[devin-ai-integration]

What and why

docs/project/spec-compliance.md was audited clause by clause against the pinned SysML v2.1 Beta 2 / KerML 1.1 Beta 2 texts, the current tree and the pinned pilot. This change applies the documentation corrections that audit produced, as one commit on develop (the first pass, #1526, re-pointed paths and clause numbers; this one corrects what rows say). No implementation code changes; the rows that describe a gap now say so and the fixes are separate work.

Rows changed, by category (173 edits, plus the structural edits below):

• Stale text corrected (DOC-STALE): 162 edits (several rows took more than one) — wrong clause numbers (KerML 1.0 numbering, SysML clauses that moved), renamed or removed Go symbols and test names, two library quotations that are not in the Beta 2 text, two edition stamps, the pilot pin (2026-08 / 0.62.0), stale counts, and internal work-item references rewritten as what the change did.
• Status corrected (STATUS-WRONG): 3
  • Calculation purity: ✅ → ⚠️ Approximate. SysML v2 §7.19.1 only recommends that a calculation "should be pure" and allows action usages as steps; KerML §7.4.8.1 allows a function side effects. The row also dropped perform from the refused forms (a performed effect-free action runs; nothing tests the refusal the row claimed).
  • Bare import without a visibility indicator: ⚠️ → ✅ Faithful. SysML v2.1 Beta 2 §8.2.2.5.2 makes the indicator mandatory, import-visibility is reported (non-gating error at the syntax tier) and the indicator is honoured for visibility filtering.
  • gRPC Query: ✅ → ⚠️ Approximate. The standard's Query has seven PrimitiveConstraint operators and orderBy; =, <, > are implemented, <=, >=, in, instanceOf are refused as malformed and orderBy is absent.
• Implementation gaps now stated truthfully (IMPL-GAP, text + ⚠️, no code change): 6
  • Default multiplicity: a KerML feature with no multiplicity is held to 1..1 (AssumedRange) where KerML §7.4.12 / §7.4.5.2 default to 0..*; the SysML §7.6.3 implicit [1..1] conditions are stated.
  • Nested subrequirements: not materialized into subrequirements/constraints, implicit subject not bound to the container's, verdict does not reach the container's result (SysML §7.21.2).
  • x + null / x - null: refused by the operator evaluator while the function form '+'(x, null) answers (KerML §8.4.4.9.6 makes the operator the invocation; y: [0..1]).
  • excludingAt with an end past the end or an inverted range: ErrIndexOutOfRange where the vendored library body answers the sequence unchanged.
  • Query PrimitiveConstraint operators: <=, >=, in, instanceOf unimplemented (API & Services §7.1.4).
  • Query.orderBy: unimplemented; the known-limitations list now says so.
• Review finding fixed: the row on a one-ended first <node>; in a state body described constraint- and notation-tier diagnostics and cited CodeFirstNamesNoTarget / TestTransitionToFirstMarkerIsIllegal, neither of which exists. It now records the parser-level error[syntax] ("a one-ended first <node>; is only admitted by an action body; elsewhere a succession names both ends") with tests/parser/negative_test.go:TestNegative (one_ended_first_in_state_body), verified with sysml -validate.

Structural edits: six duplicate connector rows and one duplicate view-rendering row removed (the widest copies kept); the empty "Decisions to Reassess" table and its "issues disabled" sentence dropped (issues are enabled); the Lines column of the three implementation-file tables removed (2–6× stale, no gate keeps it); the bounds table gains OPENSYSML_MAX_INTEGER_BITS (eight bounds, not seven); the RPC table notes which RPCs are mapped elsewhere and which eight have no row yet.

Report proposals corrected against the evidence

• Cast/classification rows keep their §8.4.4.9.2 cites (that clause's note — a finite literal is a Rational — is what they cite); only the nonexistent 8.3.4.9 CastExpression and 8.4.4.2 ClassificationExpression cites moved to §8.2.5.8.1 / §8.3.4.8.17 / §8.4.4.9.6.
• Transition-scope row: executeAction does not exist in runtime/state_executor.go and was dropped with the other stale names.
• View-rendering duplicates: the copy naming the case/mixed/DOT/PlantUML forms was kept and the "table or matrix"/GridView wording folded into it.
• DesertKite prose re-measured on the InitialDesign branch: OOSEM.sysml is clean; DesertKite.sysml reports ten 'SysML Standard Diagrams'::gv references, one endpoint-not-a-node and one cross-file reference — not the three gv errors proposed — and its qualified bind ends and requirement-body connection connect parse.
• API JSON round-trip row written without hard counts: the expectation file holds 832 stable / 8 whitespace-only / 1 unwritable / 38 refused:syntax / 1 graph-diff, not the proposed 830/7/1/38/0.

Decisions pending (status left unchanged; stale text only corrected)

• Unbound calculation result: typed error (runCalcBody, errCalcNoReturn) vs no value — the specification is silent (KerML §7.4.8.2); the missing-return-expression row should carry the same answer.
• Exit behaviour default payload (grounded on PSSM) and an unperformed owned action never running (grounded partly on fUML): re-ground on KerML §7.4.11 / SysML §8.4.7, §8.4.13.11, or mark as extension/approximate.
• seq#(i) out of range: typed error here, null by the library bodies of head/last and the pilot.
• Real finite-only: §9.3.2.2.9 defines the extended reals (±∞); the runtime's refusal of non-finite results is a tool decision the prose states as the specification's.
• The four validation-rule sections and constraint-manifest.json are keyed to the pilot's constraint names and the formal 1.0/2.0 editions, not Beta 2 (renamed constraints, three Beta 2-only constraints satisfied by construction).
• One-member CompositeConstraint: accepted as a degenerate conjunction vs refused per [2..*].

Overlap with other PRs and the develop merge

#1522, #1530, #1471 and #1484 landed in develop while this was open; origin/develop (11c4699) is merged in as a merge commit (no rebase). The one conflict was #1471's reworded HappensBefore partial-order row: develop's new row text is kept with this change's clause-number correction (§7.4.9 → §9.2.4 Occurrences, §9.2.6 Performances) re-applied to it. #1484's Census line and validatePartUsagePartDefinition row and #1530's checkConnectorTypeFeaturing row merged clean and are kept as develop has them. #1529 (pilot pin prose) is still open; one row here now names 2026-08 / 0.62.0, which agrees with it.

How it was verified

Every replacement was checked against the tree (grep for each path, symbol and test name) and the rendered specifications (build/spec-docs/sysml.txt, build/spec-docs/kerml.txt). Gates, all passing on the pushed commit:

go run -C tools ./cmd/doc-counts -check          → doc-counts: already current
go run -C tools ./cmd/validation-census -check   → 217 constraints, census document, compliance map and manifest agree
python3 scripts/check-doc-ids.py                 → No internal work-item labels in reader-facing documentation
python3 scripts/check-doc-links.py               → 0 broken link(s)
python3 scripts/check-doc-figures.py             → every quoted oracle figure names the round it measured
python3 scripts/changelog.py check               → exit 0
python -m mkdocs build --strict                  → Documentation built

Also checked by hand: sysml -validate on a state body with first a; (the syntax error above), the DesertKite models, and that every Status cell carries one marker and every | inside a code span in a table cell is escaped.

Checklist

• make test and make lint pass locally (docs-only change; the seven documentation gates above)
• Tests added or updated for the change (not applicable — documentation only)
• 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

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/0cadee59142045a995c05b473ed44534
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/0cadee59142045a995c05b473ed44534?variant=devin
Requested by: @HuiJun

[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.