AutoForge/prompts/Step-06-Report.md

14 KiB

# Step-06 - Imperial-default display units

Status: COMPLETE on worker implementation and verification evidence; pending
manual orchestrator review. Step-07 was not started.

## Summary

Added one shared Imperial (US)/Metric and Road/Engineering selection, with
Imperial (US) / Road as the first-use default. The native Client persists this
small selection through MAUI Preferences. The pure conversion boundary projects
finite source numerics into supported display units without mutating source
signals, SI model state, configuration JSON/fingerprints, exact dashboard
bindings/bounds, source quality/range/time, or acquisition ownership.

The shared shell exposes accessible controls and truthful first-use/restored/
error state before a vehicle session opens. Diagnostics, simulation scenario
targets, simulation telemetry and live dashboard readings use the same
conversion logic. Dashboard edit bounds are displayed in a draft-pinned unit
and converted to source units only for changed inputs. Unknown units are
preserved with an explicit unconverted notice; conversion/storage failures are
visible and do not fabricate data or publish a failed selection.

## Implemented code and documentation

- `src/AutoForge.Application/Features/DisplayUnits/DisplayUnitSettings.cs`
  defines the immutable selection, exact mapping/inverse/bounds conversions,
  preference port/service, validation and visible storage-failure behavior.
- `src/AutoForge.Client/DisplayUnits/MauiDisplayUnitPreferenceStore.cs` adapts
  the application port to MAUI `Preferences`; `MauiProgram.cs` registers it
  before UI composition.
- `src/AutoForge.UI/Features/DisplayUnits/DisplayUnitPresentation.cs` converts
  JSON/dynamics numeric values from their raw numeric representation before
  formatting. `MainLayout.razor` owns the shared controls, status/error and
  cascading selection.
- `LiveDashboardPanel.razor` projects values/ranges without changing canonical
  definitions, preserves source reading state on projection failure, pins
  editor selection, converts edited bounds once, retains exact unedited bounds,
  and exposes explicit reset/cancel/error context.
- `DiagnosticsPage.razor`, `SimulationPage.razor` and
  `SimulationPageViewModel.cs` retain raw/source columns and add display-only
  projections for assigned human-readable measurements.
- `src/tests/AutoForge.UnitTests/DisplayUnitSettingsTests.cs` adds pure mapping,
  inverse, bounds, overflow, unknown/pass-through, rounding and preference-store
  cases. `src/tests/AutoForge.IntegrationTests/IsolatedDisplayUnitPreferenceStore.cs`
  supplies explicit portable test persistence. Integration composition uses
  that store rather than the unavailable UI fallback.
- Related composition tests and `DynamicsWorkflowTests.cs` cover real providers,
  shared rendering, paused-session toggles and source/wire invariants.
- Updated `README.md`, `docs/README.md`, `docs/architecture.md`,
  `docs/build-and-test.md`, `docs/live-dashboard.md` and `docs/simulation.md`.
  Step-05/05A/05B are identified as accepted; Step-06 is worker-complete and
  pending manual review. Step-07 remains blocked.

No project, package, host, project reference, protocol, schema, physics,
configuration fingerprint, history file or automation was added or changed.
No commit/stage/push was performed.

## Contract matrix and named evidence

| ID | Result | Executed evidence |
|---|---|---|
| C-01 | PASSED | `PreferenceService_UsesImperialRoadDefaultAndRestoresPersistedChoice` and native first-use shell: selectors were present before a session, the missing-key state identified Imperial (US) / Road as the first-use default, and no ambient culture/region was used. |
| C-02 | PASSED | `Project_KnownMappingsMatchIndependentVectors` and `ConvertToSource_RoundTripsSupportedVectorsWithinTolerance`: exact independent vectors and inverse conversions, absolute target tolerance 1e-9. Covered 20.1168 m/s -> 45 mph / 66 ft/s / 72.42048 km/h; 0.3048 m -> 1 ft; 0.45359237 kg -> 1 lb; (0.45359237/3600) kg/s -> 1 lb/h; 1.3558179483314004 N*m -> 1 lbf*ft; 0/100/-40 degC -> 32/212/-40 degF; and 0.3048 m/s^2 -> 1 ft/s^2. |
| C-03 | PASSED | `SpeedBoundsAndValueProjectionPreserveCanonicalGeometry` asserts all four raw projections and 0..40.2336 m/s bounds yield 50% geometry. Native real dynamic v2 sample was exactly 20.1168 m/s; observed display/range pairs were 45 mph / 0..90 mph, 66 ft/s / 0..132 ft/s, 72 km/h / 0..144.84096 km/h at the gauge's configured precision 0, and 20 m/s / 0..40.2336 m/s. All showed 50% meter geometry and unchanged source range/value. Exact unrounded 72.42048 km/h is independently asserted by the pure vector test. |
| C-04 | PASSED | `DisplayUnitTogglesOnPausedDynamicsFeedRerenderWithoutSourceOrModelMutation` performs 20 shared-preference toggles/rerenders on an actual paused dynamics session. It compares definitions, feed source context/values/states/range/time, immutable dynamics status, exported configuration, trace entries and byte-for-byte request/response payloads; actual `ReadSignals` trace count does not increase. |
| C-05 | PASSED | `Bounds_ConvertOffsetsAndRoundTripWithoutRoundingCanonicalValues` asserts 32..212 degF -> 0..100 degC, and `InvalidDefinitionReplacementIsAtomicAndEnforcesZeroToEightLimit` preserves feed state on rejected definitions. Native editor saved 0..90 mph as 0..40.2336 m/s; rejected 200..100 with visible alert and unchanged gauge; an unedited Save retained prior canonical bounds exactly. |
| C-06 | PASSED | Native editor opened on a real dynamic gauge with 10..110 mph typed, then global units changed to Metric/Road. Labels and values remained mph / 10..110 until save; saved source bounds were 4.470400000000001..49.174400000000006 m/s. The rendered draft-context message identified captured Imperial/Road units. Cancel and unedited Save were also exercised. |
| C-07 | PASSED | Existing real-feed coverage in `LiveDashboardAcquisitionTests` and `DynamicsWorkflowTests` verifies Pending/Good/Stale/Stopped, source quality/range/time and actual source context. Projection failures now keep `gauge.State` and separately show an alert/unavailable display; no conversion changes quality or clamps the value. |
| C-08 | PASSED | `Project_PreservesAssignedPassThroughUnits`, `Project_PassesThroughKnownUnitsAndExplicitlyLabelsUnknownUnits`, `Project_RejectsNonFiniteInputAndOverflowWithoutFabricatedValue` and `UnknownBoundsRemainSourceUnitsAndInvalidSelectionRejects` cover exact unchanged units, unsupported pass-through/notice, invalid enums, infinities, conversion overflow and finite ordered bounds. Native invalid edited bounds displayed an alert without changing the prior definition. |
| C-09 | PASSED | `PreferenceService_CorruptStoredChoiceShowsDefaultAndDoesNotOverwrite`, `PreferenceService_ReadFailureIsVisibleAndExplicitSelectionCanRecover`, and `PreferenceService_InvalidChoiceAndFailedWritePreserveActiveSelection` use controlled read/corrupt/write failures on the real service/port, including write atomicity. Native MAUI adapter started with no saved key; after explicit selection, closing/reopening the rebuilt Client restored saved Imperial (US) / Road. |
| C-10 | PASSED | The dynamics route test asserts displayed Imperial values and retained source `degC`, `kg/s`, `m/s`; configuration export is byte/string-identical across selection changes. The 20-toggle integration test compares copied raw trace request/response byte arrays. Native telemetry retained source 20.1168 m/s and 88.82446676508809 degC while showing display values; exported configuration remained 2,789 characters with SI target 20.1168 and no mph/km/h text. |
| C-11 | PASSED | `Project_ConvertsBeforeDisplayRounding` verifies raw 0.4 m/s -> approximately 0.894774516821761 mph within 1e-9 before precision-zero formatting to `1`, not `0`. |
| C-12 | PASSED | `StaticProviderCannotOpenDynamicsControls`, `StaticSession_FirstStartGroupsReadsAndSharesOneOwner`, `RealApplicationAndRenderedRoute_RunAndReadDynamicTelemetryWithoutOracle`, and the paused dynamics toggle test exercise real static v1/dynamic v2 composition. Native Client smoke configured a static rpm gauge and a dynamic speed bar, exercised existing routes, reset and close. |

## Acceptance criteria

| ID | Result | Evidence |
|---|---|---|
| AC-01 | PASSED | C-01/C-09 tests above plus real MAUI Preferences save/restart evidence. |
| AC-02 | PASSED | C-02/C-03/C-08/C-11 vector, inverse, precision, bounds, invalid and native rendered evidence above. |
| AC-03 | PASSED | C-04..C-07 real-feed source/state/wire invariants and native pinned editor, invalid save, unedited save and reset evidence above. |
| AC-04 | PASSED | Diagnostics, simulation scenario/telemetry, shared dashboard, raw/source inspectors, configuration JSON and native static/dynamic paths all use or preserve the assigned source/display boundary. |
| AC-05 | PASSED | Final nine-project Release build and all 270 preserved/added tests passed. Rebuilt native Client smoke covered static/dynamic gauges, four speed modes, draft/save/failure, restart, reset/close and legacy routes. |
| AC-06 | PASSED | Frozen matrix and exact named evidence are recorded here. Self-review and documentation changes are below. |
| AC-07 | PASSED | No protected/history/orchestration files, dependencies, stage/commit/push, agent or future-step work. Only this assignment report was added under `prompts`. |

## Commands and outcomes

Focused unit selection:

- `dotnet test .\src\tests\AutoForge.UnitTests\AutoForge.UnitTests.csproj --configuration Release --no-restore --filter FullyQualifiedName~DisplayUnitSettingsTests` — exit 0; 44 passed, 0 failed/skipped.

Focused real-provider/rendered integration selection:

- `dotnet test .\src\tests\AutoForge.IntegrationTests\AutoForge.IntegrationTests.csproj --configuration Release --no-restore --filter "FullyQualifiedName~DisplayUnitTogglesOnPausedDynamicsFeedRerenderWithoutSourceOrModelMutation|FullyQualifiedName~RealApplicationAndRenderedRoute_RunAndReadDynamicTelemetryWithoutOracle|FullyQualifiedName~StaticSession_FirstStartGroupsReadsAndSharesOneOwner"` — exit 0; 3 passed, 0 failed/skipped.
- After adding byte-for-byte trace-payload assertions, the focused `DisplayUnitTogglesOnPausedDynamicsFeedRerenderWithoutSourceOrModelMutation` selection — exit 0; 1 passed, 0 failed/skipped.

Required final sequential full checks:

- `dotnet build .\src\AutoForge.slnx --configuration Release --no-restore` — exit 0; all nine solution projects including the Windows Client; 0 warnings, 0 errors. MAUI XAML source-generation emitted its standard informational notice.
- `dotnet test .\src\AutoForge.slnx --configuration Release --no-build` — exit 0; AutoForge.UnitTests 178 passed; AutoForge.IntegrationTests 92 passed; 270 total, 0 failed, 0 skipped. The accepted Step-05B baseline was 134 unit + 91 integration = 225; all cases remain, with 44 new unit and one new integration case.

An intermediate attempt to run focused unit and integration test builds in
parallel contended on the shared Application output. That invocation was not
counted as passing evidence; the standalone focused integration rerun and the
required final sequential build/test commands above all completed successfully.

## Native Client smoke versus controlled evidence

Native evidence used the actual rebuilt executable at
`src\AutoForge.Client\bin\Release\net10.0-windows10.0.26100.0\win-x64\AutoForge.Client.exe`,
task-local WebView2 debug port 9361 and an isolated temporary WebView profile.
The initial app showed the explicit first-use/default status before any session.
The static synthetic session rendered 850 rpm unchanged through preference
changes. The dynamic synthetic session was paused at step 203 / 20.3 s; protocol
v2 telemetry showed source 20.1168 m/s beside Imperial display, and source
temperature/fuel/torque units beside their converted display values. The native
dashboard/editor results and route checks are mapped in C-03..C-06 and C-10/C-12.

Native persistence was checked by closing process 44360 and reopening process
48432 against the same temporary WebView profile. The stored Imperial/Road
selection appeared as saved after restart. The exact temporary WebView profile,
the only task-created profile folder, was removed. Because the preference key
was absent on first launch, the smoke's explicit save created the app's
`imperial-us|road` value; that sensible task-owned preference value is left in
place. No existing preference or unrelated user data was wiped or replaced.
Both native processes and listener port 9361 were closed.

Storage read/write failure, corrupt values, undefined enum values and conversion
overflow are controlled tests using the real preference/conversion services
and explicit isolated/faulting stores. They are not represented as native
platform-failure observations. No platform storage fault injection was added to
the Client. Native gauge precision was 0 for the exact speed demonstration, so
the UI showed the rounded 72 km/h; the raw 72.42048 km/h mapping and tolerance
are independently asserted before display rounding in unit tests.

## Source-path self-review and preservation

After the first green focused run, I walked the store->service->shell/cascade,
raw source->pure conversion->each readout, and gauge edit->validate->replace
paths. This review found that an unavailable display projection could replace
the underlying dashboard source state with Error. The panel was corrected to
retain the original source state and show a distinct display-unavailable alert;
the conversion error remains explicit, the value is `--`, and geometry is not
fabricated. Repeated focused and full checks then passed. The review also
distinguished first-use default from a persisted default so native smoke can
truthfully report restoration.

The native adapter is registered before `AddAutoForgeUi`; the UI fallback is
explicitly unavailable and reports storage failure, not memory-backed success.
Selection enum validation occurs before writes; writes precede publication.
Mappings use exact unit strings and raw numeric values. Source unit binding,
quality/range, diagnostic/model time, configuration identity/fingerprint,
exported SI JSON, raw trace bytes, one feed owner and paused model state remain
unchanged. Unedited gauge bounds use the original canonical values; conversion
and ordered finite-span validation happen before definition replacement.

`git diff --check` passed. No dependency manifest, package, project, Core,
Diagnostics, Simulation, Components, `prompts/STATE.md`, `prompts/ROADMAP.md`,
historical report or protected instruction was changed. The repository began
clean; all remaining modifications are Step-06 implementation/tests/docs and
this required report. No stage, commit, push, restore/install or Step-07 work
was performed.