Copre le parti nuove e una trappola che sarebbe passata inosservata.
- unit: completamento del profilo (30/30/30/10), stato di scadenza dei
documenti, export CSV, conversione riga <-> modello, anagrafica di squadra.
- unit: risoluzione dei permessi admin, incluso il fatto che con una sessione
attiva decide il database e la lista di nomi non conta più.
- integration (schema-profili): verifica su un database vero le colonne che il
codice legge, il bucket privato e la chiusura verso l'utente anonimo.
- e2e: /admin entra nell'elenco delle schermate verificate.
schema-profili tenta scritture da anonimo per dimostrare che la RLS le respinge,
e poi rilegge la riga: su un UPDATE che tocca zero righe PostgREST risponde 2xx,
quindi fidarsi del codice di risposta darebbe un falso verde. Si salta da solo
dove M2/M3 non sono ancora applicate, indicandolo nel motivo.
Verificato: 8/8 contro lo stack locale (che ha M2/M3), 21/21 file con
npm run test:all contro il progetto cloud.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
test/README.md covers the commands, what each folder verifies, whether it
needs the network, and the one real limit: the app renders after
hydration, so the end-to-end tests check HTTP responses and the data
behind the pages, not the rendered interface.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
sessioneScaduta compared Date.now() against NaN when aggiornato_il was
not a valid date, and every comparison with NaN is false: the session
was reported as still active, so the Scout Live table stayed locked to a
player who could no longer release it.
An unreadable timestamp now frees the session, which is the safe
direction: at worst someone takes over a scouting session that was
already unattended.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The project had no automated verification at all, so the "Test" step in
the AGENTS.md workflow rested entirely on clicking through the app.
The suite runs on bun with node:assert and adds no dependency: every file
is a script that exits non-zero when a check fails, and test/run.ts runs
each one in its own process. Unit tests cover the rules that decide what
players see (badges, streaks, ball duty rotation, ratings, MVP ties,
scouting totals, goals, notifications, CSI parsing); integration and
end-to-end tests drive the real dev server. Nothing writes to the
database, so both can be pointed at a live environment.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
davideandClaude Opus 5