Files
scadaproj/docs/plans/2026-11-12-scada-ui-consistency/ignitionoee.md
T

45 KiB
Raw Blame History

ignitionoee UI consistency Implementation Plan

For Codex: REQUIRED SUB-SKILL: Use executeplan to implement this plan task-by-task.

Goal: Converge router, registry and native Perspective UI behavior and visual tokens while preserving their separate runtimes, roles and operational truth.

Architecture: The router adopts the verified Theme package; the independent public-feed registry imports the versioned canonical static asset export without private NuGet prerequisites. Perspective maps exported semantic tokens to native view styles, with no Blazor package or admin-rail requirement on operator panels.

Tech Stack: Router/registry .NET 10 Blazor + xUnit, registry Playwright 1.58.0, Ignition 8.3+ Perspective/Jython 2.7, CPython offline tests and vendored view schemas.


Required testing prerequisite

Complete bUnit standardization and close GB before any task in this chapter, including local fixes and candidate UI pilots. All bUnit snippets below use the post-GB BunitContext / Render / AddAuthorization API; their inherited helpers must be migrated in GB. Preserve framework HtmlRenderer methods and xUnit runner APIs where used. Non-bUnit runtimes retain their existing harnesses, but their UI work also waits for the family gate. Old package pins are discovery baselines, not targets for this chapter.

Execution contract

  • Repository: /Users/dohertj2/Desktop/ignitionoee. All source/test paths below are relative to that root; use it as cwd unless a step states otherwise.
  • Planning date: 2026-11-12. This chapter specifies future work; no implementation or test execution is claimed.
  • Read the corresponding /Users/dohertj2/Desktop/scadaproj/impeccable/ignitionoee.md, the shared design-system report, the master plan and applicable AGENTS.md before execution.
  • Reuse a suitable checkout; inspect git status --short and preserve user edits. No new worktree, commit, push, package publication or deployment is performed by authoring this plan.
  • Dependency gate: complete shared Theme G1 compatibility audit, assign THEME_VERSION, and verify its artifact before consumer migration. THEME_VERSION is a placeholder, never a literal package version to publish.
  • Shared G1 supplies semantic status ink (--ok-ink:#236f34, --warn-ink:#855000, --bad-ink:#b42121, --idle-ink:#535d66), --ink-faint:#626970, and independent blue --info-ink; product accent never recolors semantic informational notices.
  • G1 adds TechField.For, HintId, ErrorId; callers supply input IDs, aria-describedby and aria-invalid. G1 adds ThemeShell.AccentDeep, AccentTint, AccentBorder, AccentFocus while preserving Accent.
  • Shared accent aliases must resolve in shell scope: G1 uses --accent-tint directly or rebinds inherited aliases at .app-shell. Bootstrap .btn-primary/ghost component variables must follow shell accent for normal/hover/active/disabled states. Compare computed states before deleting any consumer override.
  • G1 also supplies shared focus/busy/reduced-motion and filter/table/dialog recipes. TechButton.BusyText is optional; busy adds aria-busy. Caller-owned status/alert text announces outcomes, never secret values or every polling update.
  • Pilot native dialogs in two consumers before extracting a common dialog component. Do not build a new component package during a first consumer migration.
  • Restore/build first and record actual dependency/runtime failures. Complete GB bUnit/parser standardization first. Preserve xUnit, test SDK and Playwright versions unless a specific demonstrated compatibility fix is recorded in GB; do not perform later incidental framework upgrades. Never treat a skipped native/SQL/COM/live test as a pass.
  • Each numbered step is one short action. When a step discovers broader product behavior, stop that task at its boundary and update this written plan before expanding implementation scope.

Current source and hard boundaries

  • Read /Users/dohertj2/Desktop/ignitionoee/AGENTS.md, relevant CLAUDE.md sections, router/README.md, registry/README.md, and docs/reference/ignition-8.3-resource-format.md.
  • Router central package pin: router/Directory.Packages.props:85 Theme 0.5.0; Secrets.Ui 0.6.2. Update after G1 dependency audit; preserve shared Secrets compatibility through SECRETS_UI_VERSION.
  • Registry has no Theme package. registry/README.md explicitly keeps private router packages out of its prerequisites. Decision: consume verified canonical static exports; do not introduce a private feed or import router authentication/package configuration as a styling shortcut.
  • Router tests use xUnit and the existing PageRenderTest static renderer, not bUnit. Registry BrowserTests already have Playwright and disposable hosts/SQLite fixture data with cleanup. Reuse them without upgrades.
  • Router EF InMemory tests establish semantic behavior, not SQL Server translation/enforcement. Retain UiSqlTranslationTests whenever a query/deep-link task touches EF predicates.
  • Perspective browser UI is native Ignition; keep Jython 2.7 event code, current tag paths, named query contracts, roles and source-of-count semantics intact. JSON defaultSize is design-time metadata, not proof of runtime responsiveness.
  • Protected CR history and gateway heartbeat/forwarding rules in AGENTS.md remain untouched. No rebuild, re-forward, backfill, schema migration, PLC write, live driver action or downtime reclassification is part of styling validation.

Findings coverage

Report finding Planned task Runtime
OEE-01 click-only Jobs rows OEE-01 Router
OEE-02 unmanaged TechModal OEE-02 Router, using registry reference
OEE-03 copied registry theme OEE-03 Registry static asset import
OEE-04 Perspective token drift OEE-04 Native views
OEE-05 long nav footer visibility OEE-05 Router shared shell
OEE-06 scheduling context switches OEE-06 Router
OEE-07 issue recovery consistency OEE-07 Router
OEE-08 responsive toolbar/table reuse OEE-08 Router
OEE-09 eligibility vs runtime health OEE-09 Registry
OEE-10 operator touch density OEE-10 Native views

Package/export gate and baseline

  1. From repo root run dotnet restore router/ZB.MOM.Oee.slnx; save dotnet list router/src/ZB.MOM.Oee.Host/ZB.MOM.Oee.Host.csproj package --include-transitive to the validation record.
  2. After G1 audit assigns THEME_VERSION, update the central router pin, restore and prove the resolved graph includes its expected shared Theme/Secrets assets.
  3. Use Theme TH-08 export technical-light-web-assets.<THEME_VERSION>.zip; require its sourceTag, sourceRevision, sourcePackageSha256, version, per-file SHA256 and semantic token manifest. Verify the immutable artifact before importing into registry or generating native mappings.
  4. Run dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj as baseline. Record infrastructure failures and keep them distinct from design assertions.
  5. Run dotnet build registry/ZB.MOM.MachineRegistry.slnx, then dotnet test registry/tests/Registry.BrowserTests --filter FullyQualifiedName~RegistryShellTests after installing the pinned browser if needed.
  6. Browser setup when required: dotnet build registry/tests/Registry.BrowserTests; pwsh registry/tests/Registry.BrowserTests/bin/Debug/net10.0/playwright.ps1 install chromium. No silent browser version substitution. If PowerShell/browser install is unavailable, report the unmet browser gate.
  7. Add docs/ui-validation/technical-light-family.md to record fixture endpoints, process stop methods, screenshots and actual state/keyboard results. Keep real credentials and operator databases out of fixtures.

OEE-01: Give Jobs details a semantic keyboard disclosure

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Jobs.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Jobs.razor.cs
  • Create router/tests/ZB.MOM.Oee.Ui.Tests/JobsRowDisclosureTests.cs
  • Reuse router/tests/ZB.MOM.Oee.Ui.Tests/PageRenderTest.cs

Step 1 — Establish the failing behavior. Create the xUnit class with imports for Jobs, entities and PageRenderTest. Seed a current completed job so the page default date range includes it. The test below guards the actual emitted button semantics, not CSS structure.

[Fact]
public async Task Job_rows_emit_named_details_buttons()
{
    var id = Guid.Parse("11111111-2222-3333-4444-555555555555");
    var html = await PageRenderTest.RenderAsync<Jobs>(db => {
        var cell = PageRenderTest.AddHierarchy(db);
        var now = DateTime.UtcNow;
        db.Jobs.Add(new Job { JobId = id, CellId = cell.CellId,
            SourceName = "Operator", Status = JobStatuses.Completed,
            JobStepNumber = "0100", StartTime = now.AddMinutes(-10),
            EndTime = now, CreatedAt = now });
    });
    Assert.Contains("Details for job", html);
    Assert.Contains("aria-expanded=", html);
    Assert.DoesNotContain("<tr role=" + '"' + "button" + '"', html);
}

Step 2 — Run the focused check. From the repository root:

dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj --filter FullyQualifiedName~JobsRowDisclosureTests

Expected before: Jobs renders a click-only tr with role=button and no semantic details button.

Step 3 — Implement in small edits.

  1. Keep table row/cell semantics; add a real named details button with aria-expanded and stable controls ID based on full Guid.
  2. Use existing Toggle/_expanded state and retain full copyable job ID/work-order context; row click may remain only as an optional accelerator with propagation handled to avoid double toggles.
  3. Preserve SearchHref full-Guid behavior, 500-row window, server predicate and role/policy checks; no query rewrite for a UI disclosure.
  4. Keep focus on the trigger after expand/collapse; collapsing a row containing focused content returns focus to its trigger.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. On the isolated router fixture Tab to Details, activate with Enter and Space, verify aria-expanded and full ID/work-order detail, then collapse. Assert no double toggle. Re-run JobsQueryTests and JobsSearchWiringTests; source display labels retain real Active/Completed/Superseded statuses rather than the illustrative mockup vocabulary.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "fix: make OEE job details keyboard accessible". Do not include generated credentials, runtime data or unrelated edits.

OEE-02: Pilot complete native modal behavior in router

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Shared/TechModal.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/AddMachineModal.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/AddMachineModal.razor.cs
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/MachineEditorModal.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/MachineEditorModal.razor.cs
  • Create router/src/ZB.MOM.Oee.Ui/wwwroot/js/tech-modal.js
  • Create router/tests/ZB.MOM.Oee.Ui.Tests/TechModalRenderTests.cs
  • Modify router/tests/ZB.MOM.Oee.Ui.Tests/PageRenderTest.cs
  • Create docs/ui-validation/router-modal-browser.md

Step 1 — Establish the failing behavior. Add a static render test for the requested native dialog, plus actual browser assertions in the validation record. Reuse the existing registry driverCatalog.js behavior as reference, but do not import registry state or write handlers into the router.

[Fact]
public async Task Visible_modal_emits_a_named_native_dialog()
{
    // Extend PageRenderTest.RenderAsync with optional ParameterView support once;
    // preserve existing callers using ParameterView.Empty as the default.
    var html = await PageRenderTest.RenderAsync<TechModal>(
        parameters: ParameterView.FromDictionary(new Dictionary<string, object?> {
            ["Visible"] = true, ["Title"] = "Review machine mapping"
        }));
    Assert.Contains("<dialog", html);
    Assert.Contains("aria-labelledby=", html);
    Assert.Contains("Review machine mapping", html);
}

Step 2 — Run the focused check. From the repository root:

dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj --filter FullyQualifiedName~TechModalRenderTests

Expected before: TechModal emits an aria-modal div rather than native dialog; browser Tab can leave the overlay.

Step 3 — Implement in small edits.

  1. Add additive Busy and safe OnCancel/return-focus contract; preserve current callers until each is explicitly wired.
  2. Render native dialog, show it after interactive render, name title uniquely per instance and preserve typed-state cancel behavior.
  3. JS listeners use their real DOM event argument; cancel handler prevents default synchronously and consults current busy state before invoking Blazor. Never use window.event or treat aria-modal as inertness.
  4. Capture/restore trigger; dispose listeners and module; retain data on failed save. Use native showModal inertness, adding bounded fallback focus only where browser validation proves necessary.
  5. Inventory every TechModal caller: AddMachineModal, MachineEditorModal, MachineIdentifiersModal, FaultOverridesModal, ErpInstances and MachineTypes. Wire each safe cancel/busy path and verify no behind-dialog write is reachable.
  6. Complete this as consumer pilot 2 after MX pilot 1. Compare behavior before proposing shared extraction; extraction is a separate G1-follow-up, not an excuse to ship an untested common wrapper.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Run real browser focus checks in the isolated router host: trigger → open → Tab/Shift+Tab contained; outside button cannot focus; Escape cancels while idle; busy prevents dismissal/double-submit; closing restores trigger. Exercise a field validation error and disposed circuit. Static HtmlRenderer cannot prove these outcomes.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "fix: manage router modal focus and cancellation". Do not include generated credentials, runtime data or unrelated edits.

Exact static-render helper change for OEE-02

In router/tests/ZB.MOM.Oee.Ui.Tests/PageRenderTest.cs, append an optional argument to the existing method signature:

internal static async Task<string> RenderAsync<T>(
    Action<OeeDbContext>? seed = null,
    Action<IServiceCollection>? configure = null,
    ParameterView? parameters = null)
    where T : IComponent

Keep the existing body/fixtures unchanged except replace renderer.BeginRenderingComponent(typeof(T), ParameterView.Empty) with renderer.BeginRenderingComponent(typeof(T), parameters ?? ParameterView.Empty). Current Microsoft.AspNetCore.Components imports already supply ParameterView; existing callers keep their default behavior.

Actual browser focus acceptance for OEE-02

Run this in the existing browser automation session against an isolated router page with an editor account, a machine fixture and no real writes. page is the connected Playwright page; supply the actual accessible trigger name from that page. This is executable browser validation code, not an xUnit/static renderer assertion:

const trigger = page.getByRole('button', { name: 'Add machine or line', exact: true });
await trigger.focus();
await trigger.press('Enter');
const dialog = page.getByRole('dialog');
await dialog.waitFor({ state: 'visible' });
if (!(await dialog.evaluate(d => d.contains(document.activeElement))))
  throw new Error('Initial focus is outside dialog');
for (let i = 0; i < 30; i++) {
  await page.keyboard.press('Tab');
  if (!(await dialog.evaluate(d => d.contains(document.activeElement))))
    throw new Error('Tab escaped modal');
}
await page.keyboard.press('Escape');
await dialog.waitFor({ state: 'hidden' });
if (!(await trigger.evaluate(e => e === document.activeElement)))
  throw new Error('Trigger focus was not restored');

Use the exact accessible name found in the fixture; if the name is not “Add machine or line”, update the recorded selector to the real control without weakening the focus assertions. Add Shift+Tab, cancel-button and delayed-busy variants in the same browser pass. Stop fixture processes after validation.

OEE-03: Import canonical static assets without a private registry feed

Files:

  • Modify registry/src/Registry.Host/Components/App.razor
  • Modify registry/src/Registry.Host/wwwroot/app.css
  • Create registry/src/Registry.Host/wwwroot/shared-theme/css/theme.css
  • Create registry/src/Registry.Host/wwwroot/shared-theme/css/layout.css
  • Create registry/src/Registry.Host/wwwroot/shared-theme/fonts/ibm-plex-sans-400.woff2
  • Create registry/src/Registry.Host/wwwroot/shared-theme/fonts/ibm-plex-sans-600.woff2
  • Create registry/src/Registry.Host/wwwroot/shared-theme/fonts/ibm-plex-mono-500.woff2
  • Create registry/src/Registry.Host/wwwroot/shared-theme/tokens.json
  • Create registry/src/Registry.Host/wwwroot/shared-theme/manifest.json
  • Create deploy/scripts/import-technical-light-assets.py
  • Create tests/test_technical_light_asset_import.py
  • Create docs/design/technical-light/registry-asset-import.json
  • Modify registry/tests/Registry.BrowserTests/RegistryShellTests.cs

Step 1 — Establish the failing behavior. Capture current registry pages on a disposable seeded database; assert the intended imported CSS/fonts URLs will return 200 and the browser uses Plex with shared semantic ink. Before import these URLs are absent. This is a visual/static-asset migration, not a reason to add mirrored CSS unit tests.

Step 2 — Run the focused check. From the repository root:

dotnet build registry/ZB.MOM.MachineRegistry.slnx
dotnet test registry/tests/Registry.BrowserTests/Registry.BrowserTests.csproj --filter FullyQualifiedName~RegistryShellTests

Expected before: the new shared-theme asset requests are absent before import; current copied palette/font rules remain.

Step 3 — Implement in small edits.

  1. Verify TH-08 zip manifest version/sourceTag/sourceRevision/sourcePackageSha256 and every file SHA256; reject path traversal, missing files and unexpected versions in the importer before writing.
  2. Expose importer CLI --artifact ZIP --destination DIRECTORY --expected-version VERSION --check and require manifest validation before writing; omit --check only for the intentional import after a verified dry run. Extract only documented css/, fonts/, tokens.json and manifest.json under wwwroot/shared-theme; save artifact checksum/import provenance in docs/design/technical-light/registry-asset-import.json.
  3. Load shared-theme/css/theme.css then layout.css before registry app/domain sheets. Verify relative ../fonts URLs; do not load a second Plex declaration afterward.
  4. Remove duplicate :root palette/font/base rules from app.css only after comparing computed declarations. Registry currently uses shell/side-rail/main/button/table selectors; preserve necessary registry chassis geometry and local dialog/floorplan/evidence selectors until explicitly migrated.
  5. Map local --hover/--active aliases to canonical --hover-bg/--active-bg where necessary. Use semantic --*-ink for text and independent blue --info-ink. Keep eligibility meaning local.
  6. Update RegistryShellTests with asset-load/contrast/overflow assertions in the existing disposable process harness. Do not change public NuGet feeds or startup permissions.

Step 4 — Verify the change. Add/run python3 -m unittest discover -s tests -p test_technical_light_asset_import.py -v for manifest/hash rejection, archive path escape rejection, check-mode non-mutation and successful verified import. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Run registry Inventory, Machines, Drivers/editor, Identity, Verification and Floorplans at all specified sizes. Preserve keyboard map alternatives and 200% zoom. Verify one active nav item, no font 404s, native editor focus, existing button geometry and no document overflow. Keep current domain CSS if a shared generic selector changes its behavior until mapped deliberately.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "style: import canonical Technical-Light registry assets". Do not include generated credentials, runtime data or unrelated edits.

OEE-04: Map shared tokens into native Perspective views

Files:

  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/KpiTile/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/CellTile/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/NavHeader/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/Overview/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/ReasonPicker/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/JobEntry/view.json
  • Create docs/design/technical-light/perspective-token-map.json
  • Create deploy/scripts/apply-perspective-theme.py
  • Create docs/ui-validation/perspective-technical-light.md

Step 1 — Establish the failing behavior. Use a bounded source/visual inventory of literal surface/ink/status colors and compare them with the verified export manifest. The first pass changes style properties only; no new style-class resource format is invented. Test the migration tool dry-run and schema/live acceptance rather than assertions mirroring every color declaration.

Step 2 — Run the focused check. From the repository root:

python3 deploy/scripts/apply-perspective-theme.py --check
bash deploy/scripts/ign-deploy.sh validate --all projects/OEE_Site

Expected before: before tool creation the command is missing; after tool creation but before applying it reports only allowlisted divergent style paths. Validator baseline may already pass and is not evidence of visual conformity.

Step 3 — Implement in small edits.

  1. Consume TH-08 tokens.json and source/hash provenance; define explicit native props.style mappings by view/component path, not arbitrary global string replacement.
  2. Implement --check and --apply using Python stdlib; --check does not mutate. Reject attempts to alter events, propConfig bindings, params other than approved density inputs, tag paths or named-query references.
  3. Apply paper/card/ink/semantic status/spacing/font mappings to the six exact views listed. Preserve dynamic status expressions; map their values through canonical semantics without converting unavailable to zero.
  4. Use existing native props.style initially. The repository has no verified named-style resource example; defer named-style extraction until Designer export establishes its manifest/payload format and this file table is updated.
  5. Do not add Theme/Blazor classes, Bootstrap CSS or admin side rail to Perspective. Use the existing native operator navigation and preserve Jython 2.7 event scripts.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Run offline view validation, then import into a disposable Ignition Designer/session using the repository workflow. Check kiosk and handheld rendering, computed contrast, fonts/fallback availability and dynamic state colors. Designer/live resource acceptance is a required gate; offline JSON validation alone is insufficient.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "style: map shared Technical-Light tokens to Perspective views". Do not include generated credentials, runtime data or unrelated edits.

OEE-05: Keep long router navigation and sign-out reachable

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Layout/MainLayout.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Layout/NavMenu.razor
  • Update docs/ui-validation/technical-light-family.md

Step 1 — Establish the failing behavior. At 1024×768 and 200% zoom, expand all authorized nav sections and record whether footer/sign-out are visible/reachable. Do not add a test that only searches source for ScrollNav.

Step 2 — Run the focused check. From the repository root:

dotnet build router/src/ZB.MOM.Oee.Host/ZB.MOM.Oee.Host.csproj

Expected before: the current layout does not opt into scroll nav; the full expanded rail can push footer beyond the viewport.

Step 3 — Implement in small edits.

  1. After THEME_VERSION gate, opt into ThemeShell ScrollNav=true.
  2. Keep nav keys and policy gates unchanged. If changing initial group expansion, preserve stored state and never hide current destination.
  3. Supply correctly bounded page/nav heights through G1 only; do not reintroduce local height:100vh wrappers that fight shared mobile-shell behavior.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Keyboard-reach every permitted nav item and Sign out at 1024×768 and zoom; compare mobile collapsed/open content remains visible. Confirm no double-scrolling trap or footer overlap and Viewer/Admin nav differences remain correct.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "style: keep router navigation footer reachable". Do not include generated credentials, runtime data or unrelated edits.

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Schedules.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleAssignments.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleAssignments.razor.cs
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleCalendar.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleCalendar.razor.cs
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/SchedulePreview.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/SchedulePreview.razor.cs
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleEditor.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ScheduleEditor.razor.cs
  • Create router/src/ZB.MOM.Oee.Ui/Logic/ScheduleScopeLinks.cs
  • Create router/tests/ZB.MOM.Oee.Ui.Tests/ScheduleScopeLinksTests.cs

Step 1 — Establish the failing behavior. Create a small pure link builder with named schedule/cell/site inputs and tests for encoding, missing optional scope and route names. This is navigation context, not a new schedule state store.

[Fact]
public void Preview_link_preserves_cell_scope_and_encodes_site_code()
{
    var href = ScheduleScopeLinks.Preview(cellId: 104, siteCode: "A&B");
    Assert.StartsWith("/schedule-preview?", href);
    Assert.Contains("cellId=104", href);
    Assert.Contains("site=A%26B", href);
}
[Fact]
public void Preview_without_scope_has_no_fabricated_selection()
{
    Assert.Equal("/schedule-preview", ScheduleScopeLinks.Preview(null, null));
}

Step 2 — Run the focused check. From the repository root:

dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj --filter FullyQualifiedName~ScheduleScopeLinksTests

Expected before: ScheduleScopeLinks does not exist; initially compilation fails, then behavioral expectations fail until implemented.

Step 3 — Implement in small edits.

  1. Current SchedulePreview/Calendar keep private selected IDs and expose no query hydration; ScheduleEditor has the route ScheduleId. Introduce a documented additive navigation contract: cellId is the selected cell ID, site is an optional encoded site code, and the editor retains its route ScheduleId. Resolve cell ancestors from validated loaded data and reject mismatching supplied site/cell context. Reuse existing query names if a later source audit finds they have been added; update the helper tests together before implementation.
  2. Build related-task links carrying applicable selected schedule/cell/site via query parameters. Treat them as related pages, not mandatory linear wizard steps.
  3. Hydrate only valid supplied context with server-validated IDs/codes; display absent/retired/denied scope clearly and keep existing selection fallback.
  4. Keep all schedule write guards, effective-time and denominator logic unchanged; read-only preview must not submit a schedule mutation.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Select one cell/schedule, visit assignments → calendar → preview and verify scope persists and denominator explanation links to the correct source/override. Check deep-link refresh, missing/retired scope and denied Editor action. Run ScheduleWriteGuardTests and existing schedule tests.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "feat: retain scheduling context across related pages". Do not include generated credentials, runtime data or unrelated edits.

OEE-07: Standardize issue-state feedback without hiding consequences

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ConfigIssues.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/ConfigIssues.razor.cs
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/EnrichmentIssues.razor
  • Create router/tests/ZB.MOM.Oee.Ui.Tests/ConfigIssuesFeedbackTests.cs
  • Modify router/tests/ZB.MOM.Oee.Ui.Tests/EnrichmentIssueBuilderTests.cs

Step 1 — Establish the failing behavior. Render ConfigIssues using existing PageRenderTest with an empty source and with a failing IDbContextFactory registration. Assert no-issues success differs from evaluation failure and that failure text is safe/recoverable. Add a sequential success/failure browser fixture for retained stale results.

Step 2 — Run the focused check. From the repository root:

dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj --filter FullyQualifiedName~ConfigIssuesFeedbackTests

Expected before: the current failed-load notice includes raw error text and has no consistent stale/retry summary.

Step 3 — Implement in small edits.

  1. Keep evaluated timestamp/count beside page heading; introduce explicit never-loaded/loading/loaded/refresh-failed state.
  2. On refresh failure retain previous result set marked stale, and expose Retry; first-load failure cannot say no issues.
  3. Use safe user summary with expandable sanitized technical detail; remove raw exception text from primary notice without suppressing server diagnostics.
  4. Preserve severity, affected scope, consequence and exact fixing link; align enrichment feedback to the same state/notice contract.
  5. Announce one caller-owned result summary after user re-evaluation, not individual rows or every polling tick.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Exercise no issues, one error, capped/large queue, denied fix target, DB failure then recovery and late completion after navigation. Confirm existing whole-Guid Jobs links remain correct and no exception leaks credentials/connection strings.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "fix: unify router issue queue recovery states". Do not include generated credentials, runtime data or unrelated edits.

OEE-08: Reuse wrapping toolbar/table contracts across router

Files:

  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Oee.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Jobs.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/SchedulePreview.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Rates/RatePreview.razor
  • Modify router/src/ZB.MOM.Oee.Ui/Components/Pages/Rates/RateCoverage.razor
  • Update docs/ui-validation/technical-light-family.md

Step 1 — Establish the failing behavior. Capture the already-repaired OEE toolbar as a reference, plus Jobs/schedule/rate screens at 390px and zoom. This is responsive composition: verify actual bounding boxes/overflow rather than copying styles into assertion strings.

Step 2 — Run the focused check. From the repository root:

dotnet build router/src/ZB.MOM.Oee.Host/ZB.MOM.Oee.Host.csproj
dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj --filter FullyQualifiedName~UiSqlTranslationTests

Expected before: OEE has a local container query while related toolbars need separate measured comparison; record any baseline overflow instead of assuming it.

Step 3 — Implement in small edits.

  1. Use G1 wrapping filter utility where it matches the existing OEE fix; remove only the superseded page-local query.
  2. Keep controls labeled and dates/scopes visible when wrapped; apply coarse-pointer density through shared recipe.
  3. Contain wide tables in named focusable scrolling regions; keep critical identifiers and explicit details controls reachable.
  4. Preserve OEE fact-window cap, unknown-value handling, aggregation, missing-row semantics and SQL queries exactly.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Verify desktop/mobile/zoom/coarse-pointer states and actual document width. Recheck OEE count integrity and scope totals using existing tests; no layout-only unit tests. At most one batched visual correction pass after the initial comparison.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "style: reuse responsive router filter and table patterns". Do not include generated credentials, runtime data or unrelated edits.

OEE-09: Keep eligibility and verification semantics distinct from health

Files:

  • Modify registry/src/Registry.Host/Components/Pages/Drivers.razor
  • Modify registry/src/Registry.Host/wwwroot/app.css
  • Modify registry/tests/Registry.BrowserTests/RegistryWorkflowsTests.cs
  • Modify registry/tests/Registry.BrowserTests/RegistryOperationsTests.cs

Step 1 — Establish the failing behavior. The current catalog already states the correct eligibility meaning. Add preservation assertions in an existing disposable browser workflow; this is a regression guard around the asset migration, so it can legitimately pass before styling. Do not manufacture a failing behavior when the semantic behavior already exists.

// Insert in the existing async browser workflow after its fixture starts:
await page.GotoAsync(url + "/drivers");
await Assertions.Expect(page.GetByRole(AriaRole.Heading,
    new() { Name = "Driver catalog", Exact = true })).ToBeVisibleAsync();
await Assertions.Expect(page.GetByText(
    "Production eligibility permits selection for a mapping. It does not deploy or commission a driver.",
    new() { Exact = true })).ToBeVisibleAsync();
Assert.Equal(0, await page.Locator(".driver-status")
    .Filter(new() { HasText = "Healthy" }).CountAsync());

Step 2 — Run the focused check. From the repository root:

dotnet test registry/tests/Registry.BrowserTests/Registry.BrowserTests.csproj --filter FullyQualifiedName~RegistryWorkflowsTests

Expected before: the semantic preservation test may pass at baseline; the gate is that it must remain green after OEE-03. Any styling change that replaces eligibility with Healthy must fail it.

Step 3 — Implement in small edits.

  1. Keep Eligible/Not eligible labels and copy explaining mapping selection rather than deployment/commissioning.
  2. Map the local eligibility/status class to appropriate shared semantic ink without renaming it to runtime health.
  3. Retain prerequisites, evidence, revisions, change reason and concurrency reload in the existing native editor.
  4. Preserve the floorplan keyboard location list and evidence distinctions under the same asset migration; no new operational-state inference from catalog records.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Use existing Playwright seeded registry harness and insert assertions immediately after navigating to /drivers. Open a driver, test field validation, cancel/return focus and revision conflict, then compare current mobile editor layout.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "test: preserve registry eligibility meaning during theme adoption". Do not include generated credentials, runtime data or unrelated edits.

OEE-10: Propagate operator touch density through reason entry

Files:

  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/ReasonPicker/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/JobEntry/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/PanelCell/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/CellDetail/view.json
  • Modify projects/OEE_Site/com.inductiveautomation.perspective/views/NavHeader/view.json
  • Create tests/test_operator_touch_context.py
  • Update docs/ui-validation/perspective-technical-light.md

Step 1 — Establish the failing behavior. Add a CPython unittest for the cross-view touch-context contract, not exact CSS values. ReasonPicker must accept an input touch flag and the native caller must explicitly pass it without changing downtime event routing. Use existing JSON walkers from test_cell_detail_view.py; do not add a new test framework.

Step 2 — Run the focused check. From the repository root:

python3 -m unittest discover -s tests -p test_operator_touch_context.py -v
bash deploy/scripts/ign-deploy.sh validate --all projects/OEE_Site

Expected before: ReasonPicker currently has no touch input contract, so the focused assertion fails before implementation.

Step 3 — Implement in small edits.

  1. Add a native input touch parameter with false default and proper propConfig; propagate existing caller touch context into ReasonPicker popups/embeds.
  2. Keep JobEntry touch behavior (already uses 64px controls) at least as usable; do not shrink working operator targets merely to match compact desktop admin rows.
  3. Apply G1 density mapping to dropdown/comment/save/cancel fields while keeping cell/current-job/stop-time context visible.
  4. Preserve role check and basePath/eventId/eventStart/mode/popupId contracts, reclassification behavior and close-only-on-clean-save event logic.
  5. Keep unavailable/unknown KPI values explicit and ensure top navigation remains usable on kiosk/handheld; do not replace process-panel bindings or introduce PLC actions.

Step 4 — Verify the change. Re-run Step 2; expected after: PASS for the named tests or all stated visual checks. Validate offline schemas and existing cell-count/manual-entry/downtime tests, then use disposable Ignition sessions at 1024/1280px panel and 390px handheld. Test live and reclassify modes, touch true/false, unavailable tag value, denied role and failed save (popup remains with readable feedback). No real downtime reclassification is required.

Step 5 — Commit. Stage only the listed production/test/evidence files, inspect git diff --cached, then git commit -m "style: carry operator touch context into reason entry". Do not include generated credentials, runtime data or unrelated edits.

OEE-10 failing native-context test

Create tests/test_operator_touch_context.py with Python 3 stdlib only. This checks whether callers can select the intended density variant; it does not assert particular CSS numbers or claim to emulate Ignition rendering.

import json
import pathlib
import unittest

ROOT = pathlib.Path(__file__).resolve().parents[1]
VIEWS = ROOT / "projects/OEE_Site/com.inductiveautomation.perspective/views"

class OperatorTouchContextTests(unittest.TestCase):
    def test_reason_picker_accepts_explicit_touch_input(self):
        view = json.loads((VIEWS / "ReasonPicker/view.json").read_text())
        self.assertIs(view["params"]["touch"], False)
        self.assertEqual("input", view["propConfig"]["params.touch"]["paramDirection"])

if __name__ == "__main__":
    unittest.main()

Add a second test that walks the actual caller event/embedded-view payload and checks touch is supplied, after inspecting its current native component path. Preserve behavioral script tests from the existing repository instead of asserting string equality for all Jython event code.

Cross-cutting state and viewport matrix

State Expected contract Verification
First load / slow work Page identity remains; busy action disabled with meaningful busy text; no false successful zero Delayed fixture and keyboard
Healthy Text label plus status token, sampled time, real scope Representative fixture
Warning / failed dependency Plain summary, affected scope and recovery; safe detail disclosure Fault fixture; no raw secret/connection string
Refresh fails after success Keep last good data visibly marked stale; retry does not clear context Sequential success/failure fake
Empty / no matches Different from loading/failure; clear-filter or next action Empty source and filtered-to-zero separately
Unavailable / unknown Show unknown/n/a, never substitute numeric zero or Healthy Null/missing fixture
Permission denied Route and action gates remain server-enforced; nav reflects existing policy Viewer/Admin/denied accounts
Long content Full identifiers available in detail/copy context; table scroll local 200-character path and error
Keyboard / modal Visible focus; enter/space activation; real modal inertness and focus restoration Browser test, not HtmlRenderer
Narrow / zoom / motion 1440×900, 1024×768, 390×844, 200% zoom; reduced motion keeps status feedback Batched visual pass and at most one confirmation pass

No runtime timestamps or status values in a mockup are deployment evidence. Keep status domain mapping app-owned. Test color contrast against actual computed backgrounds; screenshots alone do not establish WCAG compliance.

Native validation, release and rollback

Router

  • After focused tasks, run dotnet test router/tests/ZB.MOM.Oee.Ui.Tests/ZB.MOM.Oee.Ui.Tests.csproj and relevant Host/security tests once. Query changes require SqlServer translation tests; no InMemory green result proves a SQL execution plan or production permissions.
  • Validate all TechModal caller families, keyboard Jobs details, Viewer/Editor/Admin nav, long rail, mobile shell and shared accent/status states in an isolated router host.
  • Record actual package graph and THEME_VERSION; retain compatible Secrets.Ui rather than accidentally resolving two Theme generations.

Registry

  • Run dotnet build registry/ZB.MOM.MachineRegistry.slnx, dotnet test registry/tests/Registry.BrowserTests, and dotnet test registry/tests/Registry.Tests --filter 'Category!=RealInputs' for synthetic checks on a code-only checkout.
  • Category=RealInputs requires local migration workbooks and SourceInfo files. Run it only with those inputs supplied and explicitly record if missing; do not claim full suite green after filtering.
  • Existing browser harnesses seed their own data directories and terminate their host in finally. Preserve this cleanup; never point them at an operator database, shared SQL server or existing vault.
  • Record static export zip/version/source-tag/hash manifest and imported file checksums. Registry remains buildable on its public-feed graph.

Perspective

  • Run bash deploy/scripts/ign-deploy.sh validate --all projects/OEE_Site, focused new tests, python3 tests/test_cell_detail_view.py -v, and existing impacted downtime/manual-entry tests; then one appropriate broader offline run.
  • Use the repository deployment tools only for a separately authorized disposable validation gateway, retaining view/schema checks. This plan does not execute deployment. Designer resource acceptance, actual font availability, native input behavior, screen-reader output and panel interaction need live session evidence.
  • No oee.rebuild, CR history re-forward, DB/UDT schema change, heartbeat network call or real production job/downtime write is required. Preserve protected history periods verbatim.

Rollback

  1. Keep separate commits for router package/assets, router interactions, registry import/domain compatibility and native view mappings.
  2. Router rollback restores the prior coherent UI/package graph (Theme currently 0.5.0) and its corresponding JS/markup together; validate modal and role behavior.
  3. Registry rollback restores the prior app.css/font declarations/App.razor load order and removes only imported generated assets. It must not touch registry DB, revisions, vault or publication journals.
  4. Perspective rollback restores prior view resources and associated approved style mapping artifacts together through the existing deployment workflow. Never revert runtime facts, job IDs, reason classifications or tag providers to undo a visual change.
  5. Compare final deployed/source commit IDs and rerun the documented read-only smoke paths after any authorized rollout/rollback. Keep live limitations visible in the release record.

Done: all ten findings have evidence-backed implementation/validation paths; shared visual tokens remain canonical while the router, independent registry and operator runtime retain their proper behavior and platform boundaries.