Files
lmxopcua/docs/ui-consistency/evidence/ot09-deployments/README.md
T

4.0 KiB
Raw Blame History

OT09 deployment request versus application status

Frozen scope: Deployments.razor, new DeploymentFeedback, and three test files listed in source-freeze.json. Shared Theme stays published 1.0.1. No actor/backend/authorization policy changes.

Actual StartDeploymentOutcome is separate from recorded DeploymentStatus. Accepted means in progress, NoChanges is informational, an unknown transport result invents no verdict, and a new accepted ID never borrows the previous deployment’s terminal status. Sealed degradation and full failure metadata remain visible. Short live status is separate from escaped native diagnostic disclosures. Requested ID/revision and UTC machine-readable last-fetched time remain available. Manual Refresh reads the existing bounded scalar history; no automatic polling or artifact fetch was added. Failed refresh preserves rows, timestamp and accepted outcome. Initial unavailability does not masquerade as empty history. Pending request/read guards prevent reentry. Disposal cancels view-owned reads, never an already-dispatched mutation; disposal during auth prevents a new dispatch.

Verification

  • tests.log/tests.trx: 37 passed, 0 failed/skipped. Includes 18 feedback cases, 11 page cases, existing drift/API contracts and a real-provider page test. The latter seeds 60 synthetic rows in the owned disposable SQL instance, verifies the real translated query returns 50 rows and excludes ArtifactBlob, then removes its own IDs. The test runner receives private credentials only through child environment; no credentials are retained here.
  • Command from repository root: python3 infra/tools/ui-consistency/with_private_sql.py --state <private-state-path> -- dotnet test tests/Server/ZB.MOM.WW.OtOpcUa.AdminUI.Tests/ZB.MOM.WW.OtOpcUa.AdminUI.Tests.csproj -c Release --no-restore --filter "FullyQualifiedName~DeploymentFeedbackTests|FullyQualifiedName~DeploymentRefreshTests|FullyQualifiedName~DeploymentPageSqlServerTests|FullyQualifiedName~DeploymentDriftProbeTests|FullyQualifiedName~DeploymentStatusEndpointTests" --logger "trx;LogFileName=ot09-final-deployment.trx" --results-directory /tmp/ot09-final-deployment-results. Raw TRX records the exact five classes and all 37 identities.
  • fixture-build.log: actual AdminUI component fixture compiles with zero warnings/errors.
  • confirmation-matrix.json: 24/24 browser cases, 1440/390 at 100% and real browser 200% (720/195 effective CSS widths), synthetic accept/no-change/rejected/unknown/blocked/initial-read-failure. Actual double-click dispatches once; manual status recovery, failed-refresh retention, no polling/reload/body overflow and retained per-node link are checked. Synthetic fixture uses memory-only records/fabricated actor replies; no fleet writes.

Retained failures and corrections

Two initial harness runs ended on embedded-JavaScript selector quoting SyntaxErrors; the historically named first-fixture-hydration-reset.log is not evidence of a product hydration defect. Both logs are retained. The first complete matrix passed 20/24: case-sensitive checks failed on visually uppercase chip text; one real 195px chip exceeded its panel by 13.24px. accepted-status-195-before.json/png preserve that defect. The local display is now short “Awaiting acks” plus a wrapped full “Awaiting apply acknowledgements” caption; domain enum/status is unchanged. Corrected comparison checks and bounded confirmation pass all24. Zoomed matrix photographs may be poorly framed by Chromium viewport capture after scrolling; DOM geometry remains the recorded measurement. A separate tall-viewport 195px photo is framing confirmation only, not another behavior matrix.

Limits / rollout

This is a component canary and focused test slice, not a full production fleet or deployment test. Full combined consumer suite, whole-route accessibility/zoom matrix, package provenance and rollback remain OT14. No deployment or production configuration mutation was performed. Roll back this scoped source commit if regression is found; keep published package versions immutable.