[devin-ai-integration]

What and why

Merge order

Merge after #2121 (fix/then-succession-member). This branch was cut from develop with #2121 merged in and builds on it; once #2121 lands, develop is merged back here so the diff shows only this change.

What this fixes

Follow-up to the audit of #2121 against the pinned specification (SysML v2.1 Beta 2, KerML 1.1 Beta 2) and the pinned pilot (scripts/pilot-pin.sh: tag 2026-09, artifact 0.63.0). Each finding was re-verified against the spec text, the pinned SysML.xtext and the pilot's export-json before changing anything.

1. State-body then x; after a state is a TransitionUsage (fixed)

SysML §8.2.2.18.1 StateBodyItem (BehaviorUsageMember (TargetTransitionUsageMember)*), §8.2.2.18.3 TargetTransitionUsage : TransitionUsage, §7.18.3. state idle; then running; is a TransitionUsage the state owns through a FeatureMembership, which owns the SuccessionAsUsage to its target through an OwningMembership (TransitionSuccessionMember). #2121 reflected and wrote it as a SuccessionAsUsage of the state.

• ast.IsStateTargetTransition names the form (after a state; not after entry/do/exit, not member-attached); SuccessionAdmitsBody uses it.
• Symbols: buildDecl builds such an edge as an anonymous action-usage symbol whose DeclaresTransition() is true; records carry NodeTransition (Symbol.NodeKind).
• Semantic model: metaclass is TransitionUsage, the implicit base Actions::transitionActions; every transition (this form and transition …/accept …) owns its succession as a synthesized part (ExpressionTransitionSuccession) reflected as OwningMembership → SuccessionAsUsage, answered as succession, with relatedFeature = source state, target state. The query leg reads these through OwnedRelationships.
• Writers: the RDF/API JSON writer emits a TransitionUsage (sysx:transitionSyntax "target") owning its succession with two ends, sysml:source the preceding state and sysml:target; the notation reader writes then x; / then x { … } back (the body is the transition's ActionBody).
• entry; then idle; (EntryTransitionMember) and then state off; / then part p2; (SourceSuccessionMember) stay SuccessionAsUsage (§8.2.2.9.3, §8.3.13.6; KerML §8.3.4.5.4).

Pilot evidence (export-json, state def Mach { entry; then idle; state idle; then running; … then state off; }): FeatureMembership → SuccessionAsUsage (entry), FeatureMembership → StateUsage(idle), FeatureMembership → TransitionUsage { Membership ⇒ idle, OwningMembership → SuccessionAsUsage(relatedFeature idle, running), FeatureMembership → transitionLink }, FeatureMembership → SuccessionAsUsage (then state off); a named transition t first off accept Sig then idle; owns its succession the same way.

Tests: semantics/state_then_transition_test.go (metaclass, memberships, relatedFeature, succession, nested state body), parser/succession_body_admitted_test.go TestIsStateTargetTransition, tests/export/behavior_test.go TestStateMembersRoundTrip (then idle;, then idle { … }, entry; then idle;, then state busy;), reflective_transition_kinds_test.go now also expects the transition's succession membership.

2. Connector ends and relatedFeature (approximate; known limitation)

No branch or open PR covers fix/succession-connector-ends-related-features; issue #2068 stays open. Not done here: the succession shorthand still owns no end features (TargetSuccession's SourceEndMember/ConnectorEndMember, §8.2.2.17.8; KerML §8.3.4.5.3 derives relatedFeature from them; the pilot exports both as ReferenceUsages under EndFeatureMemberships). relatedFeature stays derived from the members the then is written between: since #2143 a bare entry; performs an anonymous action, so entry; then idle; now answers (that action, idle), while then assign …; remains unanswered because the statement has no element on the symbol side (pre-existing, outside #2121). Recorded as ⚠️ in both compliance areas.

3. SuccessionAdmitsBody and then done { … } (fixed)

then done { … } derives from TargetSuccession UsageBody (§8.2.2.17.8) and now admits its body; the comment/test that attributed a body after a state in a state body to ActionTargetSuccession now says it is the ActionBody of TargetTransitionUsage.

4. Compliance map (fixed)

Rows added in 13-action.md (owned body-less SuccessionAsUsage, ⚠️ for the ends) and 15-state-machine.md (state-body then metaclass ✅; successions' owned ends ⚠️), with the pinned edition's clauses: SysML §8.2.2.17.1/.8, §8.2.2.9.3, §8.2.2.18.1/.3, §8.3.13.6; KerML §8.3.4.5.3/.4. Changelog fragment changes/unreleased/state-then-transition-usage.fixed.md.

Known limitations

• Successions written as then shorthand own no end features; relatedFeature is unanswered where an end is a statement (then assign …;) — Successions lose their related features unless named; transition successions, action-node ends and body-declared connection ends have none #2068.
• A two-ended first x then y; in an action body has no SuccessionAsUsage symbol; if x then a; / else b; after a decide have no membership on the symbol side (pre-existing, noted by the audit; the bare entry; case is closed by fix(symbols,semantics,export): a bare entry;, do; or exit; performs the anonymous empty ActionUsage its StateSubactionMembership owns — the symbol structure, the semantic model, the query leg and the RDF writer state it, both readers take it back, as the pilot exports it #2143).
• The editor still refuses an inline Route/Style on then done; although its production admits a body: it has no insertion point for one (the done is declared between the keyword and the ;), stated as such in inlineAnnotationAdmitted.
• Records (cached documents) carry the transition's metaclass but not its parts, so the owned succession is reflected from the live declaration only — as a transition's trigger and guard already are.
• A then x; after a non-behavior member in a state body (entry; attribute a; then idle;), which the pinned grammar does not admit and the pilot fails to parse, is classified as a TargetTransitionUsage, following the parser's existing leniency.

Baselines moved

• internal/translate/export/testdata/structure_differential_expected.txt (pilot-corpora root): six state-body then successions become TransitionUsages; their synthesized successions are now mapped by the differential (TransitionUsage::succession).
• tests/export/testdata/convert/then_after_members.golden.ttl.sha256: state def Switch { entry; attribute pressed; then idle; … } is written as a TransitionUsage.
• internal/semantic/metamodel/derived_census_baseline.json / derived_census_records.tsv (derived-census -update): element and implied counts rise in every root because each transition now owns its succession and that succession's implied base; no rule verdict moves.

Specification basis

SysML v2.1 Beta 2: §8.2.2.18.1 StateBodyItem / TargetTransitionUsageMember, §8.2.2.18.3 TargetTransitionUsage : TransitionUsage, §7.18.3; §8.2.2.17.1 ActionBodyItem, §8.2.2.17.8 ActionTargetSuccession / TargetSuccession UsageBody; §8.2.2.9.3 SourceSuccessionMember; §8.3.13.6 SuccessionAsUsage. KerML 1.1 Beta 2: §8.3.4.5.3 Connector (relatedFeature from connectorEnd), §8.3.4.5.4 Succession. Rows added in docs/project/spec-compliance/02-detailed-semantic-compliance-map/13-action.md and 15-state-machine.md (sources; the combined page is not committed).

How it was verified

• gofmt -l . (empty), go build ./..., go vet ./..., make lint (staticcheck + gosec) clean.
• go test ./... green (the runtime, smt, codegen, cmd/sysml and tests/export packages re-run with -timeout 45m after hitting the 10m default under load); go test -count=1 ./tests/hygiene green.
• TestSymbolStructureMatchesGraphStructure re-recorded with -update-structure-differential after adjudication (above).
• go run -C tools ./cmd/pilot-apijson-diff -check moved three pilot files (AssignmentTest, 6-Individual and Snapshots, 10c-Fuel Economy Analysis): the StateDefinition.ownedConnection / ExhibitStateUsage.nestedConnection rows leave present-different for indeterminate, as the state no longer owns the then succession as a connection and the transition's synthesized elements are unkeyable; re-recorded with -update; go test -C tools ./referee/apijson green.
• python3 scripts/spec_compliance.py check, python3 scripts/changelog.py check, python3 scripts/check-doc-ids.py clean.

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

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/646cfbd400b24080a168f29eb053ccd9
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/646cfbd400b24080a168f29eb053ccd9?variant=devin
Requested by: @HuiJun

---

🟢 Devin Review found nothing to fix · View findings · Follow Devin's work

[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.