Commit Graph

6 Commits

Author SHA1 Message Date
cesnimda 6bfdbd9db7 perf(companies): integrate verified owner cache invalidation 2026-11-10 17:44:51 +01:00
cesnimda bca750446f perf(applications): bound correspondence reads to the current page 2026-11-10 17:42:08 +01:00
cesnimda b176a44627 docs: reorganize tree, restore architecture + research from archive, add Phase 0 reports
Active docs/ was stub scaffolding while the real docs sat in docs/_archive/.
Restore and correct them, and record the Phase 0 work.

- docs/architecture/current.md: verified system map (from archived SYSTEM_OVERVIEW,
  9 corrections against code).
- docs/research/competitors.md: sourced competitor analysis (from archived
  PRODUCT_RESEARCH, feature matrix corrected).
- docs/decisions/ADR-002-job-application-model.md: the Job/JobApplication split.
- docs/application-discovery-report.md, docs/implementation-roadmap.md,
  docs/phase-0-foundation-report.md, docs/career-workspace-branch-assessment.md.
- Remove 10 zero-byte placeholder files that advertised content that never existed.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-17 17:04:32 +02:00
cesnimda b8f8569e6e fix(hooks): stop infinite render loop in useViewResource
CI and Deploy / test (pull_request) Successful in 2m0s
CI and Deploy / deploy (pull_request) Has been skipped
useViewResource built `reload` with `load` in its useCallback deps, and the
fetch effect depended on `reload`. Callers routinely pass an inline `load`
closure (e.g. JobTable), so `load` — and therefore `reload` and the effect —
changed every render, calling setState and re-rendering: an unbounded
"Maximum update depth exceeded" loop that froze the renderer on /jobs and every
other list view (DashboardView, RemindersView, CompaniesTable).

Fix: hold `load` in a ref (like the existing hasLoadedRef) and drop it from the
dependency arrays. Re-fetching is still driven by `deps`/`enabled`; the ref
always points at the latest closure. No API/behaviour change for callers.

Runtime-verified live: /jobs went from a render storm (frozen renderer, 100s of
console errors) to 0 errors in a 2s window and a clean render. Suites that drive
JobTable→useViewResource pass in isolation; the remaining full-run flakiness is
pre-existing (state-pollution/timing in the heavy RTL suites, unrelated).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-05 15:36:55 +02:00
cesnimda 490c5b803e fix(auth): stop infinite /auth/me request loop when logged out
The axios 401 interceptor calls clearAuthClientState() on every 401, which
dispatched "auth-changed"; the App handler re-fetched /auth/me, which 401'd
again → interceptor → clearAuthClientState() → "auth-changed" → ... an unbounded
request storm (observed live: 100+ GET /auth/me and climbing) that ran whenever
the user was logged out (login page, expired session) — burning CPU, network and
battery and flooding the server.

Fix: make clearAuthClientState idempotent — only emit "auth-changed" when it
actually removes a stored user key (a real signed-in→out transition), so
repeated 401s can no longer re-trigger the fetch.

Runtime-verified in a live stack: /auth/me went from 100+ & growing to 0 &
stable. login-page/settings tests green. Documented in
docs/performance/PERFORMANCE_IMPROVEMENTS.md (Phase 3.5 runtime finding).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-05 15:13:18 +02:00
cesnimda 39266c0935 docs(perf): memory-leak investigation reports (no leak found)
Evidence-based investigation across every leak vector (timers, listeners, object
URLs, observers, websockets, static server collections, IMemoryCache, Python
caches). Verdict: no confirmed memory leak — the codebase has disciplined
cleanup. One resource-release correctness bug (over-eager blob-URL revocation in
the CV carousel) was found and fixed (eed9b1f).

Adds docs/performance/: MEMORY_LEAK_REPORT.md, ROOT_CAUSE_ANALYSIS.md,
PERFORMANCE_IMPROVEMENTS.md.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-05 14:59:28 +02:00