Files
scadaproj/ZB.MOM.WW.Overview/docs/ui-consistency-validation.md
T

5.6 KiB
Raw Blame History

Overview Technical-Light validation

Actual source implementation and loopback synthetic-component verification, 2026-11-13. This is application validation, not a production deployment or fleet health report.

Shared contract and implementation

The application adopts exact Theme 1.0.1, with the approved purple accent #6d28d9, deep/focus #4c1d95, tint #f2ecfc, border #d7c5f2. IBM Plex, paper/card/rule geometry, semantic inks and shared focus/motion remain canonical. No local shared card-grid or KPI-grid copy was introduced. The actual package and release evidence live in ../../components/ui-theme/validation/theme-1.0.1/.

Task Implemented behavior Evidence
OV-01 Full-contrast stale observations and readable local semantic ink G1 pilot plus final screenshots
OV-02 Explicit check-state text; decorative dots; unknown remains neutral Six render cases, unknown-summary follow-up
OV-03 Pause explicitly freezes view updates while background polling continues; pressed state and user-only outcome Render tests, queued-publisher races, browser pause/age/resume
OV-04 Stored unreachable and stale/pending classification remain distinct; last-known rollups qualified Seven classification regressions; mixed-state browser scene
OV-05 Native full escaped probe disclosure, keyboard-accessible raw cluster region, narrow check layout Four render regressions and long-error browser scene
OV-06 Default-off attention filter preserves fleet totals/group identity/leader disagreement; pending separately counted; real sweep UTC Eight tests including deterministic dispatcher races; browser filter/clear/pending/split scenes

Observed verification

The full application suite passes 224/224, zero failures/skips, compared with 197 after GB. Existing identities/assertions remain; 27 regressions exercise new behavior. Exact-final-package test/build output is in overview-final.log and TRX under the Theme 1.0.1 evidence directory. SDK 10.0.403, net10.0, bUnit 2.7.2; no test framework/parser migration was mixed into this phase.

The fixture references the actual application assembly, routes, layout and widgets. It registers synthetic immutable snapshots and a frozen/advanceable clock; no fleet poller, real HTTP probes, authentication, secrets or mutation service is registered. Its readiness marker belongs only to the fixture. Referencing an executable Blazor app duplicated identical SDK framework assets: the fixture deduplicates input identities before compression. Production build configuration remains unchanged.

Evidence: ../../components/ui-theme/validation/overview-ui/.

  • First batched measurements exposed shared KPI overflow at effective195 px and local long leader-chip wrapping. Both were corrected in their proper owners. The first harness also selected a hidden navigation summary and asserted an asynchronous pressed state without waiting; these harness failures are retained, not reported as product defects.
  • One corrective matrix contains 30 state/width combinations: mixed seven states, long diagnostic, split leaders, pending, empty, and500 instances, each at 1440, 1024, 390, 320, 195 px; one additional real 200% browser-zoom case gives720 CSS px and DPR 2 from a 1440 px viewport. No document overflow or browser page errors.
  • Filter shows 5 of 7 attention instances while fleet totals remain 7; Clear restores all cards and checkbox focus. Paused cards continue aging into stale; new snapshot publication stays hidden until Resume adopts the latest store snapshot.
  • Enter opens the complete 2000+ character escaped diagnostic. Reduced motion yields opacity 1 and no rise animation. Cluster evidence intentionally scrolls inside a named focusable region; raw data remains complete.
  • 500-instance navigation/render/readiness measurements were 1005–1191 ms on this machine, including an explicit 700 ms font/animation settlement. This is a fixture observation, not a render-only benchmark or performance budget. No evidence justified introducing virtualization.

Impeccable audit and polish

The bounded desktop/mobile/zoom inspection preserves the existing Technical-Light identity and domain hierarchy. Status stripes supplement explicit labels; values, last-known classifications and raw diagnostics remain truthful. Full contrast is retained for stale cards. Font/rule/spacing and purple navigation align with shared chrome. The polish batch corrected narrow reflow at the shared owner and wrapped Overview's leader/address and very narrow card header/footer anatomy locally. No alternate card system, theme fork or ornamental content was introduced.

The active Impeccable UI edit hook reported no deterministic issue for the shared CSS edits; this is not an accessibility certification. Keyboard, accessible roles, names/states, live-region source semantics and browser accessibility-tree evidence are distinct from a human screen-reader listening session. Human VoiceOver/NVDA, additional browser engines and real production latency remain rollout validation limits. Only the supported light Overview surface is claimed; ScadaBridge's separate dark adapter is not an Overview theme.

Rollout and rollback

No application deployment has occurred. Package 1.0.1 passed exact-byte publication and independent-feed restore; see its acceptance record. Keep its graph and the previous application artifact together. Roll back application artifact/pin together if a production check fails. Registry, polling, readiness/activity reports, wire contracts and database schemas are unchanged. No login or write controls were added.