fix(js): validate route_idx in v2 header check; uniform events[x].type

[ukint-vs]

Summary

Two related sails-js bugfixes from open issues without linked PRs.

• sails-js: _assertMatchingHeader should validate route_idx for multi-mounted services #1316 — _assertMatchingHeader ignored route_idx, so off-route replies/events for a multi-mounted interface decoded silently against the wrong service instance.
• sails-js v2: named-field event .type is an object while unit/unnamed variants return strings #1318 — events[x].type returned an object for named-field event variants and a string for unit/tuple variants. Consumers had to special-case named-field or got [object Object]. Same bug exists in v1 (js/src/sails.ts); fixed in both paths.

Changes

Commit	What
ef074b87	_assertMatchingHeader (v2) extended with route_idx validation.
3b02d760	events[x].type (v2) routes the already-computed typeStr to the public field; drops the unused intermediate t.
608ae352	Same one-line fix applied to v1 events at js/src/sails.ts:264. Existing service.test.ts events test gains a named-field event variant.
77549ada	Cleanup pass: replace local byte-concat with u8aConcat from @polkadot/util, hoist parser.parse(idl) into beforeAll, drop one as any cast, trim two over-explanatory comments.
5383b328	Codex review follow-up: honor received.routeIdx === 0 as the spec's inference sentinel (was previously rejected when expected was concrete); apply the same route check to events[x].is() and the subscribe predicate so they no longer disagree with decode() in multi-route programs.

Decisions worth noting

• Symmetric route_idx rule (after Codex review): throw only when both sides assert concrete (non-zero) routes that disagree. Either side being 0 is the inference sentinel per docs/sails-header-v1-spec.md §6.2 line 78 and §13.6 line 128 — accepted in both directions. This still catches the original sails-js: _assertMatchingHeader should validate route_idx for multi-mounted services #1316 bug (RMRK-style off-route delivery) while preserving spec-compliant inference for legacy/single-instance receivers.
• is() and subscribe parity: both predicates now share matchesRoute() with decode(), so the common if (event.is(e)) event.decode(e) pattern is consistent. Without this, is() returned true for the wrong service while decode() threw — a confusing surface.
• decodeResult not covered here: it doesn't call _assertMatchingHeader on master. PR fix(js): validate reply prefix in decodeResult #1315 (fix/js-decoderesult-header) adds that call site; once it lands, decodeResult automatically inherits this PR's route_idx validation.
• sails-js v2: named-field event .type is an object while unit/unnamed variants return strings #1318 is a runtime breaking change for consumers reading .type.fields on named-field events. The documented TS type is any, so it's not a TypeScript-level break. Migration: read event.typeDef.fields (already public, unchanged).

Tests

15 new test cases + 1 extended test case across 3 files. All pass:

$ pnpm jest --runInBand --testPathPatterns "route-idx-header-validation |event-type-shape|service.test"
Test Suites: 3 passed, 3 total
Tests:       17 passed, 17 total

Test plan:

• Function decodePayload rejects mismatched concrete route_idx with "Header mismatch" referencing both expected and got values
• Event decode rejects mismatched concrete route_idx
• Ctors accept any received route_idx (asymmetric expected-only rule)
• Received route_idx === 0 is accepted as the inference sentinel (function + event)
• event.is() returns false for wrong concrete route_idx (consistent with decode())
• event.is() returns true for inference sentinel route_idx === 0
• Matching route_idx decodes successfully (positive case)
• events[x].type is a string for unit, tuple, and named-field variants (v2)
• Named-field event round-trips through event.decode(...) after the fix
• event.typeDef.fields escape hatch still works
• v1 service.test.ts extended with named-field event — pre-fix returned object, post-fix returns string
• Existing offline test suites (partial-idl, idl-v2-type-resolver, encode-decode, enum, struct) still pass

Review

• Local /review pass: clean
• Codex review (round 1): 2 [P2] findings, both addressed in 5383b328
• Lint: ESLint clean

Closes #1316
Closes #1318

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request implements route index validation within the header matching logic, specifically allowing a sentinel value of zero for constructors to support inference. Additionally, it standardizes the representation of event types to ensure they are consistently returned as strings across both v1 and v2 IDL paths, resolving previous inconsistencies with named-field variants. Comprehensive test suites have been added to verify the new validation logic and the uniform shape of event types. I have no feedback to provide as the changes correctly address the requirements and include appropriate test coverage.