# OtOpcUa UI consistency implementation verification ## Scope and retained baseline Authorized checkout `/Users/dohertj2/Desktop/OtOpcUa`, branch `codex/scada-ui-consistency`, initial revision `a47ffb91` (stable bUnit migration). Impeccable Technical-Light direction, shared Theme gate and shared Secrets delivery precede this implementation. The Avalonia client is outside this web UI chapter. No production service, driver endpoint, SQL database, LDAP or fleet was connected or mutated; no deployment occurred. Baseline command (cwd is this checkout): ```sh dotnet test tests/Server/ZB.MOM.WW.OtOpcUa.AdminUI.Tests/ZB.MOM.WW.OtOpcUa.AdminUI.Tests.csproj -c Release --no-restore --filter Category=Unit ``` Before changes:1012/1012 Unit tests, zero skips/failures (`evidence/ot01-03/unit-before.log`, `ot-unit-before.trx`). After the first source batch, before the independent protocol-family test rollout began:1021/1021 (`unit-after.log`, `ot01-03-unit-after.trx`). Final owned field/modal/tree/Raw-row/delete cases:30/30 after peer lifecycle regressions (`final-generation-owned.log`). One broad filter accidentally also matched the other agent's newly red ProtocolFormAccessibilityTests; its31 expected migration failures remain in `final-first-batch-focused.log` and are not represented as an accepted aggregate. Final whole-project acceptance follows the completed OT13 rollout at OT14. ## OT01: delivered components and first field/filter associations Central pins adopt Theme1.0.1, Secrets.Ui1.0.1 and Secrets.Abstractions0.6.6. Core/umbrella Secrets and Akka replicator pins remain0.6.2; Auth.Abstractions stays0.4.1, bUnit stays2.7.2, AngleSharp stays1.5.2. Task-local package/HTTP caches restored approved feeds without a local package canary. `package-provenance.json` records the exact published nupkg hashes and official source; `runtime-provenance.json` independently matches all three test runtime DLLs against their package DLLs and finds no bUnit/AngleSharp/xUnit/NSubstitute/test SDK/InMemory provider in the AdminUI production graph. The disposable host's test-only InMemory constructor seam does not flow into that graph. Calculation timeout uses shared TechField Label/For/HintId with caller-owned input ID/description relationships. IDs are unique per component instance. Existing InputNumber parser messages are linked when an EditContext is already cascaded; the standalone modal form remains valid without one. No validator, numeric bound, default, serializer option or emitted protocol JSON changed. A native number input's incomplete-entry semantics are not claimed to be changed by this association work. Driver name uses the same shared hint/label recipe. Script log has visible Log level/Script ID labels and UNS has Filter loaded UNS names, preserving the original filters/callbacks and loaded-only scope. Toast dismiss actions name their own title and retain targeted dismissal/timers. The first test fixture hit the enforced ASP0006 rule by using a loop index as a render sequence; literal sequence plus SetKey corrected the fixture before the real red3 field failures. Four final field cases cover standalone rendering, visible label/hint relationships, repeated IDs and existing parser-feedback/JSON behavior. Actual browser label-click focused the standalone input, all three rendered timeout instances had distinct targets, and driver-name/filter labels focused their actual controls. A real effective195px negative check exposed Script log's fixed200px input/parent flex minimum, widening the document to220px. The first child-only bound was insufficient; `script-filter-overflow-cause.json` identifies the200px parent. Bounding that row and its child yields195/195 at real200% of390, while390/390 also fits. No live-tail behavior changed. ## OT02: named confirmation and native lifecycle pilot RawConfirmDialog now uses the local ModalFrame pilot. Title/body relationships are explicit, safe initial focus is Cancel, native Tab/Shift-Tab is bounded synchronously, background siblings are inert, and Escape returns through the existing cancel path. Busy confirmation disables close/cancel and rechecks server busy state; a queued second confirmation cannot start another delete. Confirm still awaits its existing caller then closes; exceptions retain the existing failure semantics. The exact existing delete-confirmation tests still prove no delete merely by opening/cancelling and real deletion only after confirmation. The supported policy is one active ModalFrame presentation. A second simultaneous enhanced overlay is rejected through its caller's cancel path; arbitrary nesting is not supported or claimed. Other legacy wrappers have not yet been migrated in this checkpoint. OT13 must sequence RawTree's intended Delete blocked/message follow-ups after confirmation closes so they cannot be dropped as an unsupported simultaneous presentation. It must retain each caller's busy/error/target-switch semantics. Native code receives the actual DOM event and does not use window.event. It restores previous inert/body state, verifies focus actually reached a connected trigger, and falls back to an available page heading/landmark for hidden/disabled/unfocusable/disconnected triggers. Temporary tabindex values are restored after blur; preexisting values remain exact. Native adapter cases cover safe focus, Tab containment, busy Escape, single-active rejection, disconnected removal cleanup, existing body/inert state, heading priority and exclusion of closing-dialog headings. Disposal releases listener/inert/reference state; no operation cancellation is implied by cleanup. A real implementation defect was caught: a comma-separated fallback selector selected `
` before its heading due to document order. The initial native JSON was2/7, even though its first script mistakenly exited0; the premature7/7 status message was explicitly corrected. `native-dialog-before.json/log` retain that result. The corrected adapter prioritizes outside-dialog headings explicitly; the extended suite is10/10, and scripts now return nonzero on false assertions. `actual-removed-trigger.json` separately verifies the real component returns focus to its page heading and serves JS byte-for-byte equal to reviewed source. ## OT03: contextual tree controls and preserved virtualization UNS expansion names/state follow the displayed `childrenShown` state, including a filter-open loaded path without mutating the stored expansion flag. Add/edit/delete/settings/open actions name their row and retain the exact node callback object. Raw expansion/paging controls name their owner; its ellipsis trigger receives an additive contextual name with a backward-compatible default. Picker checkboxes are labelled by the existing visible node-name span; the technical key remains in their title. Per-tree scope plus encoded node key avoids duplicate DOM targets. Virtualize, ItemSize30, Overscan20, lazy loading, row flattening, paging limits, selection and callback expressions remain unchanged. An initially added absolute visually-hidden context span did affect framework range measurement: a synthetic jump expanded estimated height to174272px instead of14914px when the diagnostic removed it. `raw-hidden-context-diagnostic.json` preserves the controlled diagnostic. The source correction reuses the existing visible span and adds attributes only, avoiding the extra per-row DOM node. No virtualization algorithm was replaced. Actual RawTree rendered800 already-loaded synthetic tags in the local fixture. At1440/390 it rendered55 initial and56 final checkboxes, retained selection across collapse/expand, expanded the real500-child Show more boundary to the final tag, and kept estimated height23717px instead of the inflated range. Names resolve to visible row IDs; document widths fit. The empty local InMemory factory exists only for child-constructor compatibility; browse/protocol sessions and mutations reject calls. A missing fixture DriverOperator policy initially blocked rendering; adding the source-matching anonymous-denial policy fixed the fixture without changing production authorization. Pointer confirmation uses native mouse coordinates on the already visible virtualized paging control rather than locator auto-scroll disturbing that range. ## Accepted browser scope and limitations `evidence/ot01-03/acceptance.json` combines8 accepted rendered interactions (first fields/modal/tree cases at1440/390/real200%, loaded-UNS filter, corrected Script log390/effective195, Raw1440/390),10 native-adapter cases and one real removed-trigger case. The first broad browser run stopped on an ambiguous close selector matching a toast and modal; scoping the selector and dismissing remaining toasts corrected the harness. The retained subsequent4/5 result's Script log overflow is explicitly superseded by its focused correction. Raw fixture/paging diagnostics and native failures remain separate; no all-first-pass-success claim is made. Mobile field/Raw screenshots were visually inspected. This is actual-component behavior over synthetic state, not every-route or deployed-host/authentication acceptance. App.razor retains static Routes, per-page interactive boundaries, SSR cookie authentication, antiforgery and the wrapper-owned Secrets route; only the local adapter script is added. `binding-preservation.json` confirms unchanged callback/binding/route attributes across affected surfaces; RawConfirm's header close is deliberately relocated into ModalFrame and remains mapped to its existing Cancel path, with busy guards covered by tests. The17 owned production/test files are hashed in `source-freeze.json`; the independent35 protocol-family files/tests are a separate scope. No SQL-integration/real driver acceptance is claimed by Unit results. OT04–12, remaining OT13 common controls/modal wrappers and OT14 route/consumer rollout gates remain pending this checkpoint. Production name/tombstone inventory for shared Secrets legacy names over450, legacy DIM exact-target restrictions and any deployment remain separate rollout requirements. No production inventory or deployment was performed. ## OT02 peer lifecycle/correct-target corrections Peer review reproduced a delayed native-activation reply arriving after hide/reopen or disposal. Red2 in red-modal-lifecycle-race.log is retained. ModalFrame now serializes native lifecycle work, captures presentation generations, rechecks current/disposed state after awaits, cleans stale accepted activations, and gives native updates/deactivation/close requests the matching generation. Busy-only renders update without reactivation/refocus. Optional nonconfidential Target identity supports editor reuse. Header handlers capture their rendered generation; stale Escape cannot cancel a reopened presentation. Seven focused frame/confirmation cases initially passed after this correction; final native10/10 and actual five-case field/modal confirmation remain green under the new ABI, with served-source byte parity/real removed-trigger return recorded again. Peer review also reproduced RawConfirm's old action completing after a replacement target and closing/resetting the replacement. Red2 in red-confirmation-target-race.log is retained. RawConfirm now stamps each render's confirm/cancel handlers with its presentation identity and captured callback; visibility/title/message/optionalTarget changes invalidate the prior presentation. An already-dispatched A operation continues, but cannot close B or clear B's busy flag. Shared-caption TargetA→B and stale rendered-footer regressions verify correct callback identity; neither action is replayed/cancelled. Internal close invalidates old handlers before a parent parameter pass. Disposal invalidates UI callbacks while allowing the dispatched operation to finish. Its first disposal fixture used cut.Dispose, which only detaches rendered tracking; switching to the existing DisposeComponentsAsync convention preserved the zero-stale-close assertion and verified actual component teardown. Raw failure logs remain; existing callback exception propagation is preserved. Final modal+existing-delete suite15/15, all owned cases30/30. This supersedes earlier21/14 checkpoints. The independent protocol-family rollout remains in progress, so the1021 Unit checkpoint is explicitly pre-parallel, not a claimed current aggregate. Existing root/common Monaco name work was accepted and committed separately as8cc053ce:19 targeted tests plus two real focused/editable non-ignored internal AX textbox cases, with unchanged model text/language/equipment context. Its preexisting late-create/disposal behavior is not newly certified beyond that bounded change. ## OT06 grammar checkpoint and navigation refinement Owned page anatomy is committed in ee8ac73e (first4),75ff3810 (secondary runtime8),78a39a10 (administration6 including additive Login title) and1fe4dbf5 (cluster9 plus tested overview destination policy gates). All27 owned page C# bodies/events/route attributes were preserved in those grammar changes. First4 rendered12cases; secondary runtime8 rendered24cases over synthetic state. Administrative/cluster populated/loading/error/whole-app/SQL coverage remains an explicit OT14 gate. Parent/peer owns Home/navigation/two retired Namespace compatibility routes and later ACL/list destination gates; those are separately evidenced and must not be counted as this agent's changes. The overview's Edit cluster/New node/per-node Edit/Deployments links now use the exact existing destination ConfigEditor policy. Real ASP.NET policy-engine4-role tests retain readable cluster/node content while denying authenticated readers/operators and permitting Designers/Administrators. Intermediate bUnit guard-registration/fixture diagnostics remain in ot05-cluster-actions; no endpoint policy changed. ## OT08 complete native diagnostic text checkpoint DiagnosticDetails provides native details/summary plus a named bounded focusable escaped text region. Fleet, Deployments, Script log, Alarms historian and Cluster audit expose complete existing metadata without hover/whole-artifact fetch. Existing previews/statuses remain; audit correlation IDs and UTC log timestamps retain complete context. DriverStatusPanel applies the same recipe to status-read failures, per-node errors and long operation results, retaining all action/controller code. Its restart confirmation is separate OT13 work. Focused helper/deployment10 and driver/by-node/wiring23 pass (three helper cases overlap); fresh combined owned lifecycle/labels/policy/diagnostic31 also pass against the current peer commits. Actual27 native scenes cover all five route compositions, standalone helper and driver read/last/failed-operation diagnostics at1440/390/real200%. Native Enter/Tab/ArrowDown expose/scroll complete escaped5000-character payloads with no document overflow. The first15 opening/full-text cases passed; stronger single PageDown probes timed out before a completed confirmation scene. Final natural keyboard navigation/repeated ArrowDown is recorded without a JS scroll setter; failed probe logs remain. A last text-align:start declaration is verified in the historian's right-aligned KV composition; prior24scenes already use left alignment. Fixtures use synthetic in-memory records/Administrator and a fabricated failed reconnect response. Historian uses a read-only local GetStatus fixture actor with no sink/queue/writer/fleet actors. This does not certify real authentication, SQL, fleet operations, deployment or all future callback/timer/exception states. Source/evidence scope is frozen in ot08-diagnostics; remaining operational-state tasks and final whole-app gates continue independently. Parent reviewed/accepted the OT08 source boundary and allowed its two ScriptLog OT07 toolbar/header class edits in the same commit; current source hashes include that markup. OT07 CSS/density acceptance remains separate.