Match the concise style of the other field tooltips instead of
explaining the whole update/rollback flow.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Found by running the installer for real rather than testing the hook in
isolation. Stalwart never started: it failed to open its database with
"Permission denied" on /var/lib/stalwart, which meant no mail could be
stored and the setup wizard could not be completed by hand either.
The image runs as its own uid 2000. LibrePortal gives container directories
to the docker install user under rootless and to the manager under rooted,
and 2000 is neither, so the bind mounts were unwritable in both modes. This
was not something the new provisioning introduced — it predates it, and the
app has never been able to hold mail.
Running as container-root maps to whichever host user owns those
directories. Under rootless that is the unprivileged docker install user,
not host root.
Also stop discarding the server's error when setup fails. Both failures
that actually occur — a hostname under a TLD that does not resolve, and the
unwritable data directory above — name themselves precisely, and a bare
"setup failed" turns a one-line fix into guesswork.
Verified end to end through `libreportal app install stalwart` on a clean
install: setup applied, DKIM keys generated, postmaster mailbox created,
and the full record set printed from the server's own zone data.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A new Stalwart drops you into a five-screen wizard — hostname, domain,
storage backend, directory, logging, DNS — before it will do anything.
LibrePortal already knows the two answers that matter and the rest have
sane defaults, so asking is asking a question we can answer ourselves.
v0.16 exposes those wizard fields as a `Bootstrap` singleton, so the whole
thing is one `update` applied through the Stalwart CLI. The CLI is not in
the server image (upstream split it into its own repo), but it publishes a
multi-arch container, so we borrow the server's network namespace and run
it there — nothing installed on the host, nothing to clean up, arm64 works.
Setup now also:
- generates DKIM keys (Ed25519 + RSA) with rotation left switched on, and
requests a TLS certificate. That last one is easy to miss: Traefik only
fronts the admin port, so 25/465/587/993 never see its certificate and
clients would hit a self-signed one on 993.
- creates postmaster@<domain>. The generated zone points DMARC and TLS-RPT
reports there and nothing was creating it, so those reports bounced.
- prints the record set read back from the server rather than composed
here, so it includes the real DKIM public keys, MTA-STS, TLS-RPT and the
SRV records clients autoconfigure from. This hook used to tell the user
to go and fetch DKIM themselves; by that point the keys exist.
Optionally hands DNS to a provider API (Cloudflare/DigitalOcean/DeSEC),
which keeps the whole record set in sync and makes DKIM rotation safe to
leave on. Off by default: the token can write to your zone and lives in
the mail server's database.
Re-running is safe — provisioning is skipped once config.json exists, and
the plans use upsert so they reconcile rather than duplicate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A failed verify makes the engine abort and restore, and a restore cannot
put back a bundle that was never downloaded — it would roll a working
mail server back a version to fix a missing web page, then hit the same
empty GitHub fetch next time. So the console check now warns loudly and
returns 0; readiness stays the only gate.
Renamed to stalwart_upgrade_check_admin_ui so the name cannot be read as
part of the gate, and bounded its poll to a 60s grace window (capped by
the caller's deadline) — the upgrade result is already decided by then,
so there is no reason to hold the run open on a web asset. The unreach-
able-probe branch is advisory for the same reason.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Stalwart v0.16 does not ship the WebUI in its Docker image — the admin
console is fetched from GitHub on first start. With no outbound HTTPS at
that moment the fetch fails silently: /healthz/ready still answers 200
because the mail server genuinely is serving, so both the installer and
the upgrade verifier reported success while /admin and /account 404'd
with nothing to explain why.
Install hook now probes /admin after the port-25 and PTR checks and, on
404, names the GitHub download as the cause rather than emitting a
generic failure. Upgrade verifier treats stable readiness as necessary
but not sufficient and confirms /admin before returning 0; the console
is polled under the same deadline because the bundle download runs
behind the server coming up, and failing on the first 404 would abort an
upgrade that was seconds from finishing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Replaces the drawn shield-and-envelope stand-in with the real mark from
stalw.art (/favicon.svg), in their #DB2D54.
Padded from the source's 159.95x139.07 to a square 159.95 viewBox with
the art vertically centred, matching every other catalogue icon — all of
which are square, so a non-square box would letterbox in the app tiles.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Second correction: the earlier version was only the outer blue disc.
This keeps all 10 artwork paths (circle, body, ears, muzzle, wool, face)
and drops the <g> holding the 'vikunja' wordmark, since a catalogue tile
wants the mark alone. Cropped from the wide 872.6x256.8 logo viewBox to
the mark's own 0 0 256 256, and the web-app attributes (class,
xml:space) removed.
Extracted programmatically from the supplied SVG rather than
transcribed, so no path data could be mangled by hand.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Replaces my drawn placeholder with the actual logo path supplied by the
maintainer.
Kept as .svg rather than .ico: webuiSyncAppIcon only looks for
<app>.svg then <app>.png, so an .ico would never be copied into the
frontend and the app would silently fall back to the default icon. SVG
is also the right format here — it is a single vector path that scales
to any tile size, where an .ico would bake in fixed raster sizes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Removing Focalboard from the catalogue left its icon still being served:
the sync only ever ADDS, so every app ever dropped leaves a file behind
that the portal keeps offering for something that is gone. Same shape as
the task queue that only ever appended.
webuiPruneAppIcons runs at the end of the sync and removes only icons it
can match to a missing template — anything else in the directory is left
alone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Retired last commit, deleted now at the maintainer's call. Mattermost
ended support in 2023, the community repo is asking for maintainers, and
the image had not been rebuilt in 1042 days — the staleness signal's
worst case after speedtest. Vikunja covers the same ground and is
actively developed.
Self-contained: every reference lived inside containers/focalboard/ plus
its generated manifest entries, so nothing else needed touching. Git
keeps the history.
Anyone with it already installed keeps a running container and their
data — removing the template only stops NEW installs. Their app will now
report as unknown in the App Center rather than offering an update,
which is the honest state for software with no upstream.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Dead weight: 19 app directories, 101 files, none referenced by any live
code path and already export-ignored so they never shipped in a release.
Several were actively misleading — the mailcow attempt in there is what
the mail-server discussion kept having to explain around, and none of
them would survive contact with the current conventions (tag sentinels,
port manager, backup labels, update policy).
Verified before deleting: nothing outside the tree references it, and
none of the 19 duplicates a live app in containers/. Git keeps the
history if any of them is ever wanted back.
Function manifest and source arrays regenerated — no entries pointed
into the deleted tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Found by installing it. Two problems, one fatal and one worse.
The container crash-looped: the image runs as uid 1000, which under
rootless Docker maps to host sub-UID 232071 while the bind mounts are
owned by the install user, so Vikunja died on its first write to
/app/vikunja/files and restarted forever. Its own error message
diagnosed it exactly. Fixed with the existing USER_TAG mechanism the
portal container already uses — 0:0 under rootless (container root IS
the install user on the host), the real uid:gid under rooted — rather
than hardcoding either.
More seriously, the compose mounted the docker socket, copied in from a
template that needed it. A task manager has no business talking to the
daemon, and the socket is root-equivalent access on the host. Removed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Focalboard is the one audit finding with no successor to follow.
Mattermost ended support in 2023, the community repo is openly asking
for maintainers, and the image has not been rebuilt in 1042 days.
Adds Vikunja as the replacement for NEW installs: lists, kanban, table
and gantt — the same job Focalboard did — from a project rebuilt 14 days
ago. One container on SQLite, no database sidecar, following the
catalogue conventions (tag sentinels throughout, category/title +
backup.db/backup.files labels, traefik block, gluetun markers).
VIKUNJA_SERVICE_PUBLICURL is wired to the existing APP_URL_TAG rather
than a hand-built URL. It is not optional for this app — get it wrong
and creating the first account fails with a bare "unauthorized" — and
APP_URL_TAG already resolves to https://<domain> behind Traefik or
http://<host>:<assigned-port> otherwise, so the port is never guessed.
(First attempt invented a PORT_DATA_1 tag that does not exist; checking
what the processors actually emit found the real mechanism, which
bookstack already uses.)
Focalboard is RETIRED, not deleted. Replacing an app in place would stand
still for anyone already running it — their data does not move to Vikunja
— so it keeps working, and instead:
* the description says plainly that it is unmaintained, why, and what
to use instead
* UPDATE_TYPE drops to manual, because there is nothing to update TO
and auto-pulling a 2.8-year-old tag is pure churn
Icon is a drawn placeholder, not the upstream trademark.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
adolfintel/speedtest was last rebuilt 1624 days ago — 4.4 years, the
worst in the catalogue. LibreSpeed itself is fine: the project moved to
the librespeed org, and LinuxServer.io's build was rebuilt 2 days ago.
Chose lscr.io/linuxserver/librespeed over the org's own ghcr.io image
specifically because it is on Docker Hub: our tag enumeration and the
new staleness signal can query Hub but not ghcr, so this keeps the app
visible to the tooling that would catch it going stale again.
Not a tag swap. The LSIO image uses their house conventions, so the env
block is ported rather than copied: PUID/PGID/TZ (matching bookstack,
the catalogue's other LSIO app) and everything under /config instead of
/database. MODE, WEBPORT and ENABLE_ID_OBFUSCATION have no equivalent —
this image is standalone by design and always serves on 80 internally,
which is what our port mapping already assumed.
PASSWORD and DB_TYPE keep their names and meaning, so the existing
config keys still drive the results page and telemetry. DB_NAME is
deliberately left unset: the image picks a path inside /config, and
guessing one risks a database sitting outside the volume.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
zadam/trilium was last rebuilt 810 days ago because the original author
handed the repository to the community project and put his own in
maintenance mode. The project is very much alive — triliumnext/trilium
was rebuilt 24 days ago. Same story as Pi-hole: we were pinned to an
abandoned original, not a dead project.
A genuinely clean swap, verified against upstream's own docs rather than
assumed: identical data path (/home/node/trilium-data), identical port
(8080), and TriliumNext states there are no special migration steps — it
opens an existing zadam database as-is. TRILIUM_DATA_DIR is now set
explicitly, matching upstream's compose.
URL metadata repointed at the maintained repo so the App Center links
somewhere that still gets commits.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
mvance/unbound was last rebuilt 668 days ago. Replaced with
madnuttah/unbound (14 days): distroless, runs unprivileged as non-root,
listens on 5335 by default — exactly the "upstream behind a blocker"
shape — and publishes clean semver tags. klutchell/unbound is equally
fresh but defaults to port 53 (fighting Pi-hole/AdGuard for it) and its
tag namespace is CI build soup.
The shipped config was worse than the stale image. It was not a
recursive resolver at all:
interface: 0.0.0.0@53
forward-addr: 10.100.0.3@53 # "Local AdGuard" — a hardcoded IP
forward-addr: 9.9.9.9@853
So it listened on 53 (conflicting with any blocker on the same host),
forwarded to Quad9 — surrendering the "nobody sees my queries" property
that is the only reason to run Unbound in front of a blocker — and
pointed at AdGuard, inverting the dependency: AdGuard should point HERE.
Replaced with a drop-in at conf.d/libreportal.conf. The image's own
unbound.conf ends with `include-toplevel: conf.d/*.conf`, so ours ADDS
to a working recursive config the image author maintains rather than
replacing it — upstream keeps owning the parts that change between
Unbound releases. It contributes access-control (private ranges allow,
everything else REFUSE, so this can never become an open resolver for
amplification attacks), DNSSEC hardening, rebinding protection, and
cache sizing suited to a small VPS. Forwarding is included commented
out, with the trade stated rather than silently chosen.
Ports corrected to 5335:5335 — the old mapping assumed an image
listening on 53 internally. Added the libreportal.category/title labels
the app was missing (no traefik labels: it has no web interface).
Install hook copies the drop-in and repairs a stub directory first, the
same trap that kept Nextcloud's nginx from starting.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The app shipped cbcrowe/pihole-unbound — a third-party bundle last
rebuilt 841 days ago. Pi-hole itself is fine: the official
pihole/pihole was rebuilt 42 days ago. We packaged an abandoned fork,
not a dead project.
Not a tag swap. The official image is v6, which replaced nearly every
v5 environment variable with an FTLCONF_ equivalent — and the container
ACCEPTS the old names and ignores them, so a v5-style block looks
correct while configuring nothing, including the admin password:
WEBPASSWORD -> FTLCONF_webserver_api_password
WEBTHEME -> FTLCONF_webserver_interface_theme
PIHOLE_DNS_ -> FTLCONF_dns_upstreams (";" separates values)
DNSSEC -> FTLCONF_dns_dnssec
DNSMASQ_LISTENING -> FTLCONF_dns_listeningMode
REV_SERVER{,_TARGET,_DOMAIN,_CIDR} -> FTLCONF_dns_revServers, one
combined "<enabled>,<cidr>,<target>,<domain>"
FTLCONF_LOCAL_IPV4 -> gone in v6
listeningMode is ALL rather than the old "single": on a bridge network
queries arrive via the docker gateway, and "single" drops them.
The bundled unbound is gone, so PIHOLE_DNS_=127.0.0.1#5335 pointed at
nothing. New CFG_PIHOLE_UPSTREAM_DNS defaults to Quad9, with the config
documenting how to point it at the unbound app for full recursion. The
freed port slot becomes the (disabled) DHCP port, which v6 supports.
Volumes: v6 keeps config, databases and gravity under /etc/pihole, and
ignores /etc/dnsmasq.d unless explicitly re-enabled — so the old
two-mount layout is replaced by a single ./etc-pihole.
Variable names taken from the official v5->v6 upgrade doc, not memory.
Substitution verified end to end: every placeholder resolves and
revServers renders in the documented format.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
"Up to date" answers one question — has the tag I track moved? — and an
abandoned project answers it reassuringly forever. The tag stays put, the
digest never changes, and the app reports as current while receiving no
security patches at all. Nothing in the UI could tell a healthy stable
app from a dead one.
An audit of all 34 anchor images found five in exactly that state:
speedtest (4.4y since rebuild), focalboard (2.8y — Mattermost dropped
support in 2023), pihole-unbound (2.3y), trilium (2.2y), unbound (1.8y).
The scan now records image_updated_at per app (one cheap Hub call inside
the existing registry window, cached between windows like everything
else) and emits stale_after_days from CFG_UPDATER_STALE_DAYS (365, 0
disables) so the UI and the config agree on one number.
Surfaced as an "unmaintained?" severity chip on the fleet row and a
dated explanation in the app detail. Phrased as an observation rather
than an accusation — plenty of small tools are simply finished — but it
does spell out the security consequence, because that is the part a user
cannot infer from "up to date".
Deliberately NOT a "needs action" row on the Overview board: it is not
fixable by pressing anything, and a permanently amber board teaches
people to ignore the board.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The registry helpers rejected any repo containing a dotted host segment,
which caught 'docker.io/authelia/authelia' — Docker Hub spelled out in
full. Those apps were silently skipped by tag enumeration and version
laddering. Strip the docker.io/ and index.docker.io/ prefixes before the
host check; genuinely third-party registries (ghcr.io, quay.io, lscr.io)
are still correctly skipped.
Found by auditing every app's anchor image.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The live Nextcloud 31→34 climb left 4.4 GB of images behind — one per
rung, each ~1.5 GB, all still present after it finished. On a small VPS
that is the difference between working and full.
`system reclaim` cannot help: it collects DANGLING images, and every rung
is a real tag, so all of them stay tagged and stay on disk. (Rolling apps
never hit this — moving a floating tag orphans the old image, which
reclaim then collects. It is specific to laddering.)
After a SUCCESSFUL climb only, remove the images stepped through, keeping
the immediately-previous version so a roll-back needs no download.
CFG_UPDATER_UPGRADE_PRUNE=false keeps everything. Never runs on failure,
where the older images are exactly what recovery may need.
Tested: a 3-rung climb removes 31 and 32 and keeps 33; a single-step
climb removes nothing (its previous version IS the rollback target); the
config switch disables it.
Found by looking at the box after the first real ladder run — the feature
worked, and then quietly cost 4.4 GB.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`updater upgrade <app> --detach` parsed "--detach" as the target version
and refused with "no safe path from 31-fpm-alpine to --detach". It failed
safe, but blaming the version for a misplaced flag is a poor way to say
the flag is not supported here. Unknown options now say so.
Found during the first live ladder run.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Found by actually installing it. The compose bind-mounts
./resources/nginx.conf into the web container, but nothing ever copied
that file into the container tree, so Docker created a DIRECTORY in its
place and nginx died with:
error mounting ".../resources/nginx.conf" to rootfs at
"/etc/nginx/nginx.conf": not a directory
Worse than a hard failure: the app still recorded as installed. Three of
four containers came up, the DB and the app itself were fine, and only
the web front end was missing — a quiet, partial install.
Apps needing a resource file declare the copy in a hook (authelia does
exactly this); Nextcloud simply never had one. Adds
nextcloud_install_post_compose — after the compose file is written,
before permissions and `up` — which repairs any stub directory left by a
previous attempt and then copies the file.
The stub repair matters: without it the copy lands INSIDE the directory
(resources/nginx.conf/nginx.conf) and the mount fails identically.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Puts the stepped engine behind the "34 available" chip so it is not
CLI-only. Wired end to end: updater_upgrade task type -> task-router ->
updaterUpgrade action -> `libreportal updater upgrade <app> [version]`,
with a label and icon in the tasks list.
The button always confirms, and the dialog states the plan and the
guarantee rather than asking "are you sure?" — which app, from which
version to which, that every step snapshots first and waits for the app
to confirm it is serving with no migration outstanding, that a failure
stops the ladder on the last version that verified, and that it can take
a long time because each release runs its own migration.
Both delegated dispatchers (per-app Updates tab and the fleet Overview
rows) learn the action, so the button works wherever the chip appears.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Ties the ladder and the verifiers together behind a new verb:
libreportal updater upgrade <app> [version] [--dry-run]
Per rung, and every part is load-bearing:
snapshot (fail-closed) -> set version -> pull -> up -> VERIFY -> next
On failure anywhere: restore THIS rung's snapshot, put the version back,
stop, and leave the app on the last version it actually verified at. The
ladder never continues past a doubt.
A snapshot PER RUNG rather than one at the start, because upstream
migrations are usually one-way — Nextcloud 32's schema cannot be undone
by putting the 31 image back. The recovery guarantee is "restore the
snapshot from sixty seconds ago", which only holds if every rung has one.
Two gates before anything moves. An app with no <app>_upgrade_verify is
refused outright: the generic health check cannot see a half-finished
migration, so laddering on it would be a guess wearing a safety label.
And a ladder that cannot be computed end to end refuses rather than
attempting a partial climb.
`updater upgrade` is a separate verb from `apply` on purpose: apply moves
you WITHIN a release line (and may be automatic), upgrade moves you
BETWEEN lines and is always a deliberate act. Dry runs execute inline so
the plan is instant to read.
updaterSetAnchorVersion rewrites the image tag AND its version sentinel
together — updating only the image would leave the sentinel advertising
the old version, and the next config regeneration would silently revert
the app.
Tested with stubs against the real code paths: the no-verifier gate holds
and changes nothing; a dry run has zero side effects; the happy path
snapshots at each current version before moving; a verify failure on rung
2 of 3 stops with the app on rung 1, restored, and never touches rung 3;
a failed snapshot moves no version and pulls nothing; a container that
will not start is rolled back.
NOT yet exercised on a live install — no app here needs a ladder. The
first real run should be a dry run.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Stepping 31 -> 32 -> 33 is arithmetic. Knowing 32 FINISHED before
touching 33 is the whole safety story, and it is invisible from outside
the app: Nextcloud runs its migration on boot and sits in maintenance
mode — or fails halfway — while Docker reports the container perfectly
healthy. Advance a rung there and a migration has been skipped on live
data.
Contract: <app>_upgrade_verify <app> <expected-tag> <deadline> -> 0
Returns 0 ONLY on positive confirmation that the app serves at the
expected version with nothing outstanding. Unhealthy, indeterminate and
timed-out all return non-zero — uncertainty is a failure, not a maybe,
because the alternative gambles with data.
nextcloud `occ status`: installed, NOT in maintenance, no pending DB
upgrade, and the running major matches the tag. Maintenance
mid-migration is expected and simply keeps waiting.
mastodon /health serving, ZERO "down" rows in db:migrate:status, and
the version from /api/v1/instance matching. /health alone is
insufficient — Puma answers before migrations finish.
stalwart /healthz/ready (per its documented probes), required to hold
stable rather than flash once. Weaker by design: the probes
confirm serving but report no version, and the file says so
rather than implying more.
updaterVerifyGeneric (running + healthy + no restart during a settle
window) is the fallback for everything else, and is explicitly NOT
sufficient to justify climbing a rung — the engine will refuse to ladder
an app with no declared verifier.
9 tests drive the dangerous states directly: maintenance mode, pending DB
upgrade, and a wrong major all correctly REFUSE to verify; clean states
pass. Those three negatives are the ones that would have corrupted data.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Foundation for stepped upgrades. Answers one question only — WHICH
versions, in WHICH order — with no side effects, so it can be tested
exhaustively. Applying the rungs is a separate job.
Nextcloud refuses to skip a major ("Updates between multiple major
versions and downgrades are unsupported") and will not start; databases
behave the same way about their data directory. For those apps 31 -> 34
is three upgrades, each with a migration that must finish before the
next begins.
Built by PROBING each candidate rung, not by enumerating tags — because
enumeration is provably unsafe here. Docker Hub pages at 100 ordered by
recency, and the first real-registry run proved the danger: it produced
v4.2 -> v4.4 -> v4.5 -> v4.6 for mastodon, silently skipping v4.3, which
exists (HTTP 200) but had fallen off the newest-100 listing. Skipping a
rung is the precise failure this file exists to prevent, so the ladder is
now built by incrementing and probing: v4.2 -> v4.3 -> v4.4 -> v4.5 ->
v4.6, 4 steps.
Guarantees: same shape only (never 31-fpm-alpine onto 31-apache),
strictly ascending, never a downgrade, rolling tags refused outright, and
a version upstream never published is stepped over only because the probe
said so. If a continuous path to the target cannot be constructed it
returns 1 and prints nothing — refusing to guess, because a wrong ladder
means a skipped migration.
20 unit tests, including the exact listing-truncation case above and the
numeric ordering that would otherwise drive an app backwards (0.9 vs
0.10). Real registry: nextcloud 3 steps, mastodon 4, stalwart current.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both templates were stale in different ways, and the new tag enumeration
surfaced it: nextcloud sat on 31-fpm-alpine with 34 out, mastodon on
v4.2.0 with v4.6 out.
nextcloud 31-fpm-alpine -> 34-fpm-alpine
mastodon v4.2.0 -> v4.6
The mastodon one was the real problem: v4.2.0 is an EXACT patch pin, so
it never moved at all — no security patches, ever. v4.6 is a moving
minor-line tag (the same shape as stalwart's v0.16), so auto-update now
delivers patches within the line.
Deliberately NOT floated to :latest or :stable. Both projects require
stepped upgrades — Nextcloud in particular refuses to skip a major — so
a tag that crosses majors on its own would break the app on a routine
container recreate. A major-pinned, patch-moving tag is the correct
shape here, not a limitation.
Verified against the registry: all three pinned apps now report nothing
newer, and each tag still moves (v0.16 2026-08-10, 34-fpm-alpine
2026-08-03, v4.6 2026-08-06).
Templates only, so this changes NEW installs. Existing installs keep
their tag and will now be told a newer line exists.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The digest compare only ever asks about the tag already pinned, so it
answers "has my tag been rebuilt?" and can never answer "does a newer
version exist?". An app on v0.16 reports up to date forever while 0.17
ships. That is the gap between an app that updates and an app that is
current, and it silently affects every pinned app.
Adds tag enumeration for VERSIONED tags only (rolling tags already move
on their own): list the repo's tags, keep those sharing the current tag's
SHAPE, and pick the numerically greatest.
Shape matching is the whole safety story — v0.16 -> v#.# so it can never
"upgrade" you onto v0.16-alpine, 31-fpm-alpine onto 31-apache, or a date
tag onto a semver one. Comparison is component-wise numeric, so 0.10 > 0.9
and 1.0 > 0.99 (a string sort gets both wrong), with 10# forcing base ten
so an upstream "08" cannot be read as octal. 15 unit tests cover it.
Docker Hub only, deliberately: all three pinned apps live there, it needs
no auth, and the generic OCI tags/list wants a per-registry token dance.
Other registries stay quiet rather than guess. Throttled inside the
existing registry window and cached between windows so it cannot flicker.
Surfaced as INFORMATION, never an action: no button applies it, because a
version move can carry a data migration. `update_available` and the "up
to date" badge keep their exact meaning; the new state sits beside them
and points at the Version field.
Against the live registry: stalwart v0.16 is current, nextcloud is on
31-fpm-alpine with 34-fpm-alpine out, mastodon on v4.2.0 with v4.6.5 out.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
It sets the image tag, so a wrong value stops the app starting — that
belongs with the other expert settings, not beside feature toggles.
Tooltip now explains the split it participates in: automatic updates
apply rebuilds OF this version, changing it moves between releases.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Reverses the manual default from 4ee2529, which was over-cautious once
the tag pin is taken into account.
Two things were conflated. Auto-update does not reinstall anything: it
snapshots, `compose pull`, `up -d` — the container is recreated from the
new image and the data volume is untouched. And because the image is
pinned to v0.16, the updater compares the digest of THAT tag, so auto
can only ever apply rebuilds of 0.16 (security/bug patches). It cannot
jump to 0.17. That is the safe half of updating, and there is no good
reason to withhold it.
Adds CFG_STALWART_VERSION=v0.16, which drives the image tag through the
existing #LIBREPORTAL|STALWART_VERSION_TAG| sentinel (verified: setting
it to v0.17 rewrites the image line). Moving between releases is now a
config change a user can make from the app's config page — the roadmap's
config-first version identity, used for real.
Net behaviour: patches land unattended inside the update window; a
version jump stays a deliberate decision, which is what pre-1.0 software
with a settling storage schema warrants.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every themed dropdown on the app config page was dead: it rendered, but
clicking did nothing. Only that page — every other dropdown in the WebUI
worked.
renderAppDetail captured the config section's own innerHTML right after
displayConfigForm() had rendered it...
const configHTML = document.getElementById('config-section')?.innerHTML;
...67 lines later...
configSection.innerHTML = configHTML;
...and wrote the same string straight back. That is a no-op for the
markup and a catastrophe for behaviour: re-assigning innerHTML re-parses
the subtree, so every listener in it is destroyed.
custom-select.js had already wrapped each <select>, so the captured
string contained the .custom-select wrapper and the custom-select-native
class. The re-inserted copy therefore looked enhanced — which made the
enhancer correctly skip it as already-done — while having no click
handler at all. A dropdown that renders perfectly and does nothing.
The container is never wholesale-rewritten in that function (it updates
header/config/console individually), so the capture-and-restore had no
purpose. Both lines removed, with a comment stating that anything added
there must mutate the section rather than re-assign its innerHTML.
Diagnosed in a real browser rather than by reading: instrumenting
CustomSelect.build() showed the enhancer DID build a widget for the field
while the wrapper in the DOM was not the one it built, and patching the
innerHTML setter named apps-manager.js:785 as the writer. Verified live:
the dropdown opens, and picking an option sets the value (auto) and the
label.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both found in a user's console log.
1. ReferenceError: setupMobileMenu is not defined (dashboard.js:98)
core/topbar/js/mobile-menu.js defines that global, and index.html
never loaded it. dashboard.js called it unguarded as the FIRST line
of setupEventListeners, so dashboard init threw every page load and
took loadInstalledApps() with it — and the burger menu was dead on
mobile. system-loader already guarded its own call with a typeof
check, which is why this survived unnoticed.
Loads the script (before dashboard.js) and guards the call, so
optional nav chrome can never take down the page below it again.
2. Endless 404s on /api/tasks/<id> for tasks that no longer exist
queue.json is append-only from the enqueue side and nothing ever
pruned it, so any task file removed afterwards left an id the WebUI
re-fetched forever, one 404 per poll per orphan. Adds
cleanupOrphanQueueEntries to the idle housekeeping pass: entries with
no task file are dropped and logged. Self-heals existing strays.
(Provoked by my own clean-up of two test tasks earlier in this
session, but the gap is real and predates it.)
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Reported: dropdowns dead on the app config page only, in a private window
(so not cache), with the served frontend confirmed identical to source.
The enhancer ran in a bare forEach with no try/catch anywhere, and
build() inserts its wrapper via `select.parentNode.insertBefore(...)`.
A detached select makes that a null deref, and one throw abandoned the
rest of the pass — every select AFTER it silently stayed native. The
MutationObserver callback had the same exposure for the remaining
mutation records in a batch.
The app config page is the one that can produce a detached select: it
builds its category panels in an async loop, so the observer can see a
node a later render already replaced. That matches "only app config".
Worse than losing the theme: if the throw landed after build() added
.custom-select-native, the select was left opacity:0 / pointer-events:
none behind a button with no listeners — a dropdown that looks right and
does nothing.
Now: detached/unconnected selects are skipped (they get enhanced when
their subtree is attached and the observer fires again), every
enhancement is individually guarded, a failed one is rolled back so it
can never be left invisible, and failures console.warn with the field
name instead of vanishing.
Verified with jsdom against the real file: detached nodes skipped without
throwing, a failure mid-pass leaves later selects working, and the
survivors open and set their value.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
One container providing SMTP/IMAP/POP3/JMAP plus CalDAV/CardDAV, an admin
UI and spam filtering — chosen over mailcow (owns its own installer, which
is what killed the earlier attempt now sitting in scripts/unused/) and
over Mailu (~7 containers) because a single image with a single data dir
is the only shape that fits the existing conventions cleanly: one anchor
service the updater can version, one path the backup engine can snapshot.
Mail-specific departures from the usual app template, each deliberate:
* Ports are FIXED, not random. Other mail servers connect to :25 by
number and clients expect 465/587/993 — a randomised external port
would silently make the server unreachable. Only the admin UI takes a
random port, since that one really is just a browser behind Traefik.
143/995/4190/443 ship disabled; the port processor comments them out.
* UPDATE_TYPE=manual and the image pinned to v0.16, not :latest.
Stalwart is pre-1.0 and has said the storage schema is still being
finalised, so an unattended minor bump could carry a data migration on
the message store. This is the one app where the auto default is wrong.
* BACKUP_STRATEGY=stop-snapshot-start. The message store is written
continuously; a live copy can land mid-transaction. Seconds of queued
delivery (senders retry) buys a consistent snapshot.
* The install hook checks outbound port 25 and reverse DNS, then prints
the MX/SPF/DMARC records with real values. A mail server whose
container started is not a working mail server, and every remaining
requirement lives at the registrar or the VPS provider.
Admin credentials are seeded via STALWART_RECOVERY_ADMIN from the app
config rather than left to Stalwart's first-run random password, which
would otherwise exist only in the container log.
Icon is a drawn placeholder, not the upstream trademark.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
getFieldMappings/getConfigCategories fetched host-GENERATED files with
default caching, so a browser that had the page open before a release
kept rendering the previous release's config UI — a newly shipped field
(UPDATE_TYPE) simply never appeared, with nothing on screen to hint the
page was stale. Only a hard refresh fixed it.
Adds {cache:'no-store'} to both, plus the one configs.json read in this
file that was missing it while two others already had it. Cost is a
conditional request per config-page open.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The portal was found crash-looping with EACCES on /app/webui_logins.
configs/webui/* are bind-mounted into the container, which reads them
through its GROUP (files 0640, group = container owner). Two paths reset
that group to the manager and never gave it back:
* init.sh setupConfigsFromRepo — `chown -R manager:manager` over the
whole configs tree on every install/redeploy (the documented local-
mode deploy), and
* the runtime config reconcile — rewriting a live config replaces the
file as the manager, so ANY release that merely adds a key to a
webui_* config would break the portal.
Neither breaks anything immediately: the running container holds its
open files, so the failure only appears at the next restart, long after
the change that caused it. That is exactly how it surfaced here — a
deploy in the evening, a dead WebUI later.
init.sh gains restoreWebuiBindAccess (prefers the root ownership helper,
inline chown fallback for the first install, no-op when the container
user does not exist yet) called right after the chown; the reconcile
calls the existing reconcileWebuiDirOwnership when it is in scope.
Verified by reproducing the break (chown -R manager over configs), then
running the fixed deploy and force-recreating the container: group is
restored to the container owner and the portal serves on 3179.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The missing piece of hands-off updates/backups: when a task fails while
nobody has the WebUI open, LibrePortal now says so — email (via the
existing Mail settings), ntfy, Gotify, Discord, Slack, Telegram, or
Pushover, configured under Settings → Notifications.
One hook, everywhere: the task processor reports every terminal task to
`libreportal notify task <id>` (detached, never load-bearing — hard curl
timeouts, failures ignored). The POLICY lives in the notify command, not
the daemon: CFG_NOTIFY_EVENTS = failures (default) | all | off, and
cancelled tasks never notify. Failure copy is task-aware — a failed
update says the app was already rolled back and won't be retried, so the
reader knows the box is safe before opening the WebUI.
`libreportal notify test` sends to every enabled channel with per-channel
results. Verified against a local mock endpoint: all webhook payloads,
JSON escaping (quotes/newlines), the events policy, and fail-fast on
dead endpoints (8ms, exit nonzero).
The v0.1.0 per-app NOTIFY_* field-mapping scaffolding (never wired to a
sender) stays as-is; this global channel is the system it was waiting on.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Four fixes that make the auto-updater a trustworthy background system:
* CFG_UPDATER_WINDOW (default 06:00-08:00 host time, right after the
05:00 backup cron; HH:MM-HH:MM wraps midnight, 'always' = any time).
Gates only the enqueue — scans keep running all day, so the Updates
page stays current and pending updates visibly wait for the window.
Malformed values fail closed and are rejected by the WebUI validator.
* "Check now" actually checks: an explicit `updater check` sets
UPDATER_REGISTRY_FORCE=1. The flag existed but nothing ever set it,
so the button silently reused the 6h digest cache and could not find
a build the user knew had shipped. Force also overrides interval 0,
which now means "manual-only" as documented in the roadmap.
* Registry stamp moved from /tmp to <system>/logs: the task processor
runs under PrivateTmp, so daemon and CLI each kept a separate 6h
clock and the daemon's reset on every service restart.
* A failed automatic attempt is no longer invisible: the scan emits
auto_attempted_digest (the one-shot no-retry stamp), and when it
matches the available build the UI stops promising an install that
will never come — per-app detail explains, the fleet row gets an
"auto failed" chip, and the Overview board counts it as needing you.
Also corrects the CFG_TIMEZONE label: it sets the containers' TZ only;
scheduled tasks follow the host clock (timedatectl), and the old
"Timezone for scheduled tasks" wording promised a knob that never
existed. The window + auto_window display state plainly WHEN updates
land, answering "how does the user know when the next update happens".
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
navidrome's install died on 2026-08-01 and left no explanation: the log
had only "Started container for navidrome (exit 1, up_app.sh:130)". The
compose output was captured into `result` and never read, and stderr was
not captured at all — so the one thing that says WHY (image pull EOF,
port clash, missing external network) was thrown away at the moment it
mattered.
Capture stderr and print the tail of the output on failure, before
checkSuccess (which can exit). Both the rootless and rooted call sites.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
First real end-to-end auto-update on a live install failed like this:
Automatically updating trivy (a recovery snapshot is taken first)
Snapshotting trivy before update…
Pulling new image(s) for …
Update of failed — rolling back…
Could not roll back automatically
The app name went empty after the snapshot. Cause: bash is dynamically
scoped, so a callee assigning an undeclared variable writes the CALLER's
local of that name — and a `while read app` loop leaves it EMPTY at EOF.
webuiBackupAppStatus's dashboard generator runs at the end of every backup
and did exactly that to updaterApplyApp's `app`.
Nothing was damaged: the pull ran against an empty name, failed before
touching the image, and the rollback was a no-op on a nonexistent app.
Fixed both ends. The generator (and three gluetun loops with the same
latent leak) now declare `local app`. updaterApplyApp/updaterRollbackApp
hold the name in `_upd_app` so they no longer depend on every callee's
hygiene, and updaterApplyAll stops leaking its own loop var.
This is exactly the untested path the roadmap flagged: "apply/revert not
yet exercised end-to-end on a live install with a pending update."
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds the decision half of the app updater. Detection (P2) and the
snapshot-first apply/revert (P3) were already real, but nothing ever
pressed the button — every update waited for a click.
CFG_<APP>_UPDATE_TYPE=auto|manual per app, default auto (33 templates)
CFG_UPDATER_AUTO=true|false master switch, default true
updaterAppPolicy resolves the two the way backupResolveStrategy already
resolves backup strategy: the global switch can only make things more
manual. updaterApplyAuto runs at the end of `updater check` and enqueues
the ordinary updater_apply task for each auto app that has an update —
never applies inline, so an automatic update is the same code path, task
log, History entry and Roll back button as a manual one.
Safety: each attempt stamps its target digest under generated/auto/, so a
build that fails is rolled back and then left alone rather than retried on
every scan; in-flight updater tasks are skipped so scans can't stack.
Tracked end to end: updates.json carries each app's resolved update_type,
History entries carry trigger=manual|auto. The WebUI says whether updates
install themselves, chips only the apps that opted out, labels automatic
history, and — since an auto app's pending update needs no decision — keeps
it off the Overview board's "Needs action" view.
Also fixes artifactApplyAuto enqueueing without --detach: it runs inside
the single-threaded task processor's own poll, so following the new task in
the foreground waits for a task that cannot start until it returns.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The board landed on "Everything", so the rows that want a decision were
mixed in with the healthy one-liners on arrival. Default to the "Needs
action" chip instead, falling back to "Everything" when nothing is
pending (that view would otherwise be empty). An explicit chip click
still sticks for the session.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Cut the 36-word blurb to 23 without losing either fact that matters: it's
DNS-based, and blocked domains resolve to a local blackhole.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Tags and Paths were stacked full-width blocks, so a single tag chip and a
single path each burned a whole row and left the panel mostly empty. Wrap
them in a .bsm-blocks auto-fit grid that seats them side by side and falls
back to stacking under ~460px. The wrapper now owns the divider, so it
renders once for the pair instead of once per block, and is omitted
entirely when neither block has content.
Applied to both renderers of this markup: the global Backups view and the
per-app backup card.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The App Center blurb ran three sentences, two of which restated things
the Security view already makes obvious. Keep the what, drop the rest.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
sudo-rs — the default sudo from Ubuntu 25.10, so on 26.04 — does not
implement bare -E. It does not reject it either: it warns to stderr
("preserving the entire environment is not supported, '-E' is ignored")
and runs the command with the environment DROPPED, leaving the exit
status untouched. Callers capture stderr, so the warning is invisible and
the backup engines simply never receive RESTIC_PASSWORD / BORG_PASSPHRASE
/ KOPIA_PASSWORD and cannot open the repository.
Name the nine vars explicitly via --preserve-env=<list>, which sudo-rs
and classic sudo (>=1.8.21, so Debian 10's 1.8.27) both honour, so this
needs no version gate. The list is cross-checked against every
RESTIC_/BORG_/KOPIA_ var the engine env scripts export.
The list lives in variables.sh with a literal fallback in runBackupOp,
because init.sh sources run_privileged.sh directly during install without
ever loading variables.sh — an unguarded empty list would silently
reproduce the same dropped-credential bug.
restoreFirstRunDiscover now goes through runBackupOp rather than issuing
its own sudo. It was the only backup-engine call bypassing that funnel,
which is why it missed this fix by construction; routing it back also
gives it the -H that keeps restic's cache under the backup user's HOME.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
get.docker.com/rootless aborts when ip_tables/ip6_tables aren't loaded.
Ubuntu 24.04 and 26.04 ship both modules but don't autoload them on a
fresh box, so rootless setup died there — and because the caller captures
its output into $result, the reason never reached the console or the
error report. The install continued, reported success, and printed
credentials for a WebUI that was never running.
initPrerequires now modprobes both modules and persists them to
/etc/modules-load.d/libreportal-rootless.conf for subsequent boots,
failing with an actionable message when the kernel genuinely lacks them
(container/VM kernels without netfilter).
installDockerRootless gets its own guard, since it also runs outside
init.sh via start_docker / rootless_start_setup. It only attempts
modprobe when it can — the de-sudoed manager has no modprobe in the
LP_SYSTEM allowlist, matching how ubuntu.sh handles sudo-apt — and
returns non-zero rather than proceeding into a failure whose message
would be swallowed. Already-loaded modules are a clean no-op, so the
normal post-install re-run path is unaffected.
Uninstall removes the drop-in alongside the sysctl ones, and it's listed
in the footprint summary.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`apt install p7zip*` only still resolves on Debian 13 / Ubuntu 24.04+
because the renamed `7zip` package happens to declare `Provides: p7zip`.
That is an alias we don't control, so pick the real package name against
the freshly-updated lists instead: `7zip` where it exists, `p7zip-full`
otherwise. Match on a real package stanza rather than apt-cache's exit
status, which returns 0 with empty output for provided-only names.
Also in this path:
- apt -> apt-get for the scripted calls, so the "apt does not have a
stable CLI interface" warning stops polluting the captured $result.
- drop a duplicated `pv` from the package list.
- move the package list below `apt-get update` so the 7-Zip probe reads
current lists.
Debian 10 sysctl check now reads OS_TYPE/OS_VERSION from detectOS rather
than shelling out to lsb_release, which minimal images don't ship. This
also stops a non-Debian release numbered "10" from matching.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
