jobtrackingapp/docs/redesign/audit/TEST_INVENTORY.md

Test baseline and gaps

The test-file census lists every discovered tracked frontend test file (76: 67 unit/component and nine browser/HTTP specification files), 106 backend test-source files, plus extension/parser/repository safety files. Counts describe files, not executed cases or coverage percentages. The source snapshot is 97fce89; subsequent bounded CI packaging corrections are tracked in the B06 release log, not redesign work.

Test layers

Layer	Setup and demonstrated scope	Limit
Frontend unit/component	Jest, jsdom, Testing Library; config. Account switching, routes, workspace/drafts, CV, mail, admin, theme, translations, dialogs and imports/exports	Babel strips TypeScript; mocks do not prove server ownership or real provider behavior
General browser	Playwright config: one Chromium project, isolated .NET API/SQLite data and local frontend; production static build supported	Chromium only; no actual Traefik/nginx; no cross-browser or screen-reader certification
Browser smoke	Nine cases in e2e/smoke.spec.ts: health/readiness, public plans, login, Free restrictions, reviewed job create, deep-linked job workspace/dirty guard, career, public CV/PDF, long searchable Code PDF	Selected journeys, not exhaustive editor states or all providers
Account isolation	Four e2e/account-isolation.spec.ts cases: original A→B replay, missing/forged context, crafted CV ownership/association, two-tab private-state disposal/Back	Retain API and database assertions when UI changes
Accessibility	Two axe cases over eight sampled routes	See missing state matrix below
Other general browser	One each discovery responsive review, dark Kanban, CSV security/UTF-8 byte preservation	Does not establish whole-app mobile/theme parity
PDF HTTP probe	pdf-tests/admission.spec.ts and dedicated config	Synthetic bounded concurrent requests; admission differs from time-window limiting
Rate-limit HTTP probe	rate-limit-tests/isolation.spec.ts and dedicated config	Nine account/session/IP/auth/forwarding scenarios; native API topology
Production ingress HTTP	ingress-tests/production.spec.ts and dedicated config	Six native production-mode origin/proxy checks; not running nginx or Traefik
Backend	xUnit/Moq, SQLite/InMemory and HTTP test seams; API/service ownership, auth, parser, CV/history/PDF, queue, billing, export, migration and rate/proxy tests	Provider-host tests and fault/probe tests have separate runtime requirements; SQLite is not full MariaDB proof
Repository/dependencies	Compose rendering + 17 hostile mutations, secret scan/SBOM, evaluation and supply-chain scripts	Config validation cannot prove host firewall or actual socket peers
Adjacent clients/services	Browser-extension node tests and Python parser/service tests	Preserve capture/document boundary; avoid pretending these are covered by React Jest

The default browser suite currently contains 18 tests. The nine special browser/HTTP files count includes six general e2e files and three separately configured suites. Special configurations must be invoked explicitly; a plain general Playwright run does not run all of them.

Existing evidence versus fresh execution

Evidence	Result and classification
Previous B06 unchanged application regression	166 focused backend, 3 Compose, 6 production HTTP, 9 rate-limit HTTP passed; exact prior commands/results
Previous B06 broader baseline	915 selected backend tests, 298 frontend tests/67 suites, 18 browser tests, Release builds and targeted lint passed; external-DNS-dependent IMAP test excluded locally and explicitly documented
Fresh Phase 1 type reproduction	From job-tracker-ui: node node_modules/typescript/bin/tsc --noEmit --incremental false → exit 1, src/interview-prep.test.tsx(120,10): error TS2571: Object is of type 'unknown'.
Fresh release-prerequisite checks	python scripts/test-ingress.py → 3/3 pass; real standalone Compose path also tested; official downloaded Linux binary digest verified; tracked-secret scan and diff check pass
Fresh packaging regression at 23cc250	Explicit installed .NET 9.0.317 SDK: Release build + 22/22 PDF admission tests pass, then 915/915 selected backend tests pass (same documented external-DNS IMAP exclusion). Results: .artifacts/redesign-audit/backend.trx
Actual CI push verification	Runs 736/737 exposed missing Compose/token prerequisites; corrected. Run 738 passed Compose/safety/extension/parser/API build, then exposed Linux PDF test-fixture packaging collision. Follow-up result belongs in release log
Fresh Linux backend confirmation	Run 741, 2024c37, passed 916/916 backend cases, zero skipped, after test packaging/runtime-location and TLS EOF-or-no-read portability corrections. This includes the locally excluded DNS-dependent case. Application behavior and all security assertions remain protected; deployment has separate gates
Live/container boundary	Local Docker engine unavailable; F04 production evidence remains incomplete. No audit claim of real proxy-chain success

The selected broader backend suite was repeated for the PDF fixture packaging change; the frontend suite was not repeated merely for documentation edits. Preserve the distinction between prior unchanged-source evidence, newly executed checks and actual CI failures. Run 739 subsequently failed earlier in the unchanged API compiler with csc.dll exit 139 on both existing attempts; it did not reach the corrected test-project build. No test gate was bypassed.

Exact TypeScript root cause

Interview test:120 reads .content from mockedApi.post.mock.calls[0][1]. Axios's mocked generic request-data argument is typed unknown; the test dereferences it without narrowing. The runtime assertion can pass because Babel removes types. The production component's emitted behavior is not shown broken by this test diagnostic.

tsconfig includes src, including tests. CI runs Jest/lint/Next build but has no explicit full standalone tsc --noEmit gate. Next's successful build is not evidence that this exact standalone test-inclusive program passed. Phase 2 should narrow the test payload appropriately and introduce a deliberate full type gate. Do not weaken strictness, blanket-exclude tests or call this a P1 production outage. No type fix was made in Phase 1.

Missing regression proof before redesign

Priority	Add or preserve evidence
First	Stable type-check gate; B02 stale/queued mutation + cross-tab unmount as shared-client acceptance; legacy URL/overlay/full-page parity
First	Populated CV editor/profile/job detail plus open menus/dialogs/error/conflict states in EN/NB, light/dark, mobile/desktop; preserve underlying draft/history assertions
First	AI suggestions rejected/accepted distinctly; actual profile values unchanged when editing a CV/cover letter; unsupported/locked capabilities behave honestly
First	Typed date values, JSON import/export and B05 CSV bytes; attachments/private/public download distinctions; package history remains frozen
Before mail migration	Draft revisions, uncertain send/new attempt, provider reconnect/read failure, ownership and explicit send confirmation; use synthetic/sandbox providers
Before admin migration	Non-admin denied APIs, role revocation and last-admin guards, sensitive diagnostics excluded from personal pages
Before public release	Keyboard + NVDA/Narrator, preview scroll/focus, zoom/reflow, forced colors, real mobile keyboard and touch; provider/production gates separately

Axe currently scans /, /login, /dashboard, /jobs, /kanban, /career, /settings, /admin/system, mostly after body/main appears, in English. Missing builder library/editor, job detail, discovery, offers, mail/review, public CV, account security, registration/reset/verification, operations, admin users/audit and interactive loaded states must be selected intentionally. Do not convert a green shell sample into a WCAG compliance claim.

Recommended later commands: locked dependency restore; Release backend build/test; full Jest; standalone tsc; lint; production frontend build; general Playwright static-export suite; explicit PDF/rate/ingress configurations; Compose validator and git diff --check. Use the environment/setup commands from the completion reports rather than guessing SDK paths or running probes against production.