jobtrackingapp/docs/pre-phase10-ux/D_ADDENDUM_ACCEPTANCE.md

# D owner-addendum qualification

29 October 2026. This adds evidence to the preserved initial D acceptance; final cumulative acceptance remains pending.

## Completed requirements

The administrator list now filters before paging by UTC creation interval (inclusive start, exclusive end), actual owner-verified job, configured processing family, fixed failure code and bounded search. Search covers safe account identity, owned job title/company and exact operation GUID only. Invalid identifiers, intervals, worker names and failure codes return 400. The UI explains invalid filters and stops automatic retries until a correction. Dates earlier than MariaDB's supported range are rejected. Existing owner/status/task filters and bounded paging remain intact.

Details expose the internal owner identifier, actual extraction-run identifier and explicit absence of a linked job. Job navigation is offered only when the authenticated administrator also owns that job; administrative metadata access grants no new ordinary-user job-reading permission. The list displays elapsed duration and current failure state. A named account-partitioned read limit permits 120 requests/minute without queueing; normal polling uses approximately 36/minute. Existing authorization remains mandatory.

Failure metadata uses the latest retained, owner-bound typed failure event with a sanitized matching code. Successful/retrying operations label retained failures as previous failures. Missing historical timestamps stay unknown; a later success timestamp is never presented as the failure time. Only an actually failed terminal operation may fall back to its recorded completion time. No arbitrary messages, payloads or source text enter this projection.

## Verification

Evidence paths are relative to `.artifacts/pre-phase10-ux/`. Focused reruns are not added to full-suite totals.

| Check | Result | Evidence |
| --- | --- | --- |
| Full backend | 1,447 passed, zero failed/skipped | `d-addendum-backend-final/backend.trx` |
| Full frontend | 1,031 passed across 112 suites, zero failed/pending | `d-addendum-frontend-full-02.json` |
| TypeScript, full lint, production frontend build | exit 0 each | `d-addendum-{typecheck,lint,build}-final.log` |
| Admin boundaries, filters, recovery, responsive EN/NB/themes | two passed | `d-addendum-browser-02.json` |
| Real browser zoom 100/200/400%, EN/light and NB/dark | two passed | same browser report; original geometry assertions retained |
| Real MariaDB, including final failure-event projection | 90 passed, zero failed/skipped; cleanup exit 0 | `d-addendum-mariadb-final-02/` |
| Follow-up static review | timestamp finding resolved; no further blocker | existing server review worker, 29 October |

The initial RED API/UI cases remain in `d-addendum-api-red/` and `d-addendum-ui-red.json`. The timestamp regression was independently reproduced with the former completion-time binding (`d-failure-time-red.json`) before restoring the corrected binding. The first provider attempt found an actual collation mismatch in the new cross-table ownership filters. Explicit provider-appropriate binary collation fixed those comparisons without rewriting data or schemas; `d-addendum-mariadb-02/` passed 90 tests before the final failure-time query was added. A subsequent final attempt could not start because Docker Desktop was stopped (`d-addendum-mariadb-final/`); it did not reach tests or MariaDB. This infrastructure failure is not a test pass.

The initial zoom recovery run passed both admin cases but was stopped while waiting for extension startup; retain `d-addendum-browser-final.log` as an incomplete attempt. Isolated diagnostic launches ruled out the long worktree path. The test context now enables its deliberate local extension and captures its service worker before navigation; the fresh four-case report passes. Application code, zoom percentages, timeout and geometry assertions were not relaxed.

## Scope and gates

Later integration replay found a remaining zoom-extension lifecycle race after the first four-case pass. The original `integration-03/` attempt was stopped and remains incomplete; its runtime hashes stayed unchanged and its test-source change is explicit. A direct trace records the worker closing before evaluation. The final test harness verifies the generated worker has executed before installing its network guard, permits only that exact local worker URL in addition to its original local routes, and uses short owned temporary paths with cleanup. `zoom-lifecycle-04.json` records all six repeated EN/light and NB/dark real-zoom cases passing with retries disabled. The interrupted `zoom-lifecycle-03` attempt remains partial. No application or geometry assertion changed.

The sealed security scan through `d52403d` does not include this later delta. Its bounded filters, rate policy, failure-event query and UI were reviewed separately; no new security blocker was identified. This is source review, not a new sealed scan or production acceptance.

Full cumulative ordinary/specialized browser, final Docling, PDF fidelity, source/runtime closure and local C/D activation remain pending. Historical PDF coordinate-oracle failures, human/device/language/print checks, real AI extraction qualification and production infrastructure debt remain open. No production operation, main merge, deployment or Phase 10 cleanup occurred.