Split Face records its lines on the body (split_lines): heal and the
certified-planar regularizer keep the faces apart along them, and they
make the body regularization-protected. On SMLib the boolean, the n-way
union and Shell carried the protection over but dropped the lines, so a
body with an authored split lost its record after a Union with partial
contact (issue #674/#782's case), and a shelled one after Shell. Each
result now keeps its operands' lines, each line once
(Body::inherit_split_lines), as the old kernel's coplanar join did.
regularization_protection_tests: SMLib's shell makes no section strips,
so the painted cup no longer has more faces than the ordinary one. The
ordinary cup is the box's five kept faces, their five linings and the
rim: 11. A paint over a whole face authors no boundary inside it, so the
painted cup has the same 11 faces; the split side's line stays an edge on
the wall and its lining: 13. The test now asserts those counts exactly
(it asserted more than 11 for both, the old strip route's witness) and
that the split lines survive.
Fixes kernel issue_674_tests::partial_contact_retains_an_authored_split_away_from_the_contact
and body::regularization_protection_tests::stamped_shell_keeps_authored_painted_and_split_boundaries.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The truck booleans answered a tool clear of the body without intersecting (the operand kept,
its record shared). The SMLib boolean always rebuilt the body, so a cut that could not reach a
body still handed back a new record: a hole far from a second body reported that body changed
(#2904's test), and a body clear of a pocket's cutter was no longer the same solid. A cut or
intersection whose operands are apart (their exact boxes, grown by twice the tolerance as
union_apart takes them, do not meet) now gives the body itself with its faces and edges kept,
or nothing, with no SMLib call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
transform_matrix moves a body by a matrix that is not a rigid motion through SMLib, which
numbers the moved body's faces afresh, then attached the input's face sources and paint in the
input's face order. Planar certificates landed on the wrong faces (a reflected, sheared plate's
hole walls and caps no longer lay on their certified planes). Each moved face now takes the
source and paint of the face its history names; a face with no known origin takes no source,
and with no history at all the paint stays as it was.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Booleans, blends, sweeps, lofts, shell/draft/offset, healing and sewing,
SMLib conversions, meshing and measuring; the import as parse, index,
per-solid read, placement and mesh check, convert batches and finish.
The existing per-step boolean/tessellate capture names its spans too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The doc puts exact tools into SMLib's boolean as drawn, so nothing asks whether a touching tool
can be joined, batches extrusions, or probes for room to grow a profile.
(not built: base in progress)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A boolean is SMLib's boolean, its history SMLib's tags: the jittered retries, tool nudges, prism
and coaxial routes, the retry result guard and their diagnostics are gone. A rigid placement moves
the record; any other matrix goes through SMLib's transform (new Brep::transformed over
smlib_brep_transform). Bodies apart join as their records side by side (brep part::compound),
with each piece's faces, sources and history; apart means their exact boxes do not meet. A split
is the intersection with a half-space box, with no faces pushed off the plane.
(not built: base in progress)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Body::mesh / mesh_plan / mesh_of and MeshTolerance (chord, display, display_of, export_of)
replace tessellate, tessellate_with, display_mesh*, display_plan*, reference_display_mesh_angular;
every caller in kernel, doc, engine, io, ui-egui, gpu and the CLI renamed mechanically; the
truck-mesher fallback, mesh_shells*, refined* removed; Body::new orients from the planner;
truck-meshalgo dropped from Cargo.toml with the tests of its hooks. Still to fix: trait imports
in tess/*, splitface/lumps/preview/step_in/boolean_record meshalgo uses, measure.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
All operations move to SMLib at once, so the selector is no longer the booleans' own:
`SOLIDCRAFT_KERNEL=smlib|truck` (read once, unset is truck, the same refusals for an unknown
value or SMLib in a build without it) is answered by `solidcraft_kernel::backend()`
(crates/kernel/src/backend.rs), which every operation's dispatch calls. The boolean's SMLib path
moves to smlib_boolean.rs. The regression rig's kernel check (`boolean_kernel`) now answers from
the switch, and the rig and CLI set SOLIDCRAFT_KERNEL.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
With the `smlib` feature, SOLIDCRAFT_BOOLEANS=smlib (read once per process, default truck)
sends every kernel::boolean through SMLib: each operand's schema 4 record imported and tagged
as operand 0 or 1, SmBrep's boolean, SMLib's check at every level, the result exported and
decoded (with its geometry check). A result either check refuses is an error. The setting
naming SMLib in a build without it, or anything unknown, fails every boolean instead of
quietly running truck, so a comparison never compares truck with itself.
The history is SMLib's tags renumbered into Body::history's index spaces. The export now
reports where each SMLib face and edge lands in canonical order (its arrays follow
SmBrep::GetFaces/GetEdges, the order the tags use). The result's order is checked against the
decoded solid by face and edge identity, the operands' by re-exporting the imported body and
comparing its numbering with the record's; where either differs the result has no history
rather than a wrong one. The result gets the operands' regularization protection and planar
roles as truck's does.
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
transform_matrix took any finite 4x4 matrix. A perspective entry in the
last row, or a singular linear part, was built into faces that mesh to
nonsense: a revolved ring placed with such a matrix measured hundreds of
millions of mm3, and a boolean on it ran for over half an hour before
failing with "no result". Matrices come from design files and command
parameters, so they are refused with KernelError::Invalid instead.
* claim: #3244
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Fail an unverifiable intersection instead of keeping what it should trim (#3244)
When every route of the general boolean fails, it falls back on the sampled answer
(one body whole, or nothing) with a "not verified ... that part is ignored" warning,
even when the bodies' facets cross well past their faceting error. For an
intersection that keeps exactly the material it was asked to trim: teeth standing
0.27 mm out of a trimming cylinder came back untrimmed, and `solidcraft-cli run`
exited 0 with a part 16.918 mm in radius.
An intersection whose sampled answer is contradicted by facets crossing by more than
twice the depth the warning reports (6 x the check mesh tolerance, three times the
faceting error of two meshes) now fails, and the model is kept. Cuts and unions keep
the fallback: shelling a body in two pieces relies on it (a cavity tool grazing the
other piece), with the same depth and ignored volume as this case.
Fixes#3244
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Leave an empty unverified intersection as it was (#3244)
Only a doubted intersection that keeps one body whole is refused. An empty
one ("the bodies are apart") stays a result reported unverified: the clearance
and interference checks rely on it, and main's new
a_pin_through_a_d_bore_of_its_size_still_overlaps_the_flat (#2714) expects it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Moving an occurrence places every body under it again, and each placed copy
searched every revolved face for where it sits to turn it away from its seam,
although its source had been turned already. On the Prusa Core One assembly
that search was most of the 12.7 s a move of the top occurrence took. A body
now records that its revolved faces are turned, and world placement
(`place_matrix`) skips the search for such a body. Modelling moves (pattern
copies, hole tools) keep `transform_matrix`, which still turns them in the
copy's own frame: where a face wraps all the way round the turn it finds
depends on those coordinates, and later booleans depend on it.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The retries in boolean_whole_inner mesh every candidate to compare its
volume. A nudged union at the coarse tolerance could come out with a face
on a cone whose trim loop ran metres along the cone's continuation (edges
up to 12.7 m outside a 155 mm part); meshing it took seconds to minutes per
face and once never finished, though its volume (+-1e10 mm3) failed the
check anyway.
check() first samples the candidate's edges (16 per edge) and discards a
candidate with a point further outside the operands' mesh box than 5% of
their size (or a non-finite point): no edge of a union, cut or
intersection can lie there. The reason goes into the new
BooleanDiagnostics::rejected (omitted from JSON when empty). Results of
sound candidates are unchanged.
On the reporting design the PitchTail join drops from 18.7 s to 4.3 s with
the same result; the regression tests synthesize a cone face whose loop
runs off its surface.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A cut whose tool ends exactly on a face of the target (a hole cut from a pocket's floor up to the
plate's top face) could come out of the general boolean with the target's face whole over the hole
and the tool's end face under it: a skin of no thickness that keeps the hole closed. No volume
shows it and the sampled membership checks see the hole either way, so the result was kept. The
result check now refuses a body with a planar face lying back to back on another of its own planar
faces with nothing behind part of it (`coplanar::has_skin`), and the retries take the cut.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Iterate instead of indexing in nine loops
Replace index loops that look up neighbours by position with enumerate,
zip(cycle().skip(1)) for wrap-around pairs and zip(skip(1)) for
consecutive pairs. No behaviour change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: ring_edges for the edges of a closed ring
Every walk over a polygon's edges, the last back to the first, spelled
x.iter().zip(x.iter().cycle().skip(1)); eighteen of them now call
geom::ring_edges. Three redundant take(len) calls go with them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Ring, a closed ring that knows it is closed
A Ring holds items in order around a closed polygon, loop or fan; edges()
walks each with the next and the last with the first, corners() each
with both neighbours. Ring<Vec2> has signed_area and even-odd contains.
Loop2::polyline returns a Ring, as do doc's face_loops and
boundary_loops, kernel's coplanar wire samples, CapLoop's segments,
tess pieces' rings, the gizmo's plane quads and the navigation cube's
facets. Their walks take edges() or corners() instead of wrapping an
index; sketch's text-hole test and doc's poly_area become contains and
signed_area. Loop2::joints pairs each segment with the next.
Replaces the ring_edges free function.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Chain, Ring and Path; bounds for a chain of points
Ring becomes Chain<T, Closed> beside Path = Chain<T, Open>: edges()
closes a ring and stops at a path's end, Path::close makes the ring,
and bounds() (the empty box for no points) serves both. Eight
hand-rolled min/max folds over a loop's polyline call bounds(); doc's
flat_polyline returns a Ring.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: area and inside on Ring for any planar point
Ring's signed_area and contains take any Planar point (Vec2 or
[f64; 2]). The cam profile in joints_cam is a Ring: its own even-odd
test goes, and its edge walks, gap and area take edges(). face_params'
loop area, the text path's winding, tess pieces' ring area and the
copies in three tests call signed_area or contains; the formulas are
the same term for term.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* doc: sheet's closed outlines go through Ring
dedup takes and returns a Ring (dropping a last point back on the first
is what makes it a ring's), near_outline walks a Ring's edges, and the
copy of it beside fold checks calls it. point_in goes: its callers ask
the Ring. Its formula anchored each crossing on the other end of the
edge, so the two can differ only for a point on an edge to the last
digit; one caller sets such points aside first, the others test a
region's interior point, a cut's midpoint or another outline's corners.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* doc: cam tests build their outlines with Ring::from
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: area_vector on Ring<Vec3>; three Newell sums use it
Ring<Vec3>::area_vector is half the sum of a × b round the ring. heal's
loop_area, lip's outer-loop turning and the revolve's loop orientation
summed it by hand with a wrapped index; they take it from the Ring
(the half is exact, and two of them only read its sign).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Aabb3 from points; boxes built and asked through it
Aabb3 collects from points (skipping non-finite ones, as add does) and
gains inflated, contains and meets. Fifteen hand-rolled min/max folds
over points or boxes in kernel, doc and engine become collect or union,
and their "within tol of the box" tests become inflated(tol).contains.
blend's refused-edge boxes and naming's face index hold an Aabb3.
A NaN point is now outside every box, where the old negated
comparisons let it in; an infinite point no longer stretches a box.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Aabb2, the plane's box; 2D folds and box tuples use it
Aabb2 mirrors Aabb3 (collect, add, union, inflated, contains, meets)
and adds relative, a point's place in the box. sketch's link and
link_loops each had that formula; link_loops' boundaries, prism's
segment and loop boxes, profiles' shape boxes and grid, and ends'
segment box hold an Aabb2 instead of a (low, high) pair. Folds in the
cam, sketch bounds, sheet's perforation grid and the sketch clipboard
collect one.
Boxes are built from the same terms (a round's box keeps c - r - tol).
As with Aabb3, a NaN point is outside every box and an infinite point
no longer stretches one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* kernel: prism's segment boxes are plain Aabb2s
The SegBox alias named a (low, high) pair; Aabb2 says it itself.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Chain::bounds is an Aabb2; the boxes #2382 left as pairs follow
Chain::bounds returns an Aabb2. Its callers keep it: sheet's hole and
cut boxes (loop_box), letter boxes, island outlines and line-near-cut
test use meets, contains and inflated; profile's outline overlap and
max_clearance, profiles' box tests and the cam test read its corners.
flat_size and the flat pattern's JSON size are the union of the loops'
boxes (0 × 0 for no loop, where the pair fold gave -∞).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* geom: Aabb3::intersection; hand-grown and hand-met boxes use the box
Aabb3::intersection is the box both hold. ops' four overlap boxes of a
boolean's operands take it, and seven boxes grown by hand (ops' body
boxes, the distance set's bounds, coplanar's face box, eval_more's
padded mesh box, compare's group and region boxes) call inflated with
the same amount on every side.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Two empty numerical cut attempts stopped the boolean retry sequence before
the exact geometric strategies ran. When the result guard refuses that empty
candidate, the exact strategies (sliced planar, pushed walls, local prism,
sphere lens, seam halves) now try the cut first; if none recovers it, the
original empty candidate and diagnostics are kept, before any displaced
numerical retry.
Rebased onto main as one commit (the branch's own changes, its merges of main
dropped). The only conflict was the new engine regression test next to
main's #1985 STEP export tests in boolean_regression_tests.rs; both are kept.
* inspect.clearance: pairs whose bodies did not change are not measured again
Each clearance check measured every close pair again, though between two
checks usually only a few bodies change or move. Results are now kept per pair,
keyed by both bodies' geometry and placement and the check's settings; a pair
whose two bodies are unchanged is taken from that cache. With 40 overlapping
parts, a check after moving one of them took 0.8 s instead of 23 s (741 of 780
pairs reused). Pairs that could not be worked out are tried again each time.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: boxes and placed keys exactly as before the reuse
The pair boxes are taken over the mesh's triangles with finite corners, as a
distance shape's are, so pruning is unchanged. A placed body's identity now
includes its model body's own frame, so two placements are never mistaken for
one another.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: a large body is meshed once for all its pairs
Each pair that touches cuts the larger body down to the pair's region with a
kernel boolean, and that boolean meshed the whole body twice: once for its
checks and once inside the intersection. A sheet with 600 faces paid about
2.5 s per pair for meshes that were the same every time.
A body can now keep its meshes for reuse (Body::keeping_meshes); its booleans
take the meshes from there, and the intersection takes its shell meshes
from there too (truck-shapeops: and_with_history_meshed). Clearance uses such
copies for one check. Results are unchanged to the bit. On a 3 x 2 m sheet
with bends and 300 holes, the clip took 0.23 s per pair instead of 2.6 s, and
24 rivets took 5.6 s instead of 16.5 s.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: pins on one cylinder with their hole are answered or reported in seconds
A pin seated line to line in its hole, on its axis, and with a blend running into its wall
(a fillet under a head) sent the boolean through minutes of retries that ended unchecked; a
pin a hair off its hole's axis was passed as clear. Clearance now asks first whether the pair
holds such a pin (from the faces' surfaces alone). On its axis the boolean gets a few tries,
which answer plain pins at once; a hair off, or when those tries fail, the shared volume is
worked out apart by coaxial_overlap, which now also takes the zero-offset, equal-radius case,
moves the hole's wall when a blend keeps the pin's from moving, and checks its band more
tightly. What neither answers is reported as unchecked, naming the pin.
The kernel gains with_attempt_budget: the booleans run inside it get a number of tries of
the kernel's own intersection between them, the same on every machine.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: the pin tests sit beside the other measured pairs, not at the file's end
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: the bodies left whole meet their booleans with their meshes kept
A body that meets several others at one size (a bracket with balls pressed into it) was
meshed again for each pair's intersect: only the clipping used the copies that keep their
meshes. The intersect, the pin work and the union check now take those copies too.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Near-coaxial overlap: ten tries per intersection, enough for a flush head pulled back
#1984's headed pin made before its plate needs more than eight.
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>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* Boolean: a pin in a bore of its own size is clear of it, whatever its facets
A pin whose wall is in other faces than the bore's (turned from a profile: three
faces; the bore two) is faceted at other angles. Where one body's faces end on
the shared cylinder (the pin's wall through the end face of the tube, the pin's
end inside the bore) the facets cross by the facetings' error, so the sampled
shortcut doubted its own "apart" answer, every retry failed on the coincident
walls, and the result came back unverified after seconds: inspect.clearance
left such pairs unchecked ("could not be intersected there").
The shortcut now leaves out crossings that lie on a cylinder both bodies have
faces on (axes and radii within the kernel's same-surface tolerance, where both
faces run). A pin a hair off its hole's axis, or one pushed into a narrower
bore, is still not taken for apart.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Boolean: skip only the shared cylinder's own crossings, not a flat along it
The shortcut left out every crossing near a cylinder both bodies have faces
on, from any face: a pin through a D-bore whose flat is 0.022 mm inside it
came back as contact. Mesh::crosses_except now hands the predicate the two
triangles' faces, and only crossings between the faces on the shared
cylinder, or between one of them and an end face across its axis that keeps
to its body's side of the cylinder (within the facetings' error), are left
out. The D-bore's end face reaches into the cylinder to the flat, so the
pin's wall through it counts again; a flat, key or rib along the axis always
counts.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* Boolean: a pin in a bore of its size is clear at every placement
The sampled shortcut's check that no facet has material under it on the
wrong side of the other body still caught the facetings of a shared
cylinder. A facet of a pin's end inside the bore (or of the bore part's
end face round the pin) with all three corners on the rim has its middle
just inside the cylinder. Where the other body's chords reach further in
at that angle, the point under it lies inside the other mesh, though the
bodies only touch. Which facets do so depends on the face splits and the
placement, so the same pair read as clear at one placement and as "not
verified" at another.
Such points are no evidence now when they lie on the shared cylinder,
under a facet of one of its end faces: the end faces and band the
crossing check already leaves out. A pin 0.02 mm larger is no shared
cylinder, and its overlap is still found.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* inspect.clearance: pairs whose bodies did not change are not measured again
Each clearance check measured every close pair again, though between two
checks usually only a few bodies change or move. Results are now kept per pair,
keyed by both bodies' geometry and placement and the check's settings; a pair
whose two bodies are unchanged is taken from that cache. With 40 overlapping
parts, a check after moving one of them took 0.8 s instead of 23 s (741 of 780
pairs reused). Pairs that could not be worked out are tried again each time.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: boxes and placed keys exactly as before the reuse
The pair boxes are taken over the mesh's triangles with finite corners, as a
distance shape's are, so pruning is unchanged. A placed body's identity now
includes its model body's own frame, so two placements are never mistaken for
one another.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* inspect.clearance: a large body is meshed once for all its pairs
Each pair that touches cuts the larger body down to the pair's region with a
kernel boolean, and that boolean meshed the whole body twice: once for its
checks and once inside the intersection. A sheet with 600 faces paid about
2.5 s per pair for meshes that were the same every time.
A body can now keep its meshes for reuse (Body::keeping_meshes); its booleans
take the meshes from there, and the intersection takes its shell meshes
from there too (truck-shapeops: and_with_history_meshed). Clearance uses such
copies for one check. Results are unchanged to the bit. On a 3 x 2 m sheet
with bends and 300 holes, the clip took 0.23 s per pair instead of 2.6 s, and
24 rivets took 5.6 s instead of 16.5 s.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* Boolean nudge tried both ways: no sliver where a cut's end touches a round side (#897, #827)
Robustness seed 14 cuts a cylinder whose end lies exactly on the apex of the
plate's arc side. Since the display tessellation plans (#805) the generator sees
the arc's true extent, so the cut's base lands on -10 instead of -10.000000000000002
and the end is tangent to the arc. The direct boolean fails there and the
last-resort tool nudge pushed the end into the arc, leaving a planar sliver a
tenth of a millimetre wide; shell then picked that sliver and refused it. The
13 October merges were not the cause (the bisect lands on #805).
The nudge is now tried both ways and the simpler result (fewer faces) is kept,
so the opposite nudge parts the touching faces instead. The same change joins
the hinge barrel in #827, whose top lies tangent to the lid's underside where
the mounting leaf ends. Each result still passes the same checks as before.
Goldens: seeds 14, 26 and 85 now build to the end, 31 and 78 come out with
fewer faces. Seeds 6, 43 and 73 still stop at shell: their joins now succeed and
shell reaches corners it can't offset (a cusp where a box face touches an arc,
coplanar faces facing opposite ways), which is unchanged.
Fixes#897Fixes#827
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Nudge the other way only when the first nudge fails or leaves a sliver
Trying both nudges and keeping the one with fewer faces could prefer a result
that keeps two pieces joined by a hair over one that separates them (the
centred coplanar revolve cut). The opposite nudge now runs only when the first
one failed or left a face narrower than 1/200 of the part, and it is taken
when it works with fewer of those.
Goldens re-recorded on current main: seeds 14, 26 and 85 build to the end,
31 and 78 with fewer faces. The other changed seeds are main's own since its
last full record, identical with and without this change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Nudge the other way only for slivers the nudge made
On the sensor enclosure's lid the first nudge left thin faces the operands
already had (thin walls, small rounds), so the opposite nudge ran a second
boolean for nothing and the press/pull took 79 s against main's 55 s. The
opposite nudge now runs only when the first one failed or left more slivers
than the operands brought; the press/pull test takes 36 s against 42 s on main
(same machine, one after the other).
Goldens re-recorded on top of #940.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Goldens on current main: seeds 31, 43 and 78 with the two-way nudge
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Keep the hinge and tangent-cut tests beside the #871 fillet test
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Goldens: seeds 33 and 91 as main's fillet changes build them
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Goldens: keep main's #950 record
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A through-all hole's tool is twice the largest body long, and the boolean's tolerances
followed that length: a 6 mm hole through a 920 mm part was traced at nearly 1 mm, wider than
the hole, and the cut came out empty or failed. When every other strategy has failed, the
boolean is now asked again (unshifted, then once shifted) at tolerances scaled to where the
bodies meet: the overlap of their boxes and the thinner operand's thickness. Its result is
checked as the others are and held to the retry guard. It runs last, so whatever an earlier
strategy answered is kept as it was, and only for single-shell operands (bodies in several
pieces are retried a piece at a time, each with this pass of its own).
The sampled shortcut also gave up on a slender tool lying across its box (a long hole drilled at
an angle fills little of it): it now samples past the first 2000 points until enough land
inside, so a tool that misses a body leaves that body as it is. Bodies that had enough samples
in the first 2000 are sampled exactly as before.
Baseline: the 90 actions that newly pass, rechecked one by one with the final build.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Walls swept along a spline have flat B-spline tops and bottoms, so booleans between them went to the general intersection, which fails on coincident spline faces. Such faces are now put on true planes, and two prisms of the same height and direction are decided in 2D with their spline outlines fitted by arcs within a micron. The result is kept only when its volume is plausible; otherwise the general boolean takes over. The free-form flag is cleared by a drop guard so an unwinding boolean can't leave it set.
Fixes#844.