feat(runtime): evaluate the ShapeItems geometry the library determines

[devin-ai-integration]

What and why

ShapeItems defines a shape's faces, edges and vertices through bind connectors with multiplicities. A Box's faces and edges already evaluated; every read through a Cylinder's or Cone's edges failed:

cyl.base.edges → feature value cyl.be: multiplicity violation: 1 value(s) bound to a feature with multiplicity lower bound 2

The cause was in the general binding mechanism, not in the library: binding [1] bind [0..*] base.edges = [0..*] be was read as a whole binding because its [0..*] ends neither bound nor require a count, so be [2] was bound to the disc's one edge. A binding connector's own multiplicity is the number of links it declares, so that binding identifies one value of each end — it relates the disc's edge to some value of be and determines neither end whole.

• lower.Binding now carries the connector's own multiplicity (Binding.Multiplicity, from binding [n]).
• runtime/binding.go partialBinding bounds the links of each end by the smaller of the end's and the connector's finite upper bounds; an end whose declared lower bounds already exceed the links (be [2] against one link) is partial before its value is read (endpointRequiresMoreThan), since it can never be linked whole — which also stops a read of be from recursing into its other binding. binding [2] over two [2] ends and bind a = b without a connector multiplicity stay whole bindings with the existing count check.
• Binding diagnostics quote the connector multiplicity (binding [1] bind [0..1] tf.edges = [0..1] tfe).
• wholeBindingCounts checks the identified value count against the connector's own range too (binding [2] bind a = b over one value is ErrMultiplicityViolation), and bindingLinksNothing treats a connector [0] as linking nothing.
• Binding multiplicities are read through the new semantics.Model.RangeIn(scope, mult), so a bound naming a valued feature (binding [links] with attribute links = 1) evaluates in the binding's scope instead of coming back unknown.

Per ShapeItems derived feature (library lines in Domain Libraries/Geometry/ShapeItems.sysml):

Feature	Verdict	Decided by
Box::faces (6), tf.edges (4) with length/width, edges (24), tf.vertices (8), tf.e1.vertices (2), height, isSolid, voids, shape	evaluates	Cuboid 746, faces [6] (752), edges [24] (762); RectangularCuboid 798, Rectangle/Quadrilateral edges and vertices (fixed counts, no [0..1] group)
Cylinder::faces (3), Cone::faces (2), base.edges, af.edges (one ellipse each), height; the same on a Cylinder nested as a Box's voids, isSolid false	evaluates (new)	ConeOrCylinder 384–439 faces [2..3], base : Disc [1]; Disc 237, edges : Ellipse [1] = shape (254); binding [1] bind [0..*] base.edges = [0..*] be (432) links one value, so base.edges keeps its own
Box::tfe…urre, tflv…brrv, and so Box::vertices	underdetermined, ErrBindingEnd naming the binding	CuboidOrTriangularPrism 562, binding [1] bind [0..1] tf.edges = [0..1] tfe (623) and its siblings pick one unspecified member each; Cuboid::vertices subsets tflv = tfe.vertices — recorded in docs/project/omg-issues.md
Cylinder::be, ae	ErrBindingEnd ("cf.edges": feature edges not found)	cf : Surface [1] (414); binding [1] bind [0..*] cf.edges = [0..*] be (433), … = [0..*] ae (497): Surface's edges is the Kernel StructuredSpaceObject's, a frame feature, not an object feature
Cylinder::edges = faces.edges, vertices = faces.vertices	ErrNoSuchFeature over cf (was ErrMultiplicityViolation through be)	ConeOrCylinder 416, 428 read edges/vertices on cf : Surface
matingOccurrences, spaceBoundary	frame feature, not an object feature — ErrNoSuchFeature (unchanged)	Occurrences.kerml declares both on Occurrence; Objects.kerml StructuredSpaceObject::faces restates spaceBoundary as a derived frame feature; Items::Item::shape :>> spaceBoundary redefines it, so shape answers while spaceBoundary itself is the frame's

Row moved in docs/project/spec-compliance.md: the ShapeItems row ❌ → ⚠️ (what the dimensions determine evaluates; the [0..1] groups, the curved face's edge graph and the frame features are the typed errors named there); the binding-connector row gains the connector-multiplicity semantics; Known Limitations bullet, roadmap Track L (L8) and a changelog fragment updated.

Specification basis

KerML 1.0 §7.4.6.2 Connector Declaration (a connector's own multiplicity is the number of links), §7.4.6.3 Binding Connector Declaration, §7.4.9.2 binding-connector semantics; SysML v2 §8.2.2.6 BindingConnectorAsUsage.

How it was verified

• binding_connector_multiplicity_test.go: TestConnectorMultiplicityBoundsLinks (both ends of binding [1] over a [1] and a [2] feature are UndeterminedBindingError naming the binding; an end with a value of its own keeps it), TestWholeBindingWithoutConnectorMultiplicityStillChecksCounts, TestConnectorMultiplicityWideEnoughIsWhole, TestConnectorMultiplicityDeclaresLinkCount, TestConnectorMultiplicityNamesFeature.
• robustness_partial_binding_test.go: TestRuntimeRobustnessPartialBinding (underdetermined end, lower bound above the links, mutually partial bindings terminate, chain through the undetermined end).
• Conformance: instance_library_geometry_cylinder, instance_library_geometry_box_void (new), instance_library_geometry_box (extended); TestShapeItemsUnsupportedExpressionsAreTypedErrors updated intentionally — cyl.edges moves from ErrMultiplicityViolation to ErrNoSuchFeature, cyl.faces.edges/cyl.vertices/cyl.be rows added, cyl.base.edges/cyl.af.edges asserted to evaluate.
• Pinned pilot evaluator (./scripts/download-pilot-evaluator.sh, artifact 0.62.0) over the same model, verbatim: cyl.faces → ItemUsage faces, cyl.base.edges → ReferenceUsage edges, cyl.af.edges → ReferenceUsage edges, cyl.be → ItemUsage be, cyl.edges → Feature edges, block.faces → ItemUsage faces, block.tf.vertices → ReferenceUsage vertices, block.voids.base.edges → ReferenceUsage edges, block.tfe → ItemUsage tfe, block.vertices → ItemUsage vertices, block.isSolid → LiteralBoolean false. The pilot leaves every item usage unevaluated (the usage node is its answer) and agrees on isSolid; it adjudicates none of the geometry either way.

Gates:

Gate	Result
go build ./...	pass
go vet ./...	pass
gofmt -l .	empty
OPENSYSML_REQUIRE_TRAINING_CORPUS=1 OPENSYSML_REQUIRE_PILOT_CORPORA=1 OPENSYSML_REQUIRE_PILOT_LIBRARY_XMI=1 go test -count=1 ./... (after the three download scripts)	pass, 79 packages ok
go test -race -count=1 ./internal/exec/runtime ./internal/ir/...	pass
go test -C tools ./...	pass
make docs-check	pass
python3 scripts/changelog.py check	pass
training_examples_expected.txt	unchanged

Checklist

• make test and make lint pass locally
• Tests added or updated for the change
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (compliance rows need nothing: the census is counted at docs build)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Devin Review found 1 new potential issue.