felhom.eu/REPORT.md

felhom.eu — task reports

Overwrite this file with a summary of the most recent task only (uniform with the other repos; not cumulative). The cumulative hub history lives in hub/CHANGELOG.md; the scripts history lives in scripts/CHANGELOG.md.

Polish batch — hub v0.52.0 (F-15 instant reset codes + F-16 inline confirms) — 2026-07-13

Hub half of the polish batch (take-two findings F-15/F-16); companion controller v0.123.0. Commits bad9203 (F-15), 5d66c62 (F-16), 0e05c69 (manifest 0.52.0).

What shipped

1. F-15 (internal/api/handler.go): POST /api/v1/claim/reset-request returns the ACTIVE claim state — {claim: {code_hash, generation, issued_at}}, the report ACK's exact shape and bcrypt-only guarantee — so the box applies the rotated hash in the SAME request cycle and the emailed code works immediately (was: next-ACK, ~15 min — Viktor's take-two live failure). Served on every authorized outcome; a cap-reached refusal returns the unrotated row (a controller-side no-op by generation). The operator "Kód újraküldése" keeps its ACK lag — its flash + confirm copy now state it; the unmapped claim-resent/claim-resend-failed flash codes finally render.
2. F-16 (templates/inline_confirm.html + 4 templates + scripts/hub_confirm_gate.py): every native confirm() in the hub UI replaced by the inline "question + Igen/Mégse" two-step (felhomConfirm / data-confirm + requestSubmit, so formaction buttons riding the parent form still work). Danger-zone typed-confirm untouched by design. Gate enforces zero native confirm/prompt.

Tests + red-proofs

• TestClaimResetRequest_ResponseCarriesRotatedHash (rotated hash == stored row, != pre-rotation row, bcrypt-only). Red-proof: handler mutated to serve the pre-rotation row → FAIL "response generation = 1, want the rotated one (stored 2)". Restored → green.
• hub_confirm_gate.py red-proof: one onclick="return confirm('redproof')" reintroduced → exit 1; restored → 0. Full hub green gate (build+vet+test) before each commit.

Deployed + live-validated

• ArgoCD app felhom Synced/Healthy, deploy/hub image felhom-hub:0.52.0, clean startup log.
• F-15 live: drill-box reset request → controller cached the response hash 1 s after the POST; Viktor re-ran his exact take-two failure path — emailed code accepted immediately.
• F-16 live (its own regression test): "Re-issue offsite credentials" on demo-vm-felhom driven fully by browser automation — inline strip rendered, Igen → ?flash=offsite_reissued, zero CDP freezing (with v0.51.0 this exact click froze the session).

Observations

• Take-two §4 dispositions updated: F-15 and F-16 SHIPPED (this batch); the zero-toggle obs. shipped controller-side (v0.123.0).
• Remaining from take-two: F-14 (tenancy re-attach dead-end — needs the operator ruling on auto-Reissue vs never-silently-re-key) and the S5 DR-restore drill.