9.7 KiB
Step-05B - Independent range status and decisive dashboard tests
STEP IDENTIFIER
Step-05B, focused follow-up to existing Step-05A R-03/testing requirements. Step-05/05A remain NOT ACCEPTED; Step-04 and earlier accepted work stay accepted. This is the sole active assignment. New display-unit work is NOT assigned here.
STEP TITLE
Preserve low/high status independently of Uncertain quality and complete the already-required boundary assertions and failure-safe test cleanup.
OBJECTIVE
Fix one remaining projection path and the existing executable-test/cleanup gaps. Do not redo the corrected panel/editor, session ownership, unit/origin validation, failure handling, numeric/bar rendering or native workflows.
Read STATE.md, Step-05, Step-05A, the
preserved report and WORKER-QUALITY-GATE.md.
Write only prompts/Step-05B-Report.md under prompts, then STOP for manual review.
ARCHITECTURAL CONTEXT
Keep all nine projects, one native Client, independent Components and normal
Application/Diagnostics ports. No new dependency, host, schema, physics/clock
change or resource/lifecycle framework. Tests remain under src/tests.
The owner's new imperial-default display toggle is separate future Step-06 scope. It must precede recording/replay, now Step-07, after dashboard acceptance. Do not implement it here or treat it as a Step-05A failure.
CURRENT IMPLEMENTATION STATE
Step-05A supplies the real shared editor/panel in both hosts, one feed per actual session with disposal, truthful failures/context, visible safe gauges and docs. Worker reports clean full Release build, 133 unit + 91 integration passes (224 total, zero failed/skipped), native static/dynamic workflows and final rebuilt-Client recheck. Preserve the existing 207-case baseline plus 17 new cases.
Source/test review found:
- R-01: the Uncertain branch of CreateGaugeValueLocked passes no range state. CreateValue therefore leaves RangeState null for an out-of-range Uncertain reading. UI correctly consumes range when supplied.
- R-02: the
definition test
executes unit mismatch/span overflow/count, not all C-04 vectors. The
geometry theory
combines invalid percent with Unknown/
--, masking the finite-number guard. The Stop-drain test releases its barrier/session only after assertions instead of in finally.
These are inspection findings, not orchestrator-executed reproductions. Reinspect Git and preserve existing worker changes/history before editing.
REQUIRED IMPLEMENTATION
Frozen corrective matrix
All rows enforce existing Step-05 C-01..C-05 / Step-05A R-03/R-04/testing rules. No new quality, threshold, freshness, unit or ownership behavior is assigned.
| ID | Input/precondition | Required result |
|---|---|---|
| C-01 | Uncertain finite -10 / 1100, range 0..1000 | Numeric stays -10/1100; quality remains visibly Uncertain; independent low/high status is present; bar geometry is 0/100% |
| C-02 | Uncertain in-range and exact bounds; existing Good/Stale/Stopped | No false low/high at/in bounds; original states, last-known context and existing projections remain correct |
| C-03 | Each assigned invalid definition vector, starting with valid config/running feed | Typed rejection with identical prior definitions/source/feed/snapshot; zero/eight remain valid and ninth rejects |
| C-04 | Good numeric presentation with only percent null/NaN/+infinity/-infinity; valid 85% control | Invalid geometry creates no meter/invalid CSS/false zero; valid control creates the actual visible accessible 85% bar |
| C-05 | An assertion or command fails while controlled Stop cancellation cleanup is held | Finally releases the barrier and drains/disposes the session; no abandoned owner or blocked cleanup; existing reject/retry behavior unchanged |
R-01 - Compute independent range for Uncertain values
Use the existing shared range/projection helper, without changing source numeric data, quality or unit. Supply the low/high status for finite Uncertain values just as the existing independent state contract requires.
Reproduce the missing RangeState with real-provider/rewrite transport tests before the fix. Verify projection and rendered text/accessibility for both numeric/bar composition. Preserve actual -10/1100 readings and clamped geometry; do not turn Uncertain into Good, invent thresholds or change other source states.
R-02 - Complete the assigned executable boundaries and safe cleanup
Extend the existing atomic-definition tests with independent cases for:
- NaN and both infinities in bounds, equal/reversed bounds and finite-span overflow.
- Precision below 0/above 3, undefined kind, blank/padded/duplicate identifiers, invalid label, unknown module/signal, exact-unit mismatch and more than eight.
- Accepted empty/eight definitions and unchanged prior source/feed/definitions after each rejected replacement. Keep module/signal comparison rules intact.
Isolate the component's finite-percent guard using Good state and a real numeric ValueText while changing only Percent. Keep separate Unknown/missing-value checks; they are not substitutes for the numeric-boundary proof. Assert actual role/width/accessible geometry, not merely the presence of "85".
Wrap the controlled cancellation-drain test's barrier release and session cleanup in finally so assertion failures cannot strand an owner. Audit the newly added dashboard tests for equivalent owned-resource cleanup; make only tightly coupled test changes, with no scheduler redesign or wall-time polling.
FILES OR PROJECTS EXPECTED TO CHANGE
Application dashboard projection, focused existing unit/integration tests and
directly related verification/docs text only where necessary. UI changes only
if a real producer/consumer assertion reveals a coupled defect.
Only prompts/Step-05B-Report.md under prompts; no assignment, roadmap/state,
shared gate, historical report or protected instruction edits.
INTERFACE AND DEPENDENCY REQUIREMENTS
Preserve schemas/protocols, exact source unit/binding, generation/lifetime and single-owner behavior, independent physics/model time, existing diagnostics/ verified clear/partial discovery, generic Components and all safety gates. No raw data mutation, engine oracle, fallback source/zero, package/system/workload action, staging/commit/push or destructive Git.
TESTING REQUIREMENTS
Run actual pre-fix C-01 regressions and report the assertion failure, then run corrected C-01..C-05 checks. Do not claim a compilation error as reproduction. Use real existing static/dynamic providers and controlled scheduler/barriers; do not mock away the feed, introduce sleeps or add a production fault selector.
Run focused cases, then sequentially:
dotnet build .\src\AutoForge.slnx --configuration Release --no-restoredotnet test .\src\AutoForge.slnx --configuration Release --no-build
Preserve all 224 current cases plus the added boundary assertions. Record exact per-project counts/exits/warnings and the full nine-project build. Restore/install nothing absent a real missing-dependency failure.
Preserve Step-05A's full native evidence. Recheck the rebuilt native static/dynamic dashboard baseline after the fix; controlled Uncertain/invalid-input tests prove those injected edge cases, not a human-timed native claim. If a required check is unavailable, record its exact blocker and INCOMPLETE/BLOCKED, not PASS.
DOCUMENTATION REQUIREMENTS
Keep quality/range independence and source-unit semantics accurate. Update the verification record and directly related docs if needed; do not rewrite previous reported results or claim the pending display-unit feature already exists.
ACCEPTANCE CRITERIA
| ID | Criterion | Evidence |
|---|---|---|
| AC-01 | C-01/C-02 independent range and quality are correct, without other state changes | Actual pre-fix failure plus real-feed/projection/rendered corrected assertions |
| AC-02 | C-03/C-04 decisive invalid-definition/geometry vectors execute and preserve valid state | Named independent boundary tests, not masked inputs or count-only proof |
| AC-03 | C-05 barrier/session cleanup survives failure paths | Finally-based cleanup/source-path self-review and existing controlled reject/retry proof |
| AC-04 | Current 224 regressions, new cases, complete Release build and native baseline pass | Exact worker execution evidence, native/control distinction |
| AC-05 | Original scope/history/protection preserved; matrix/self-review/report complete and worker stops | prompts/Step-05B-Report.md, no future feature implementation |
OUT-OF-SCOPE WORK
The new imperial/metric toggle/conversions/preferences, recording/replay, charts/ layouts/storage, UI/ownership redesign, clock/physics/protocol change, new host/ project/package, AI/hardware and unrelated cleanup. No future assignment/agent invocation/automatic handoff or orchestration/shared-gate edit.
REPORTING REQUIREMENTS
Create prompts/Step-05B-Report.md wholly inside exactly one plain triple-backtick
fence, no language tag/outside content/nested fences. Include identifier/title/
status, implementation/files/interfaces, R-01/R-02 and C-01..C-05/AC-01..AC-05
mapping, named actual pre/post results, exact commands/exits/warnings/test counts,
native versus controlled evidence, docs, self-review after the first green run,
limitations/blockers/deviations and preservation review.
COMPLETE requires all assigned checks; otherwise use INCOMPLETE/BLOCKED with the precise missing criterion. Save and STOP for manual review.