[devin-ai-integration]

What and why

SonarCloud reports three open Go code smells on develop that should be cleared before 0.12.0 is cut:

• go:S3776 on eachUnwrappedDeclMember in internal/exec/runtime/invoke_calc.go: cognitive complexity 33, the allowed maximum is 30.
• godre:S8193 twice in internal/frontend/grpc/validate_test.go: a got := variable declared in an if and used only in the condition.

eachUnwrappedDeclMember and unwrappedDeclMembers both opened with the same three-way dispatch — ast.OwnedConstraintOf, then a type switch over *ast.AssumeMember and *ast.RequireMember — and the iterator form repeated the yield loop in each arm. ast.OwnedConstraintOf only answers true for those same two node types and hands back their Body unchanged, so the first check never selected a member list the type switch would not have selected. Both readers now call one helper, constraintMemberBody(decl) ([]ast.Node, bool), which returns the nested members of an assume or require member in either of its forms; the iterator keeps a single yield loop over that body and the existing ast.DeclMembers walk. The sequence of members each reader produces is unchanged for every declaration kind, so no caller, error or error timing moves. The iterator's cognitive complexity drops from 33 to 17.

The two test variables are inlined into their conditions; the failure messages already print the slices, not the joined string.

No changelog fragment: a pure refactor with no user-visible change.

How it was verified

• go build ./..., go vet ./..., gofmt -l . (empty)
• go test ./internal/exec/runtime/... ./internal/frontend/grpc/...
• make test-short

eachUnwrappedDeclMember was just perf-tuned (#1742 removed allocations from the binding walks), so the three benchmarks that walk it were measured on develop (8dc84706) and on this branch, six rounds each, with go test -run '^$' -bench <name> -benchmem -count=6; InstantiateWarmModel was measured in alternating develop/branch rounds because its single-tree runs vary by ±50% on this box. benchstat finds B/op, allocs/op and sec/op unchanged within noise (the new helper allocates nothing and is inlined):

pkg: github.com/Open-MBEE/OpenSysML/internal/exec/runtime
                     │ develop.txt  │          branch.txt           │
                     │    sec/op    │    sec/op     vs base         │
AssignmentLoopStep-8   2.943m ± 20%   3.093m ± 13%  ~ (p=0.485 n=6)

                     │ develop.txt  │          branch.txt           │
                     │     B/op     │     B/op      vs base         │
AssignmentLoopStep-8   926.2Ki ± 0%   926.3Ki ± 0%  ~ (p=0.394 n=6)

                     │ develop.txt │          branch.txt          │
                     │  allocs/op  │  allocs/op   vs base         │
AssignmentLoopStep-8   8.543k ± 0%   8.545k ± 0%  ~ (p=0.537 n=6)

pkg: github.com/Open-MBEE/OpenSysML/internal/frontend/grpc
                       │ develop.txt  │          branch.txt           │
                       │    sec/op    │    sec/op     vs base         │
InstantiateWarmModel-8   363.3µ ± 39%   395.3µ ± 21%  ~ (p=0.818 n=6)

                       │ develop.txt  │          branch.txt           │
                       │     B/op     │     B/op      vs base         │
InstantiateWarmModel-8   126.1Ki ± 3%   125.8Ki ± 6%  ~ (p=0.699 n=6)

                       │ develop.txt │          branch.txt          │
                       │  allocs/op  │  allocs/op   vs base         │
InstantiateWarmModel-8   2.046k ± 1%   2.045k ± 2%  ~ (p=0.818 n=6)

pkg: github.com/Open-MBEE/OpenSysML/tests/perf
                       │ develop.txt  │          branch.txt           │
                       │    sec/op    │    sec/op     vs base         │
GRPCVerifyConstraint-8   3.056m ± 42%   2.719m ± 20%  ~ (p=0.132 n=6)

                       │ develop.txt  │          branch.txt           │
                       │     B/op     │     B/op      vs base         │
GRPCVerifyConstraint-8   766.3Ki ± 0%   766.3Ki ± 0%  ~ (p=0.184 n=6)

                       │ develop.txt │           branch.txt           │
                       │  allocs/op  │  allocs/op   vs base           │
GRPCVerifyConstraint-8   14.89k ± 0%   14.89k ± 0%  ~ (p=1.000 n=6) ¹
¹ all samples are equal

Checklist

• make test and make lint pass locally (make test-short and the race-free package suites; see above)
• Tests added or updated for the change — none needed: a behaviour-preserving refactor covered by the existing runtime and gRPC suites
• Documentation extended where it already covers the surface — no documented surface changes
• Changelog entry added — not needed for a pure refactor with no user-visible change
• baselines regenerated — no gate count moved
• 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: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.