TRUSTWEAVE SDK · 10 SEPTEMBER 2026

Testing and documentation strengthened

Ignored regression tests now execute, validation tools reject misleading evidence, and copyable examples stay synchronized with tested source.

8.5 → 9.0testing and documentation / 10
132 + 39Kotlin and validation-tool tests passed
9.0 / 10overall engineering score
Qualification update: Testing and documentation now score 9.5/10 after full SDK and committed-candidate hosted validation. Read the new evidence and remaining limits.

Target: 10/10 remains open

This review improves the existing combined category to 9.0/10. It does not assign a perfect score to partial evidence. A current-tree full SDK integration/coverage run, complete independent conformance for supported profiles, wider snippet review and hosted release validation remain necessary. Docker did not respond locally; the blocked S3 integration run was stopped and is not counted as passing.

Concrete acceptance checklist and remaining tasks · Scoring rationale

Defects fixed

The compiled JVM audit found eight annotated methods with non-void return types, which JUnit had ignored: six OIDC4VP signature/URL/error regressions, wallet recovery cancellation and typed encryption-key validation. Explicit Kotlin Unit signatures now make them discoverable, and their named XML results all pass.

The coverage checker previously accepted empty or invalid policies. It now rejects empty scopes, invalid or non-finite percentages, unknown metrics, duplicate scopes/counters, malformed counts and missing evidence. Existing global coverage floors are unchanged.

Documentation source markers could previously disappear or point to local-only sources. The checker now rejects missing/duplicated required markers, wrong fence languages, ignored/non-Kotlin/outside-repository sources and drift. Five required examples are recorded in a versioned contract.

Before: actual invalid JVM descriptors · After: compiled audit · Named executed regressions

Tests and coverage

Seven new host edge-case tests cover all host labels, client/server status classification, unknown method cardinality, queue timeout and admission, exception/cancellation cleanup, original connection and rollback semantics, absent/closed pools, and authentication boundaries. Two complete documentation test files verify fixture isolation/cleanup and authenticated host metrics.

Measured host line coverage increases from 96.52% to 97.83%, and branch coverage from 74.32% to 87.16%. The new package floor is 97% lines and 85% branches. No production-code exclusions or reduced repository-wide thresholds were used. This is host coverage, not merged SDK coverage.

ModulePassing testsExecuted scope
observability17Full module suite
credentials/plugins/oidc4vp45Full module suite
wallet/plugins/cloud3CloudRecoveryTest only
wallet/plugins/file26Full module suite
testkit13DocumentationExampleTest and TrustWeaveTestFixtureTest only
distribution/examples28Full module suite

Listed XML totals: 132 tests, zero failures/errors/skips. All five affected Kotlin modules passed lint. The documentation task also executed 24 registered example tasks. These counts exclude the interrupted Docker-backed run.

The validation tools have 39 passing tests, including real javac-compiled JUnit fixtures and adversarial policy/XML/source-marker cases. The compiled audit covers 3783 available annotated methods; that inventory is not a claim that every SDK test ran again.

Kover XML · Validation-tool results · Successful scoped Gradle log · Suites, filters, task names, source hashes and limits

Documentation and CI

Three overlapping testing pages now use accurate commands, real source references and explicit evidence boundaries. The fixture and host examples are full compiled tests. Required source markers cannot be removed silently. The checker reports 363 Markdown files, 2359 inventoried Kotlin blocks, five synchronized examples and zero errors. Other blocks remain fragments or unqualified examples; they are not all compiler-verified.

CI and release workflows now test the validation tools, check current documentation, reject invalid compiled tests, require named passing regressions, enforce host coverage, and retain diagnostics. The docs workflow executes the new examples and triggers for nested Markdown. YAML and local commands were validated; hosted jobs were not run.

Rewritten testing guide · Executable host example · Requirement-to-test contract · Required source examples

Updated scorecard

CategoryBeforeNow
Security and access control9.09.0
Observability and diagnosability9.69.6
Reliability and scale9.09.0
Configuration and data9.09.0
Deployment and release8.58.5
Testing and documentation8.59.0

The same six equal weights give 54.1 / 6 = 9.0167, rounded to 9.0/10. Only testing/documentation is reassessed. Observability remains 9.6; the other categories carry forward unchanged. Scores are engineering judgment, not certification.

Previous observability assessment · Remaining independent conformance matrix · Docker prerequisite result