and its
difference printed instead of failed on.
Records made on this branch (truck), macOS. The whole corpus builds in
about 20 s on 14 cores (designs run side by side).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The corpus naming test (next commit) records what every stored face and
edge name resolves to, and must find it exactly as a rebuild does. Three
lookups were crate-private: `locate_face_on_body` (a feature's face
points, tried on the body the feature names first), `blend_target` (the
body a fillet or chamfer rounds) and `Document::anchored_face_indices`
(the faces a face-attached sketch is on). Visibility only; no behaviour
changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
docs/native-brep.md describes schema 4: the records with their SMLib and
STEP names, regions and shells, the loop convention, every edge's uses,
the canonical order (a reversed face walked backwards), each curve and
surface kind's parameterisation, pcurve rules, validation, history, the
SCBREP05 envelope and what the truck adapter writes and refuses.
A test lifts every pcurve of a box, a filleted box with a bore and a
sphere onto its face and checks it is its own edge at both ends and the
middle; with a reversed face's pcurves left in truck's oriented order it
fails by 30 mm.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
Edges, vertices and the body carry the face tolerance (10^-6 of the
body's size, at least 10^-5 mm): truck's traced intersection edges sit
about 1e-6 mm off their surfaces, so the linear tolerance is too tight for
a pcurve to lift onto them. The box-and-fillet golden fixture is the
schema 4 envelope now (NATIVE_BREP_BLESS=1 rewrites it), and the adapter
tests cover a pair of lumps. The model cache's producer revision becomes
solidcraft-truck-adapter/4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The plain B-rep follows SMLib's flat form with STEP's names: regions
(infinite first), face/wire/vertex shells, oriented faces, faces with a
domain and edge or vertex loops, oriented edges with a pcurve each, edges
with an interval, a tolerance and their list of uses, vertices with a
tolerance. Curves and surfaces are unbounded; their bounds are the edge
interval and the face domain. Loops are relative to the surface's natural
normal. Schema 3's read-only intersection and revolution records go.
The validator keeps schema 3's caps and adds the structural rules:
references, loop shape and connectivity, regions and closed shells, the
pcurve interval, natural parameter ranges, every edge's uses (primary
first, upper = !forward) and the canonical order of shells, faces, edges
and vertices. A reversed face is walked backwards, as truck's edge_iter
walks it, so truck's face and edge indices are the canonical ones.
The truck adapter writes truck's absolute wires as the loops, numbers
everything as the kernel does, and builds regions from the shells (a
tessellated nesting test only when there are several closed shells).
Pcurves come from pcurve::pcurves only; until it supports a face, encode
fails and the body is a cache miss. The envelope signature becomes
SCBREP05.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
FaceOrigin, Change, FaceProv, Side, EdgeOrigin, EdgeProv and History move
unchanged (fields, serde derives) to solidcraft-geom, so a backend can
report history without depending on solidcraft-kernel. The pure methods
(composition, renumbering, count checks, mark_shared_split) move with
them and return brep::history::Failed, which converts to
KernelError::Failed with the same "history: …" message.
What needs a kernel body (History::identity, check, face_sources) stays
in the kernel as the BodyHistory trait; solidcraft_kernel::history
re-exports the types, so existing paths keep working.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A band of a cylinder bounded by two full circles and no seam cannot close
in (u, v): each loop's pcurve runs one period. The test records that such
loops come back continuous and a period open (the caller needs a seam
there), and that a hostile tolerance is refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
Each sample of an intersection curve is one of truck's double projections
(about 0.13 ms in a release build), and a fit evaluates the curve at every
sample and halfway between. Starting from four samples a span instead of
eight halves that; the checks stay at eight points a span, and the knots
at the curve's own breaks already carry the shape. The cylinder cut by a
tilted box now takes 70 ms at 1e-3 mm and 149 ms at 1e-5 mm for all its
faces (release), from 183 and 638 ms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
Faces with seams as imported faces have them (the kernel's own
revolutions come in pieces): a cylinder whose seam line is used twice
between two full circles, each a closed edge on one vertex, and a sphere
bounded by one meridian used twice from pole to pole. Their seam uses come
out a period apart, in the base period. The checks now also require every
face's outer loop to turn counter-clockwise in (u, v) for a face along its
surface's normal and clockwise against it, holes the other way: the rule
the pole junctions rely on, which every test body follows.
With SOLIDCRAFT_PCURVE_SVG=<dir>, a test writes each test body's face
loops in (u, v) as SVG for review.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
Every edge not exact on a plane gets a cubic 2D B-spline on its curve's own
parameter: through the parameters of the curve's ends, least squares
through samples between (banded normal equations), with spans halved and
sampled twice as densely wherever the lift misses the curve by more than
half the tolerance, at the samples or halfway between them. Knots start
where the curve's parameterisation is less smooth: a spline's own knots
and the vertices of an intersection curve's leader polyline, with the
multiplicity a cubic needs there (without them the cylinder-minus-box
intersection curves took 400 spans at 1e-3 mm). Spans and rounds are
capped; a fit that does not converge, or a curve farther from the surface
than the tolerance allows, fails naming the wire, edge and distance.
Samples are projected with the kernel's own surface parameter search (the
evaluator's form, parameterised as truck parameterises it; truck's search
for surfaces it does not take), each from the one before, unwrapped along
periodic directions. Where the surface collapses (a pole) the undefined
parameter comes from the nearest sample that defines it, so a pcurve runs
straight into the pole. Each wire's uses are then moved by whole periods so
each starts where the one before it ends, starting after a pole junction
if there is one; at further pole junctions the offset that gives the loop
a boundary's turn is taken. The wire then moves into the base period. A
seam edge used twice is fitted once.
Truck places intersection curves on their surfaces to about 1e-6 mm, so
their pcurves need a tolerance of 1e-5 mm or more; the test records that.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A plane's parameters are affine in space, so the pcurve of a line or a
(rational) B-spline on a plane is the curve's own knots and weights with
its control points mapped into (u, v): no fit. Curved surfaces and
intersection curves still return "not supported yet". A test helper checks
every pcurve the way the schema validator does, its interval against the
curve's range, the lift error and the wire's closure in (u, v); the box's
line pcurves lift onto their lines' ends bit for bit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The shared base for the schema 4 types and the truck pcurves: 2D B-spline
pcurves as SMLib keeps its UV trim curves, and the truck adapter's pcurve
function, not supported yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
* Ribbon: New Sketch on the Sheet Metal tab again; tests pin sketch enter/exit
Before the ribbon redesign the Sheet Metal tab started with New Sketch. It
gets its own Sketch panel again, like the Model tab's, and a sketch started
there keeps the tab until it opens.
Workspace tests pin the ways into and out of a sketch: New Sketch then a
plane, a double-click in the timeline or the browser, and the browser's Edit
Sketch all open the SKETCH tab looking at the sketch. Finish Sketch from the
palette or the ribbon lands on MODEL with the earlier view back. Escape and
Enter keep the sketch open (#1804), and Undo after Finish reopens it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs: ribbon layout lists Sheet Metal's Sketch panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Workspace test: press the keys well apart
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* CI: build and run the tests in one job, sccache on a Blacksmith sticky disk
The tests job compiles and runs the tests itself on the 32 vCPU runner, so a pull request or merge
group uploads and downloads no 3 GB archive (the download took 45 to 385 s per shard, and once
14 minutes) and starts beside the repository checks instead of after them. The archive is built
only for the nightly and manual jobs that need it. sccache moves from GitHub's 10 GB cache, where
about 1100 per-ref entries were evicting each other, to a sticky disk that only main commits.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
* TEMP: let this branch save caches so the warm run can be measured
* Revert "TEMP: let this branch save caches so the warm run can be measured"
This reverts commit 1bbbfcf5ff.
---------
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* claim: #3378
* Pattern and mirror copies of a join that merged its targets join the merged body (#3378)
Since #2755 a pattern or mirror copy of a primitive reuses the seed's targets. When the
seed was a join that merged two targets, the second one no longer exists, so the copies
failed with `unknown body`. The copies now keep the seed targets that are still there;
when none is, the names stay and the error says which body is missing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Rerun CI: the previous run was cancelled before any check ran
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #3242
* MCP batch: rollback restores the design the batch began in (#3242)
A failed batch with rollback: true reported rolled_back: true even when the
design never went back. Two defects, both fixed:
- file.new (and any document-replacing command) replaced the session with it,
taking its checkpoints down too, so a batch that failed after file.new had
nothing to restore. Session::replace_design now carries the checkpoints over
to the fresh session: the checkpoint belongs to the batch, not to the
session that happened to be holding the document.
- rolled_back echoed the requested flag. It now reports whether the design is
actually back: the app answers engine.batch with a new restored field, and
the MCP tool reports restored && rollback. An app that predates the field
falls back to undone/completed.
* Batch: answer with main's rolled_back field after the merge (#3242)
Upstream #2900 put the real restore check on engine.batch as
`rolled_back` while this branch named it `restored`; the merge kept
main's control test and this branch's field, and CI caught it.
Standardize on main's wire name: the app answers `rolled_back`, the
MCP tool reads it (fallback for older apps unchanged), docs and tests
follow. The checkpoint carry-over across file.new is unchanged.
---------
Co-authored-by: DahakBuilds <dahakbuilds@users.noreply.github.com>
* claim: #3141
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* 2-tangent circle: start from the touching circle, not the click (#3141)
The tool passed its second click as the place to start solving, and a
click lies on a curve: the circle began centred on that curve, where its
tangency has no side yet. The solve then moved the user's curves to fit
(a rectangle came out skewed), or failed outright when they were fixed.
The command now starts from the centre, nearest the requested point, of
a circle of that radius touching both curves (where the curves moved by
the radius meet), and the tool asks for the circle between its two clicks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* CI: give the slowest tests a shard each
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* CI: spread the slow tests on pull requests too, and keep both junit files
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* CI: pull requests skip the goldens and the action matrix sample unless the kernel changes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Docs: which pull requests run the goldens
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* CI: run the tests in one job on a big runner, slowest first
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* CI: allow the 32 vCPU runner label in actionlint, read junit.xml in the golden hint
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The origin point was drawn in the view and listed in the browser, but
nothing could pick it: the viewport never offered it as a hit and the
browser's O row had no selection. A new ORIGIN accept bit lets an input
take it as a vertex of no body; Move's point inputs (From, To, Point,
Pivot) accept it, from the view or the browser's Origin folder.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A curve written over a knot span of 1e-7 or less was refused on the STEP side
("knot vector consists single value"), and both faces it bounds were lost. Its
knots are moved and scaled onto (0, 1), which leaves its shape as it is.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
cone_point and cone_point_vertex join the sweep now that #3871 reads a cone
bounded only by its base circle up to its point.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Fix recovery for unsaved background tabs (#2884)
* Recovery: scheduled autosaves of every tab write on the worker thread (#2884)
After #3139, Autosaver::save waits for the write and Autosaver::tick runs
it on a worker. The scheduled path now goes through tick (which skips a
tab whose previous write is still running), so several dirty tabs no
longer write on the window thread each interval (#3044). save stays for
Save, close and shutdown. Each tab's autosaver has no clock of its own:
the shared cadence decides when it is due. The tests wait for the
writes a tick started.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Recovery: closing a tab waits for its running autosave before removing the entry (#2884)
With the autosave on a worker thread, discard could remove the files and
the lock while a write was still running; the write then recreated the
entry without a lock, and the next launch offered a design closed with
Don't Save. discard now ends the running write first, as Drop does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A BREP_WITH_VOIDS void's faces face into the cavity, so it encloses volume of
the opposite sign to the outer shell. Its ORIENTED_CLOSED_SHELL flag alone
doesn't tell: SOLIDWORKS writes .F. over a shell already facing into the
cavity, SolidCraft .T., and flipping by the flag added the cavity's volume. A
void enclosing volume of the outer shell's sign is now turned inside out.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* STEP import: regression sweep over the step-conformance parts of known shape
Fifteen generated parts from step-conformance (CC0): plain, NURBS, seamed,
placed, inch and far-away cylinders, domes, a grooved shaft, a plate with 100
holes, a 3.6 m disc, a 0.5 mm pin, a B-spline half-torus and a washer. Each
reads back with its exact volume, its box and its face count. The half-torus
on a TOROIDAL_SURFACE and the cones to a point are left out until they read
right.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* STEP conformance sweep: the pin's volume as FRAC_PI_8 (clippy)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A cone face written with no vertex at its point has the base circle as its only
loop (or that circle and a VERTEX_LOOP at the point, which has no edges). Its
surface was spanned over the circle alone, a sliver, so a solid cone came in as
a flat disc. Such a face is now spanned to the apex, and its circle is joined to
the apex by a seam line out and back, as a cone with a seam is written.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Boolean: a small hole in a large body's face stays a hole when the face is divided
A boolean divides faces at a tolerance that follows the larger operand (a few
millimetres on a part a few metres long). A small hole's rim, two half circles
when read from STEP, then became two chords: its polygon in the face's
parameters had no area, and the sign that tells a hole from a face piece came
out at random. The hole became a face of its own, inside out, and the face it
was cut in stayed whole. A box cut round such a hole gave a piece without it,
and the shared part of that piece and a pin in the hole came out negative
("malformed") or the boolean failed.
divide_face now divides a loop whose polygon is flat again, finer (down to a
hundredth of the loop's own size), before reading its sign. Only the polygon
that places the loop changes; loops that have an area are untouched.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Kernel: the long-plate hole test in tests.rs; a flat loop's size from its box
Review: the test moves into the kernel's test file (no file of its own),
and finer_if_flat takes the loop's size from the diagonal of the box round
its sample points instead of the largest distance between any two.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Aac7mP9rthK3WsPpPRvCsF
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* claim: #3358
* IGES import: trim curves with reversed pieces no longer drop their faces
A curve on a surface (142) gives each trim loop twice: B in the surface's
parameter space and C in model space, running the same way. Some exporters
write C as a composite curve (102) whose pieces are in order but some of
them end to start. The reader passed those pieces on as they were, so the
loop did not close and the STEP reader left the face out ("face boundary is
not a closed loop").
The pieces of C are now turned to run head to tail, starting at S(B(start)),
the point where the parameter-space curve begins on the surface. That keeps
the loop's sense from the file, so holes still run against the outer loop.
A piece whose ends are one vertex at the sewing tolerance keeps its
direction.
On a 9,276-face trimmed-surface car model, 7,887 faces now import instead
of 3,882.