Files
lmxopcua/docs/ui-consistency-validation.md
T
Joseph Doherty 8ff424fc5a
v2-ci / unit-tests (push) Failing after 3h3m20s
Record coordinator acceptance of verified OT candidate
2026-11-14 02:26:31 -05:00

103 lines
17 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# OtOpcUa shared UI consistency validation
**Execution status: accepted implementation candidate — November14,2026.** The coordinator verified the current2674/2674 full-suite result, all1517 Unit identities, final browser/build/asset evidence,274source hashes and owned cleanup. Cross-family acceptance is recorded in scadaproj’s `components/ui-theme/validation/cross-project-final/G3.md`; this does not authorize deployment.
Evidence paths below are relative to `docs/ui-consistency/`; umbrella paths are relative to the managed scadaproj root.
## Scope and source boundaries
The reviewed candidate is the existing `codex/scada-ui-consistency` OtOpcUa checkout. It adopts the published family Theme and shared Secrets UI, preserving Technical-Light/IBM Plex, neutral paper/ink, blue accent, compact rail and common semantic tokens. The illustrative Impeccable mockups guided hierarchy and shared patterns; their fabricated metrics, operations and endpoints were not implemented as product requirements.
The authored Razor `@page` inventory has **31 routable components and 36 route templates**, rather than31 distinct route templates. Extra templates include new/edit aliases. The separate code-behind ContextMenu demo route is guarded by `#if DEBUG` and has no Release route; its reusable ContextMenu behavior is covered by the native tree/menu proof rather than claiming a Release dev-page test. Runtime component fixtures deliberately use synthetic local reads and reject unlisted mutations; direct rendering does not exercise the production router, cookie login, LDAP, fleet, PLC, COM or remote actor topology. Existing exact-policy metadata and real authorization-engine tests complement the bounded browser role cases. Login retains its static POST and shared LoginCard; SecretsAdmin retains the interactive wrapper and producer-owned SecretsPage.
The original source review, plans and initial preflight are historical evidence. Their scores and unchecked inventories are not claims of live acceptance. In particular, `modal-rollout-preflight.json` describes the old wrappers; final modal acceptance is the composition evidence below.
## Delivered dependency contract
| Package | Version | Accepted official nupkg SHA256 |
|---|---|---|
| ZB.MOM.WW.Theme | 1.0.1 | `995b117b0245577598c91404ed7957c071ebe3413696c8be48b419c9deb1bd5d` |
| ZB.MOM.WW.Secrets.Ui | 1.0.1 | `cd0efe497b2efa7acb61e634b727e802e883700d7c63681f895bcc4d0c6b816a` |
| ZB.MOM.WW.Secrets.Abstractions | 0.6.6 | `5cb0386469638bda89eb6c9afa25ead397b86730c3c5e43e4e496269d316acdc` |
Core/backends, umbrella Secrets, Akka and authentication dependency pins remain as approved. bUnit2.7.2/AngleSharp1.5.2 retain this project’s existing xUnitv3 framework/runner. The root read-only graph audit records exact runtime/test dependencies, package/runtime DLL parity and absence of test-only packages from production assets. Its initial snapshot is bounded to existing restored assets, not a claim of a new restore or all-route execution. The pre-final audit is retained at umbrella `components/ui-theme/validation/ot-ui/ot14-root-audit/{README.md,snapshot.json}` (main34fc3ce); final `snapshot-final.json` (main7013b6c) verifies all31 authored pages have a PageTitle, exact accepted dependencies/official package hashes/runtime DLLs and no production test-package leak.
## Finding-to-evidence coverage
| Finding/task | Implemented contract and retained evidence |
|---|---|
| OT01 fields and filters | Per-instance labels/help/errors in driver/device/tag/common families, additive named Monaco, guarded native filter inputs; `evidence/ot01-03`, common/protocol inventories and final cluster/security corrections |
| OT02 dialog lifecycle | One active overlay, serialized generation-owned activation/deactivation, busy close guards, safe focus return including removed/hidden/disabled triggers; `evidence/ot01-03`, `ot13-dialog-focus-recovery`, `ot14-modal-observation` |
| OT03 tree semantics | Stable native row IDs, contextual checkbox/expand/menu names and keyboard metadata without virtual-range distortion; `ot01-03`, root OT07 and common tree proof |
| OT04 home | Role-safe practical landing destinations; root navigation role/viewport evidence |
| OT05 destination policies | Exact existing destination policies gate rail/home/cluster actions, ACL/New cluster residual gates; `ot05-cluster-actions` plus root navigation proof. No new policy names or privilege grants |
| OT06 page grammar | Header/title/toolbar, semantic table scopes and named local scrolling; initial `ot06-*` plus final admin/cluster matrix. Direct-component scope and alias coverage remain explicit |
| OT07 responsive/touch | Real30/44px tree rows match ItemSize; local deep-tree pan, bounded DOM/800 selection/paging, coarse44px controls, ContextMenu keyboard focus across Chromium/Firefox/WebKit; root OT07 evidence |
| OT08 diagnostics | Full escaped native disclosures separate payload/correlation/UTC and live summary; `ot08-diagnostics`, actual component read-failure proof |
| OT09 deployment outcomes | Accepted request ID distinguished from recorded application state; last-good read retention/retry and single-flight; `ot09-deployments` (37 focused tests,24 bounded native cases and owned SQL50-row/blob-exclusion query proof) |
| OT10 tree recovery | Initial root read/retry single-flight, loaded-only filter/Clear no fetch and no forced expansion; local UNS pan; `ot10-trees` |
| OT11 live/read state | Fleet/Hosts retain last-good data with independent acquisition times; ScriptLog/Alerts bounded immutable pause snapshots; `ot11-refresh` and `evidence/ot11-tails` |
| OT12 status/icons | Canonical semantic chip inks/explicit states and decorative signal separation; `ot12-visual`, root stale-panel contrast correction (9 composed pairs fail→pass, minimum5.555:1) |
| OT13 families/callers | All protocol/common fields plus composed target-owned modal callbacks/read versions, post-write read-only recovery and named inline child editors; detailed acceptance inventory below |
| OT14 family release gate | Exact graphs/assets, final current full suite2674/2674, bounded browser/AX/keyboard/zoom/reduced-motion and final Impeccable audit/polish disposition |
## Field, modal and caller inventory
Driver/device/tag families use existing domain models, parsers, emitted JSON, validation and callback signatures. Default, repeated and conditional render cases, existing serialization/parity tests and native field focus/error checks are retained in their respective inventories. This is not inferred from label regexes alone. Common controls include identity/resilience fields, static/live pickers, contextual browse/ACL trees and named Monaco creation/update. DriverFormShell/TestConnect disposition and disposal entry guards retain domain behavior.
| Family | Evidence |
|---|---|
| RawName/DriverType/Stub, Area/Line | `evidence/ot13-simple-modals/verification.md`; parent review retained in umbrella `ot-ui/simple-wrapper-review.md` |
| DriverConfig/Device/RawTag/RawCsvExport/AclScopePicker | `evidence/ot13-config-modals/verification.md`; actual invalid-name/JSON focus recovery and retiredA probe snapshot/current-liveB cases |
| Raw manual and browse | Root Raw workflow evidence: immutable submitted rows, partial-success retention, current target close/read, probe snapshot isolation, native focus reanchor and typed-error contexts |
| Raw CSV import | `evidence/ot13-csv-import`; all-or-nothing/parser/group-resolution/InputFile/virtualized review preserved. Successfully committedA invokes capturedA saved callback; RawTree guards its inner read before installation |
| AddReference/ImportEquipment/ScriptedAlarm/VirtualTag | `evidence/ot13-uns-modals/verification.md`; captured callbacks, immutable rows/CSV/scripts and current-only outcomes |
| RawTree composed callers | `evidence/ot13-raw-callers`; captured operations/targets, queued follow-up after actual originating deactivation; per-node read ownership shared by initial lazy Toggle and post-write reload, without tying ordinary reads to dialog presentation |
| GlobalUns composed callers | `evidence/ot13-uns-callers`; version-owned last-good structure, origin-only close, explicit confirmed-write/read-failure separation and readonly Retry without replay |
| Equipment page | `evidence/ot13-equipment-caller/verification.md`;15 original payload fields/maps preserved, route/selected lazy tab ownership, captured child callbacks and saved-needs-refresh recovery |
| Scripts source disclosure | `evidence/ot13-scripts-dialog/verification.md`; read-only lazy source, full literal payload, current-generation close/read and long-title fit. This page is not the script editor |
| ScriptEdit | `evidence/ot13-script-edit`; field/Monaco/route/save/delete lifecycle. Two narrowly classified vendored word-highlighter cancellation records remain documented; no blanket error suppression or vendor patch |
| Hosts controls / DriverStatus restart | `ot13-host-controls` and `evidence/ot13-driver-operations`; existing policies/actor operations/read clocks retained; restart confirmation is inline within its parent frame |
| Alerts actions | `evidence/ot13-alert-operations/verification.md`;11 new cases plus tails22/22; stable focused condition under prepend, unique duration IDs, native invalid input, UTC timed request, single-flight and policy recheck;4 native contexts pass. Direct-CDP720/195 surface captures supersede Playwright zoom framing artifacts |
| CollectionEditor / DriverTagPicker | `evidence/ot13-collection-inline/README.md` (cb26ed917):9new/24aggregate managed,4composed scenes×14checks. `evidence/ot13-driver-tag-picker/README.md` (c8edd65f3):9new/861related managed,8composed scenes×20checks +4chrome contexts. Named inline groups, original parent draft/clone/address behavior, exact return focus and one parent overlay; only existing scrollable class added to each measured long parent |
The shared overlay has an explicit single-active policy; no arbitrary nested overlay support is claimed. Nested collection/address editors must stay inside that parent. Existing already-dispatched operations are not replayed/cancelled merely because a presentation retires. Completion callbacks and refreshes obey their documented originating-target contracts.
## Final checks and retained failures
- First combined Unit run:1493 passed,1 failed,0 skipped. The failure observed JS-only `OnAfterRender` work with a render-triggered wait. The test now awaits the actual replacement-generation activation event, preserving count/generation/deactivation assertions; root focused127-case proof is retained.
- First full run:2650 passed,1 failed,0 skipped (2651 total). The remaining ScriptLog paused-filter test read old DOM immediately after a dispatched filter event; its test-only correction awaits the actual render and retains frozen/filter/cap assertions. These raw first TRXs/logs are retained in `evidence/ot14-final`.
- Final admin/cluster first matrix:72 actual component cases,54 passed/18 failed. All failures were confined to5 surfaces: field names and195px text/action geometry in RoleGrants, Certificates, ClusterEdit, NodeEdit and AclEdit. This measured correction batch preserves original business bodies (three cluster `@code` bodies byte equal; security parity separately retained). Confirmation adds12 existing-node/ACL/retired-namespace alias cases; the original72 are never relabelled as84.
- Confirmation records82/84 passed; the two ACL new/edit195px failures were an additional checkbox-label/margin overflow. Bounded wrapper/label classes fixed only that geometry; `acl-narrow-confirmation.json` passes both actual195px/DPR2 cases. The original84case result remains82pass/2fail, with the targeted correction recorded separately. No second broad84case polish sweep was performed.
- Final title coverage correction: the earlier checker exempted Account/Secrets from its title assertion; `document-title-before.json` retains8 empty actual titles. Only product-qualified PageTitle metadata was added to the two host pages; `document-title-confirmation.json` passes8/8 at1440/390/720/195. No logic/policy/model change.
- Final Unit run: **1517 passed,0 failed/skipped**. Final full run on the current metadata candidate: **2674 passed,0 failed/skipped** (`full-final-current.trx`). `test-summary.json` verifies all1517 Unit testIDs also Passed in the current full run. Earlier before-title full run is retained separately. SQL-backed collections executed against the owned loopback fixture; this does not mean all2674 cases require SQL.
- Current Release build: **0 warnings/errors** (`release-final-build.log`). Nine actual served CSS/font/JS byte checks match the exact accepted Theme nupkg/final app sources (`served-assets.json`), including frozen dialog-focus7f63.
- Final authored detector: **zero findings**, with manual interpretation and bounded16/20 audit in `impeccable-final-review.md`. The two subsequent PageTitle-only lines were manually reviewed and targeted-browser verified; no extra broad detector/polish loop was run.
Every subsequent OT `dotnet test`, including filtered test discovery, uses the private owned SQL runner. Earlier filtered UI runs selected no SQL collections, but pre-existing SqlServerFact discovery could issue SELECT1 against the repository default before the override was identified. Those earlier runs are not claimed SQL-isolated. No earlier SQL collection writes were selected by those UI filters. The final disposable fixture is loopback-only, has no volumes and does not modify pre-existing containers; its random credential is kept only in a private temporary0600 file.
Reproduction (cwd OtOpcUa; private state is provisioned separately, never committed):
```sh
python3 infra/tools/ui-consistency/with_private_sql.py --state "$TASK_PRIVATE_SQL_STATE" -- dotnet test tests/Server/ZB.MOM.WW.OtOpcUa.AdminUI.Tests/ZB.MOM.WW.OtOpcUa.AdminUI.Tests.csproj -c Release --no-restore --filter Category=Unit
python3 infra/tools/ui-consistency/with_private_sql.py --state "$TASK_PRIVATE_SQL_STATE" -- dotnet test tests/Server/ZB.MOM.WW.OtOpcUa.AdminUI.Tests/ZB.MOM.WW.OtOpcUa.AdminUI.Tests.csproj -c Release --no-restore
dotnet build src/Server/ZB.MOM.WW.OtOpcUa.AdminUI/ZB.MOM.WW.OtOpcUa.AdminUI.csproj -c Release --no-restore
python3 infra/tools/ui-consistency/check_final_grammar.py
```
The browser script reads the owned canary URL from `/tmp/ot14-final-host.log`. Its source is retained in umbrella `components/ui-theme/validation/ot-ui/host`; no copied product UI or alternate package is substituted.
## Cleanup and retained provenance
`evidence/ot14-final/source-freeze.json` hashes274 final authored AdminUI source files (aggregate `76bbdb27d153fffa498921e74a15c8d616b40fb73ca269f54f21449e68558998`). It excludes generated outputs and unchanged vendor libraries; official package/runtime byte evidence is separate. Final source maps through the accepted commits and the final two-title metadata follow-up, not an inferred deployed version.
`credential-scan-final.json` checked1480 owned documentation/script/canary/evidence files and found no fixture credential. No private credential/state contents were retained. `cleanup.json` records both remaining owned canary listeners51661/64637 stopped, validated owned SQL container/task label/loopback binding/no volumes removed, and private credential directory/marker removed. All30 other running containers remained present during cleanup. Other agents’ owned fixtures have their own shutdown records; no pre-existing rig was stopped by this cleanup. Reproduction requires a newly provisioned private SQL fixture; the deleted credential cannot be reused.
Umbrella retained source/evidence is committed under `components/ui-theme/validation/ot-ui` (65fa913, with final SQL cleanup follow-up). The root package/route audit and historical checklists remain explicitly labelled by their own scope. The earlier `OtOpcUa-verification.md` now links here with its historical SQL-discovery qualification.
## Limits, rollout and rollback
Automated AX names/relationships, keyboard operations, actual zoom and reduced-motion are checked. **VoiceOver/speech-output testing was excluded by explicit user instruction**, and no speech-output conformance claim is made. Browser matrices are finite synthetic fixtures, not production fleet scale, route-cookie E2E or protocol connectivity certification. Native tree DOM/row measurements and tail bounds are real measured limits; no general throughput benchmark is inferred.
Deployment requires the normal consumer operational review, including production inventory of secret names/tombstones exceeding450characters and the documented older backend/DIM compatibility considerations. No production scan was performed. Published artifacts remain immutable. Rollback is scoped consumer commits/package pins together with their supported host adapters; no OtOpcUa consumer schema migration or new backend operation contract was introduced. Shared Secrets package compatibility changes remain separately documented. Preserve request/audit evidence and verify the reviewed older pins if reverting. No merge, push, publication of a new consumer artifact or deployment is implied here.