cesnimdaP6-03 — Responsive, accessibility and localisation
Baseline: P6-00 register, actual Phase 5 final 312 axe scans / 39 all-enabled violation-node occurrences / 2,367 incomplete occurrences. Those are historical repeated nodes, not independent defects or compliance certification. Current source is the bounded P6-01/P6-02 correction set plus the P6-03 Career selector label repair; final integrated native evidence is recorded separately in P6-05.
Scope and preservation
Retain the established accessibility standard and manual gates, MUI keyboard/focus architecture, approved breakpoints/density and physical CV renderer. Review EN/Bokmål and light/dark at 320, 360, 390, 768, 1024, 1280, 1440 and 1920 CSS pixels, with intermediate breakpoint transitions and portrait/landscape. Existing tests cover 100/200/400% Chromium zoom and text-spacing overrides separately from normal viewport resizing. Two-dimensional Board/document canvas scrolling is considered separately from application-chrome overflow.
Semantic Login/Career/Account/CV/Admin heading fixes preserve their visual styles. Avatar initials use the existing contrasting theme pair. Settings desktop tabs now own named panel IDs; the translated name remains available when a phone uses the existing section selector. Rules drafts, inactive-panel unmounting, saves/discard and focus architecture are preserved. Company list/card dates now use the same UI locale as the existing relationship dialog; both EN and NB tests first reproduced host-locale dates. Jobs' two pending selected-company labels reuse jobTableCompany instead of hard-coded English. New language-save feedback is translated and bounded to the latest choice/request, without changing server write semantics.
Native evidence and triage
The retained first full attempt completed153/157 with four obsolete hidden-company-error assertions. Its49 detailed all-enabled scans and251 retained-rule scans are pre-Career-spacing-repair baseline, not the final integrated PASS. AXE_TRIAGE records11 repeated violation nodes and2,223 incomplete occurrences across300 scans, every rule category/context and fourteen actually inspected screenshots. Final full results are separate in P6-05; partial-scope evidence is not promoted to whole-suite acceptance.
P6-DC16 is a confirmed bounded mobile label defect. Native geometry RED in NB320/dark and EN390/light proves the floating label8px above the clipping container. The repair changes only the existing mobile FormControl from padding to margin with automatic inner width and removes the redundant label offset. The separate RED/intermediate/GREEN proof records 2/2 final focused cases passed, both labels at210–225.09375px inside the208–272px navigation surface, zero application-chrome overflow and keyboard selection of the unchanged /career Overview default. Its four hashed before/after images were actually inspected; the initial fourteen images remain attributed baseline. The intermediate post-CSS geometry pass followed by an incorrect test-only URL expectation is retained explicitly. This is a completed bounded Phase6 repair, not a full Career migration.
Fresh post-CSS checkpoint validation records 521 frontend tests /85 suites, application/browser TypeScript, lint and production build 2LDblja_XnSBWSuY6ys_V passing. The combined Settings/Career selector run passes 6/6 with no skips, retries or flaky results at2026-10-11 08:45:10.463 UTC: four named Settings variants plus the same two label cases. These six cases are a combined rerun, not eight distinct tests added to the full suite. Exact commands/report hashes are retained in the checkpoint validation and semantics result manifest. This checkpoint's bounded corrections are complete with manual/axe qualifications retained; the fresh159-case full integrated run remains separately pending in P6-05 until its actual outcome is verified.
The fresh ordinary browser run includes the existing thirteen-area audit, real zoom audit, keyboard dialog/command/back/focus/file-picker/CV preview/save workflows and four new named-Settings-panel/reflow cases (EN light1920, NB dark360, EN dark1280, NB light390). Company and language errors/retries use intercepted failures followed by real native persistence/readback. All-enabled axe results, incomplete rules and sample-specific targets are retained in evidence/axe-summary.json; representative screenshots have source test/state/viewport and hashes in evidence/visual-manifest.json.
Every incomplete category is retained for contextual review: contrast may involve disabled/overlaid/cross-frame content; prohibited/invalid ARIA attributes need actual MUI/hidden/settled contexts; hidden focus needs transition/open/closed focus behavior; target-size and table-cell relationships need actual active controls/data. No rule is removed or disabled to improve counts. Existing retained-rule scans are historical continuity evidence; current all-enabled scans provide broader triage. Remaining preview main/landmark findings span the parent and separately named iframe document; a best-practice cross-frame finding does not authorize changing A4 geometry. Settled drawer names/focus must be assessed from actual native AX snapshots, not inferred from text or an early transition sample.
Automated names, controls, dynamic status, input error and focus assertions do not replace spoken screen-reader evaluation. The M01–M13 operator packet remains NOT VERIFIED for actual reader/version/operator/device/spoken results. Physical touch/native dragging/software keyboard/mobile AT, text-only zoom/forced colours and supported alternate browsers retain their exact manual requirements. Production Linux font/photo qualification also remains open.
Relevant PRES: E02/E06/E07/E08, J05/J07/J08, W01/W08, C06–C12, S01/S03/S04. No group is promoted to full acceptance solely by finite browser or axe evidence. No route, canonical fact, CV override, frozen document, authorisation, pagination geometry or model qualification changes. Completion/push and conditional acceptance are in the chronological log and final report.