Description
During TSO-0172 (live dashboard audit), an independently inspected published Devices by OS panel at 800x600 rendered a populated pie with a readable right-side legend, but long in-chart labels overlapped and one clipped at the left boundary; small-slice percentage annotations were partly obscured. This is a layout readability issue, not a query error. No lab labels or values are needed to reproduce it. Filed during loop28; not admitted to its frozen envelope.
Acceptance Criteria
- #1 A synthetic multi-category fixture with long OS labels and small slices keeps every category readable at 800x600 and normal dashboard width.
- #2 The visualization makes the percentage-versus-count interpretation explicit without clipped or overlapping annotations.
- #3 Retain the existing query and signal coverage and capture before/after rendered fixture evidence.
Definition of Done
- #1 just check passes (the full gate; it is what CI enforces)
- #2 just gen leaves no diff (only if a generated artifact’s inputs changed)
- #3 just –fmt –check passes and every new recipe has a # doc comment and a [group(…)]
Implementation Plan
Use long-label multi-category small-slice synthetic fixture to capture baseline real Grafana pie at 800x600 and normal width; adjust native legend/labels to readable explicit percentage-vs-count semantics preserving query/coverage; capture after renders, regenerate/stage, serialized G, guarded review/CodeRabbit then root composed/CI/delivery.
loop31: implement only the frozen owner-approved seam in isolated ownership; retain exact candidate and behavior evidence, named gate where applicable; independent guarded review then root integration/CI/delivery. Read-only diagnosis has no remediation or code gate.
Implementation Notes
loop29: 0 implementation attempts; undispatched dependency park. TSO-0172 (comprehensive live dashboard audit) cannot start before 2026-10-11T21:05:15Z under its required minimum fresh window and has not landed. Resume only after that audit is accepted, landed and composed/CI/delivery green, then preceding frozen chain tasks settle. No candidate work or live call occurred.
loop30: 2 implementation attempts, no rescues or reviews; candidate uncommitted/unlanded. Native bottom Value/Percent legend, no in-slice labels, height adjustment and real query-backed 800x600/native-grid closed-navigation render proof succeed, but full G exits1 on test_specialist_panels.FleetAndRelationshipPanelContracts.test_fleet_composition_panels_are_native_pies exact default pie options and test_unit_consistency.DecimalsAndLegendConventionTest.test_no_panel_sets_an_explicit_decimals_override. Frozen shared-test edit fence blocks intended OS-specific contract changes; no test/helper bypass or half-land. Park needs owner approval of narrowly revised OS option/precision contracts, then width/reflow proof with expanded navigation, full gate and fresh reviews. Two attempts remain under current ceiling. Full G output tail only retained; complete narrow failure log and candidate/archive/final-03 render receipts remain /Users/rob/repos/tailscale2otel/codex/loop30/tso0189-retry/. No live mutation. Main remains green; parked candidate G red is not main red.
loop31 preparation, Rob 2026-10-13: recommission approved. Previous blocker was the frozen OS pie options/precision test fence. Changed premise: OS-only explicit bottom Value/Percent legend, no slice annotations and precision=1 expectations are now authorized; Compliance distribution and all other precision guards remain unchanged. Discriminating check: retained query-backed fixture at 800x600 and actual native grid with navigation expanded and closed, followed by full gate and fresh review. Lifetime implementation attempts remain 2 of 4; remaining ladder is rescue-astra-medium then rescue-astra, no reset or ceiling increase. Task remains Parked until launch admission.
loop31: rescue attempt3 reaches exact query-backed expanded-navigation width failure, despite green candidate gate. Native grid OS width falls 624px to504px, legend339px cannot fit374-454px text; all seven labels clip. Closed grid and800x600 solo pass. Actual Prometheus/Grafana response receipts distinguish scope/state categories. Final rescue attempt4 dispatched for evidence-based width/reflow, no test weakening; solo has no visible application nav, native-grid must prove both nav states.
loop31: lifetime4 implementation attempts, final rescue succeeds. Actual executed scoped Prometheus query and real Grafana responses prove long-label small-slice before/after renders,800x600 nav-free solo and native grid visibly expanded/closed navigation; full-width native pie bottom Value/Percent legend, no annotations, precision1. Query and non-OS contracts unchanged. Full-hunk security review PASS; sharded CodeRabbit Python defect explicitly unreviewed, not clean. Final published batch9927d3962b49aef5d13f3196c76f05cf722872b8 contains source963635cd9bed31e888c229db44a797f373348742; exact fresh composed gate and CI38383777876 ci-success green, sync38383777852 success; exact far-side tailnet blob48ae2e76faadff3dafdc3cada0c2f4f17e0695fa matches. Helm absent by paths. Full gate stdout unavailable beyond tail; bounded fixture/viewport proof, not arbitrary labels/cardinality. Original retained candidate untouched.
Final Summary
Readable OS count/percent presentation passes actual query/render proof with navigation expanded and closed; non-OS defaults and every scoped query retained. Attempt4 of4; integrated/published and independently reviewed.