A new box with an older hub sent `ended` in the window request; that hub reads
4 KiB of it, decided the window on a wrong count, and the weekly clean-up
stalled.
- The box sends `ended` only when the hub's last window answer said
accepts_ended (kept in settings, re-read from every answer, so a rolled-back
hub stops getting them after one answer). Otherwise the request is the one
every hub always got, and the ordinary clean-up runs.
- One request carries at most 5000 snapshot ids.
Tests seen red before the fix; 6 mutations seen red (felhom.eu
documentation/audits/build-release-4-2026-11-15/A/redproof-review-5.txt).
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
A `.fab` import installed the app and its end record survived; the restore
that installs from a unit had the same hole. Only DeployStack deleted it.
- stacks.SaveAppConfig, the only writer of an installed app's app.yaml, calls
noteAppInstalled when the record says deployed (a hook per stacks directory,
registered by SetMigrationDeps). A source pin lists every place that sets
Deployed or writes an app.yaml; a new writer fails it.
- The living check is stacks.Manager.IsLivingApp (wired in main.go): the
protected list, the manager's list (deployed, being deployed, protected) and
the app's record on disk — an installed app the manager does not list is
living. Tested by behaviour.
Tests seen red before the fix; 8 mutations seen red (felhom.eu
documentation/audits/build-release-4-2026-11-15/A/redproof-review-3.txt).
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
The due end records and the candidates were read once, before the hub was
asked, and never again before the forget. A „Mégis megtartom” or an install
while the hub answered still deleted the copies.
The window now reads the records again immediately before the forget and takes
out every id whose record is gone, changed or no longer due. From that look
until the forget has ended, a take-back is answered „not marked, or already
deleted”, never „kept”. Not covered: an install during the forget itself.
Tests seen red on f628cb5, then green; mutations in felhom.eu
documentation/audits/build-release-4-2026-11-15/A/redproof-review-2.txt.
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
Built, not released; no box has run it. felhom.eu `07-backup-architecture.md` §8.2f.
It does nothing without a hub that knows it: an older hub agrees to no id.
- The record: settings `offsite_ends` {app, removed_at, ends_at}. One writer
(backup.StartOffsiteEnd), called by a removal only with „delete backup data”
and by „Törlés 30 nap múlva” on the restore page. A second press changes
nothing. No record, no deletion. Only on the hub-provisioned tier.
- The date is on the remove dialog before the press; the restore page shows
„Törlésre kijelölve …” with „Mégis megtartom”, and one history line per
carried-out end. An install of the same app deletes the record.
- The weekly window: after the guard has passed the policy's plan under the cap
as before, it adds the snapshots whose app tags are ALL due records, that the
hub agreed to and the box named in the same request; the result names them
apart (removed_for_ended_apps). `_shares`, the box's own stacks and an
installed app are never due.
- The report carries the pending records; the household gets one mail per
press (offsite_end_scheduled), in both languages.
MinAgent unchanged (0.158.1). 143 parity fixtures gained the dialog's one new
script line (patched, not regenerated); one new parity case.
Tests TestD72_*, TestD84_*; 56 mutations seen red
(felhom.eu documentation/audits/build-release-4-2026-11-15/A/redproof-ctrl-*.txt).
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
- clear_hold: a fifth verb in the closed operator-action list; lifts a hold of any reason in the RUNNING
controller (memory + settings.json together), forgets the update's held flag, drops a waiting put-back
from the app-stop marker (copy kept). Refused on a bad name or no hold. No new event type.
- --clear-restore-hold asks the running controller over <data dir>/ctl/hold.sock (unix, 0600, peer uid
checked, no TCP); edits settings.json only when no controller answers, and says so (in-guest restart).
- R-1340: the operator's HELD STOPPED mail is built from the hold record (restore kind, mixed cause,
failed put-back step, database replay or not, kept copy) — never the option-C database story by default.
- R-1341: the OOM WARN line once per new kill (container, start time, kill count); event unchanged.
H4: a failed definition write in the off-site restore writes the live definition back before the start;
a failed write-back holds the app (restore_mixed).
H5: a put-back whose stop fails pours nothing, keeps the copy and holds the app (off-site, local, power-cut);
a failed definition write-back keeps the copy (D20); a failed 'volume rm -f' fails that volume (no merge).
H6: TryBeginRestoreOp/RestoreBusy: check every operation and record the restore in one step under m.mu;
every web handler and the kept loads begin through it; install-time kept load asks about the mirror too.
H8: SweepUnusedRestoreCopies every 6 h: removes a copy only with no hold, no marker naming it, no restore
running, and older than 14 days.
Commit only; rides the next controller release.
R-1218: escrowStart stages when OffboxRepoPasswordOnBox() (not OffboxConfigured()). R-1219: the top bar hides during an
offer; the start handler answers 409 (api.escrow.recovery_offer_open, hu+en) before staging. Red-proved. Commit only.
internal/upwatch: exact per-container counters from the Docker Engine API, per-app judging (upload >= max(5 Mbit/s,
25% of the line) for 15 min; CPU >= max(150%, 50% x cores) for 30 min), explainers qBittorrent / Transmission / Plex,
app_unexplained_upload / app_unexplained_cpu + _cleared (hub 0.169.0). The off-site push is sampled once a second
for the line (a lower bound). qBittorrent limit = 50% of the line, the household's limit always wins. 20 red-proofs.
Spike + design: felhom.eu audits/SPIKE-upload-watch-2026-10-23.md, architecture/08 section 9. Not released.
internal/privpolicy checks the RENDERED compose (privileged, cap_add, devices, host namespaces, security_opt,
sysctls, volumes_from, build/include/extends, env_file/config files outside the stack dir, aliasing named volumes,
binds outside the box's data roots, the Docker socket, propagation, ${VAR} in a privileged key) against an
allow-list compiled into the controller (one entry: homepage's read-only socket). Asked before every compose
write: the 15-minute sync, the guarded update/ladder step, the install, both settings changes, both restores
(before the stop). A refusal writes nothing, keeps the app on its last good compose, raises the operator-only
event app_definition_refused once per definition and puts one line on the app page. The sync never copies a
template named like a controller infra stack. Structural guard TestAllowList_IsWiredIntoEveryComposeWriter.
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
R-1111 reserveAttempt is the one limiter body (share password, escrow re-auth, dashboard sign-in). R-1112
family.json holds fingerprints; old files are converted at load. R-1113 the household's pass reuses its
family session. R-1109 the samba compose carries the smb.conf hash. R-1110 TestNoBareDrivePathToPathHelpers.
R-1114 shared_scripts.py resolves the felhom.eu sibling from a worktree. Each red-proved.
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
Tier-2 and the one-copy notice ask the host agent (POST /disks/same-disk, agent 0.157.0): different +
proven -> made and counted; same disk -> not made (D1); unknown (VM, hidden USB serial) -> made, not
counted, the row says why (D2); a NAS never counts and no longer hides the notice (R-1075). An older
agent keeps the st_dev rule, logged once. D4: a data-folder or media app on a box with no drive is
refused in words with a link to the drives page (R-1068's raw-error half). MinAgent 0.157.0.
Claude-Session: https://claude.ai/code/session_012qRErfCoiTkvDK9N5XHbzb
Seen on demo-hp 2026-10-10 (adventurelog x4, bookstack x4). The synthesised run skips IsUndo files; only undo copies
-> no synthesised run. Two tests, both red-proved. v0.306.0 (R-570) was built but never delivered; 0.307.0 carries it.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
Fleet condition read on each box's settings.json 2026-10-10 (offbox.last_success ~02:16Z on 0.304.0).
Red-proof: the new no-kind row of TestR553_StaleNoteDecisionSurvivesWordingChange failed against the fallback.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS