read_tor_address("bitcoin-core") was resolving through tor_service_name to
the shared "bitcoin" alias, but enrollment (install.rs auto-enroll and the
tor.create-service RPC) always names HiddenServiceDir/tor-hostnames entries
using the raw package_id verbatim — never canonicalized. On a real node
that's hidden_service_bitcoin-core, which the aliased lookup never found,
so the per-app UI Tor badge stayed empty even after the previous commit
made bitcoin-core auto-enrollable.
Give bitcoin-core its own identity-mapped arm instead of folding it into
the legacy bitcoin/bitcoin-knots/bitcoind alias, and pin all three lookup
tables (known_service_port, is_protocol_service, tor_service_name) with
regression tests so this alias-drift class of bug can't recur silently.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WxfWiFfnBkdSxwKUuV2tNy
apps/bitcoin-core/manifest.yml uses id "bitcoin-core", but
known_service_port/is_protocol_service (tor/mod.rs) and
tor_service_name (docker_packages.rs) only matched "bitcoin" and
"bitcoin-knots", so the app silently never got auto-enrolled for a
P2P (8333) hidden service at install time, and the UI's Tor address
lookup for it always returned None.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WxfWiFfnBkdSxwKUuV2tNy
Confirmed live 2026-09-08 against three real Lightning payments to a
registered @minibits.cash address: POST /claim (the only claim source
claim_and_redeem checked) always returned an empty array, no matter
how long or how often it was polled. Independently queried
wss://relay.minibits.cash and found all three payments sitting there
as NIP-04-encrypted kind-4 DMs, #p-tagged to the wallet's own Nostr
pubkey and authored by the Minibits service key — that is the actual
delivery channel for a payment made to the address, and this module
never looked at it.
fetch_relay_dms queries CLAIM_RELAY_URLS (the service's own relay plus
two public fallbacks) for kind-4 events tagged to our pubkey, feeding
matching content into the existing pending_claims retry pipeline
unchanged. A new last_dm_seen_at watermark stops the same (immutable,
never-expiring) relay event from being re-fetched and re-attempted on
every poll. The REST /claim call stays in place alongside it in case
it serves some other payment path — this only adds the missing one.
fix(ecash): trim stray whitespace before parsing a cashuA/cashuB token
Once the relay fix above surfaced the three real payments, all three
failed to redeem with "Invalid base64 in cashuB token" — the decrypted
NIP-04 content had a trailing space after the base64 payload (Minibits'
own encoding), which every base64 alphabet in decode_token_base64
rejects outright. CashuToken::deserialize now trims the whole token
string before touching the "cashuA"/"cashuB" prefix or payload. This is
a general robustness fix, not just a Minibits workaround — the same
stray-whitespace failure could hit a hand-pasted token from a clipboard
copy just as easily.
Both fixes verified end-to-end against production: all three stuck
payments (20 + 5 + 20 = 45 sats) redeemed cleanly on the first poll
after deploying this build to archy-x250-pa3.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
Same class of bug as the Minibits address label
(settings.passwordNeedSpecial: "...(!@#$%^&* etc.)" — a bare @ vue-i18n
parses as linked-message syntax). This one is live in
ChangePasswordSection.vue's password-strength validator: typing a new
password with no special character throws this exact
SyntaxError the moment the message is rendered. Fixed the same way
({'@'} escaping).
Added locales/__tests__/i18nMessagesCompile.test.ts, which walks every
string in every locale file and asks the real vue-i18n compiler to
parse it — confirmed it fails on both bad strings before their fixes
and passes clean now, with no other landmines left in either locale
file. This closes the whole bug class rather than just these two
instances; a future bad interpolation string fails `npm test` instead
of only a live crash report.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
Root cause of "click Receive, click Ecash, the modal disappears" (in
both the browser and the Android companion's WebView, since both host
the same neode-ui bundle): vue-i18n treats a bare @ as the start of
"linked message" syntax. receiveBitcoin.lnAddressLabel ("Your
@minibits.cash address:") isn't valid linked-message syntax, so
*compiling* that message throws a SyntaxError the instant it's first
rendered — i.e. the moment wallet.ecash-lnaddress resolves and the
address section becomes visible. The uncaught render-function error
blanks the whole teleported modal, which is indistinguishable from it
just closing.
Confirmed with a real (non-mocked) Vue app + real vue-i18n compiler in
a headless Chromium — a Vitest run with `t` mocked to a no-op, which is
how the existing component test suite covers this file, cannot catch a
bad message string at all. Fixed by escaping the @ as {'@'} — the same
pattern the codebase already uses for settings.domainNamePlaceholder
("user{'@'}example.com"). Added a regression test using the real
vue-i18n instance instead of the mocked one; verified it fails on the
old string and passes on the fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
Operator report (2026-09-08): clicking the Ecash tab appeared to close
the whole Receive modal. Added a regression test simulating the exact
click, both for wallet.ecash-lnaddress succeeding and failing — the
tab switch alone never emits `close` or unmounts the dialog in either
case, so this isn't reproduced by a plain component-level click; the
investigation continues with the reporter for a browser-console repro.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
archy-x250-pa3's data volume filled to 100% (cuprate at 125G, since
removed) while a client had the ecash receive tab open. save_state's
write landed mid-truncate, leaving wallet/minibits.json at 0 bytes.
load_state then hard-failed every wallet.ecash-lnaddress call with
"EOF while parsing a value", surfaced in the UI as "Lightning address
unavailable" — permanently, since nothing ever cleared the bad file.
Registration is idempotent per pubkey (re-registering returns the same
lud16 Minibits already assigned), so there's no reason a corrupt local
mirror of that state should be fatal. load_state now treats an empty
or unparseable state file the same as a missing one — re-register and
recover the same address — instead of erroring. Manually cleared the
stuck file on archy-x250-pa3 as an immediate fix; this closes the gap
so it self-heals next time.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
A Minibits /claim response consumes the payment server-side the instant
it's returned — it can never be re-fetched. claim_and_redeem previously
decrypted/redeemed each claim inline and just warn!-logged any failure,
so a mint-unreachable blip, a stale cached server key, or an operator
who'd edited their accepted-mints list to drop the default mint (via
streaming.configure-mints) could make a real payment vanish with
nothing but a log line to show for it — claimed_count/received_sats
still came back as a clean 0, identical to "nothing arrived."
Now: every fetched claim is persisted to MinibitsState.pending_claims
before decrypt/redeem is attempted, survives failures across polls
instead of being dropped, and claim_and_redeem no longer bails out on a
fetch error without first retrying whatever was already pending.
ensure_mint_accepted self-heals the accepted-mints allow-list so the
Minibits mint (the address is inherently backed by it) can't be
excluded out from under a claim. ClaimOutcome gains failed_count,
threaded through wallet.ecash-lnaddress-claim and shown in
ReceiveBitcoinModal so a stuck claim is visible instead of silent.
Also fixes the server_nostur_pubkey field-name typo (no live state to
migrate — this feature hasn't shipped yet).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EawZPP9iidXj6Tvg3EpG3a
The wallet used Minibits only as a Cashu mint, so the node could hold and
swap ecash there but had no addressable name at it. This derives a LUD-16
Lightning address (name@minibits.cash) from the node's own ecash wallet and
surfaces it in the ecash Receive tab above the existing paste-token box.
Identity reuses the NUT-13 ecash phrase, so there is no second secret:
- seedHash = sha256(mnemonic.to_seed("")) — the exact hash the Minibits app
stores, so restoring the same phrase recovers the same address both ways;
- Nostr keys via NIP-06 at m/44'/1237'/0'/0/0 (nostr-sdk Keys::from_mnemonic,
pinned by a unit test against the NIP-06 vector so a bump cannot silently
move the derivation and orphan the profile).
Backend (wallet/minibits.rs) implements the verified live /v3 flow: NIP-42
challenge/verify -> JWT, idempotent /profile registration with collision
retry, and /claim polling that NIP-04-decrypts each token (service pubkey read
from the address's own LUD-16 metadata, constant fallback) and redeems it
through ecash::receive_token. Mainnet-only; state cached 0600 in
wallet/minibits.json.
New RPC: wallet.ecash-lnaddress (register-or-read, idempotent) and
wallet.ecash-lnaddress-claim (sweep Lightning payments into ecash). The modal
fetches the address on tab open, renders QR + copy, and sweeps claims while
open; a registration failure is non-fatal so paste-token still works.
Verified end-to-end against production: registered a disposable
@minibits.cash address, confirmed it resolves via /.well-known/lnurlp, and the
claim poll returns cleanly.
Archipelago never touched /etc/tollgate/identities.json — the "owner"
payout identity was whatever the router's TollGate install happened to
default to. Confirmed live against archy-x250-pa3: an unmodified upstream
placeholder (tollgate@minibits.cash), meaning 79% of every customer payment
would auto-payout to an address the operator never chose and doesn't
control.
Adds TollGateConfig.payout_address (opt-in — None leaves the router
untouched), config::apply_payout_identity() to merge it into the "owner"
entry of identities.json without disturbing the merchant keypair or the
other profit-share identities, an RPC param on openwrt.provision-tollgate,
and a status field + reconfigure-form input in the OpenWrt Gateway panel.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KTdMfVJChCwCCYF1ZTRQLc
sweep_once() has never actually swept anything: `tollgate wallet drain
cashu` (no flags) blocks on an interactive y/N confirmation that Router::run
can never answer over a non-PTY SSH exec (empty stdin -> EOF -> defaults to
N -> "Operation cancelled." with exit code 0), so the drain_code != 0 check
can't catch it and every tick silently no-ops.
The obvious fix isn't safe either: `--json` skips the prompt, but confirmed
live against archy-x250-pa3 that on a wallet.db with a stale duplicate
per-mint entry (trailing-slash leftover from before the mint_url fix), it
completes a real swap against the good entry, then aborts on the second
(empty, stale) entry and reports "success": false without ever printing or
persisting the resulting token anywhere. 50 sats went from spendable balance
to gone in that one call. Documented so nobody "fixes" this by wiring in
--json before upstream fixes the partial-failure data loss.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KTdMfVJChCwCCYF1ZTRQLc
Two bugs found live against archy-x250-pa3: TollGate-3458 (the upstream
tollgate-module-basic-go installer's own default AP, rebranded from
OpenWrt's factory default wireless.default_radioN sections) was left
bound to `network=lan` — wide open, unmetered, and sharing the router's
admin LAN — because install_ipk() runs the upstream package's own
uci-defaults scripts but nothing reconciled the AP they create with the
separate `tollgate` network/bridge/firewall this project's own
provision_ssid() sets up for the "archipelago" SSID. Fixed by folding any
default_radioN section left on `lan` onto the `tollgate` network right
after it's created.
Separately, a caller-supplied mint_url with a trailing slash
(https://mint.minibits.cash/Bitcoin/) got written byte-for-byte into
accepted_mints[0].url, which tollgate-wrt string-compares exactly against
a token's embedded (slash-less) mint URL — rejecting every otherwise-valid
token as an "untrusted mint". Fixed by trimming trailing slashes before
the value is used anywhere.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KTdMfVJChCwCCYF1ZTRQLc
Adds a verification checklist (service running, nodogsplash bound to
br-tollgate not br-lan via the rendered config not just UCI, LAN/SSH
untouched, mint probes succeeding) plus notes on the dev-build test-mint
injection and the default-route race between a router's LAN interface
and the node's other uplinks before the router's own WAN/WISP is live.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
Archipelago's Connect form only authenticates with an existing password;
it has no flow for setting one on a fresh, passwordless router. On the
node's kiosk display there's no visible tab bar, so Ctrl+T to open a new
tab to LuCI is the way to set it before Connect will work.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
The install code was hardcoded to the Oct 2025 v0.2.0 release —
nine releases behind. Its changelog covers exactly the failures hit
live against archy-x250-pa3: a mint with an empty/broken keyset
crash-looped tollgate-wrt forever (v0.5.0 adds "graceful degradation
when Cashu mints fail"), and the bundled captive-portal JS had zero
CBOR support, hard-rejecting the cashuB (NUT-00 V4) tokens modern
wallets like Minibits generate by default.
Also: v0.5.0 publishes native .apk packages for aarch64_cortex-a53
and x86_64. install_tollgate_apk_native now prefers those directly
(apk add handles deps/postinst/uci-defaults itself) instead of always
falling back to the manual ar/tar .ipk extraction dance, which only
exists because earlier releases had no native apk build at all.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
Observed live on archy-x250-pa3: right after WAN reconnects (fresh
boot or WAN reconfigure), the first Install attempt can fail with
"apk update failed ... router may have no internet access" purely
because the WiFi-uplink STA association hasn't finished yet — it's
not a real error, just retry a few seconds later. Also cross-referenced
the now-fixed /usr/bin/opkg hardcoding bug for anyone hitting it on an
older build.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
provisionTollgate/saveTollgateConfig/scanWifi/configureWan all fell
back to the Connect form's local refs (host/sshUser/sshPassword) when
connectedParams was null. Those refs only get populated if the form
was actually submitted this session — on a normal page load the
router reconnects via the server-persisted config instead, leaving
sshPassword at its default ''. Sending that as an explicit
(empty-but-present) ssh_password overrides the backend's saved-config
fallback, so every action auths with a blank password instead of the
real saved one.
Added authParams(): omit host/ssh_user/ssh_password entirely unless
connectedParams is actually set, same as the status poll already does.
Caught live: dropbear on archy-x250-pa3's router logged a single bad
password attempt at the exact moment "Install TollGate" was clicked,
sandwiched between periodic status-poll connections succeeding with
the real saved password.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
opkg_check() and every opkg/apk invocation hardcoded /usr/bin/opkg and
/usr/bin/apk. Official OpenWrt images don't all symlink /bin into
/usr/bin — the glinet_gl-mt3000 24.10.2 build keeps them as separate
real directories with opkg living in /bin — so the check silently
missed a perfectly normal install and TollGate provisioning failed
with "this router's firmware may not support package management".
Switched every call to resolve through the router's own $PATH
(command -v / bare opkg / apk) instead. Reproduced and fixed live
against archy-x250-pa3, 2026-09-05.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
Step 4 described a single "Provision TollGate" action that prompts for
price/step/mint upfront. The real UI (OpenWrtGateway.vue) doesn't work
that way: "Install TollGate" is a one-click action with no config form
that installs with defaults, and price/step/mint/enabled are only
editable afterward via a separate "Edit" panel. Caught while walking
through a live install.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
Worked example for the Beryl AX (GL-MT3000, mediatek/filogic) verified
against the OpenWrt wiki and firmware selector: exact sysupgrade image
filename, GL.iNet UI / LuCI flash path, post-flash SSH state, and the
U-Boot recovery procedure if the flash goes sideways.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
Walks a node operator through pairing an OpenWrt router over SSH,
running the WAN/WISP wizard, and provisioning TollGate pay-as-you-go
WiFi — plus an RPC/architecture reference for developers. Distills
the openwrt crate, RPC handlers, and Vue panel into user-facing steps
that didn't exist anywhere in docs/ before.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0176RpCxFNS9ZaSJjL72W9Z5
The restricted-RPC port (18090) was `auth: none`, which the app gate
treats as fully exempt — no onion, no takeover, LAN/Tailscale IP only.
Flip it to `auth: open`: the gate still binds the external addresses
and fronts a Tor onion for the port, just without a dashboard login
challenge, since Monero wallet clients (Feather, monero-wallet-rpc,
GUI) speak plain HTTP JSON-RPC and can't hold a session cookie.
P2P (18183) stays `none` — no reason to Tor-front raw gossip.
Regenerated releases/app-catalog.json (unsigned) to embed the updated
manifest; needs scripts/sign-catalog.sh before it takes effect on any
node, since origin (catalog) wins over disk for catalog-covered apps.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NZnsiMtyJxiJBuvv7yLPUF
The default manifest baked in the exact broken config found on an
affected fleet node: no fast_sync (defaults false, forcing full ring-sig/
RandomX verification on every block) and target_max_memory capped at
~2.8GiB, which starved cuprated's DB cache into constant eviction/flush
(595GB/24h of block I/O on a node just appending ~2MB blocks every 2
minutes). A reference node with fast_sync = true and an 8GiB cache ran
at 2.8% CPU at the same chain height and block rate.
Set fast_sync = true and target_max_memory = 8GiB to match the healthy
reference config, and raise resources.memory_limit from 4Gi to 10Gi so
the container still has headroom above the larger cache.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RR7jRaicvqsJaqQQ92jpPQ
v1.8.9's move to Router.SendPaymentV2 shipped without fee_limit_sat,
and the v2 route treats an ABSENT fee limit as zero allowed fees.
Every real route carries a routing fee (the 2-hop route here: 1.5
sats), so the pathfinder rejected them all and the wallet answered
"No route to the recipient" on EVERY send — all day, on healthy
channels with plenty of liquidity both ways.
The router debug log makes it unambiguous:
wallet payment (v1.8.9 backend): fee_limit=0 mSAT -> no route
same payment by hand (lncli --fee_limit=100): fee_limit=100000 mSAT -> settles in 0.65s
My earlier "pipeline verified" claim was wrong — the manual lncli
verification set a fee limit by hand and masked this exact bug. The
400k that succeeded this morning went through the pre-update backend
on the pre-update LND.
Payments now carry lncli's own default budget — the payment amount
(100%), preferring the payer-supplied amount for zero-value invoices
and the invoice's own amount otherwise, with a nominal floor so the
limit can never be zero. Unit-pinned so it cannot regress.
LND normally reconnects channel peers after a restart, but not reliably:
after long or repeated downtime (an app update, a node reboot,
reconciler churn) the peer link can stay down for hours while BOTH
endpoints keep the channel flagged disabled in the routing graph. The
node looks perfectly healthy, the wallet shows balance, and every
payment in either direction fails "no route to the recipient" —
observed live on framework-pt (2026-09-01): its only channel sat
disabled on both policy sides for ~17 hours after the LND 0.21.2
update, while shorty had 583k spendable and the user was told, by a
mis-mapped modal, that they had 'no payment channel'.
The channel graph is desired state — every open channel should have a
live peer connection. A daemon-side watchdog now enforces it:
- every 2 minutes, list channels + peers over LND REST
- for each channel whose remote peer is not connected, look the peer's
advertised addresses up in the public graph and dial one
- per-peer retries throttled to 10 minutes so an unreachable peer is
not hammered; 'already connected' counts as done; a peer with no
advertised address is logged once per pass (cannot be dialed)
- no-ops quietly on nodes without LND (missing macaroon) and while a
wallet is locked (503 body has no channels)
Unit tests pin the selection against the live REST shapes
(remote_pubkey in /v1/channels vs pub_key in /v1/peers).
v1.8.10 CHANGELOG + What's New entries staged so the next release run
is clean first time.
"LND thinks I do not have a channel" while the wallet showed plenty of
liquidity (framework-pt, 2026-09-01): the send gate sums outbound over
FULLY-OPEN channels only, which is correct — a just-opened channel
sits in LND's pending list until it has ~3 confirmations, and an
open channel can have all its balance on the far side — but the modal
then claimed the node had NO channel at all, in every one of those
states, and pointed the user at opening another one.
The gate already fetched the full channel list; it now records WHY
liquidity is zero and the modal says the truth per state:
- pending channels -> "your new channel is waiting for on-chain
confirmations, it unlocks automatically, nothing is needed from you"
(and no "Open a channel" button — that would send the user to fix
a problem they don't have, possibly opening a second channel)
- open channels, zero on the needed side -> "balance is on the far
side — you can receive but there's nothing to send right now"
- payment refused with a routing/liquidity error -> says so, instead
of claiming no channels
- only a genuinely channel-less node keeps the open-one guidance
Eleven unit tests pin the state machine, including the regression
case (pending-only -> 'pending', not 'none') and fail-open on RPC
errors.
The release gate requires the freshly-built bundle to embed the new
version, and the version reaches the bundle through the What's New
modal in AccountInfoSection — there was no v1.8.9-alpha block yet, so
create-release.sh correctly refused to ship a bundle that looked stale.
This adds the block (the user-facing summary of today's LND/HTTPS/
launcher/NPM/Portainer fixes) and carries the version bumps the
aborted run had already written (Cargo.toml, Cargo.lock, package.json,
package-lock), so the re-run starts from a clean tree.
Verified: npm run build now produces assets containing 1.8.9-alpha
(Settings chunk), i.e. the exact check the script runs passes.
Regenerated from the fixed apps/nginx-proxy-manager/manifest.yml (the
only semantic change vs the previous signed catalog) and signed with
the release-root key. Catalog-covered nodes pick this up on their next
hourly fetch and the NPM start/die loop ends: s6 gets its /etc/letsencrypt
mount back and the internal nginx can bind 80/443/81 again under
--cap-drop=ALL.
Root-cause table, fix inventory, regression-test inventory and the
deploy/live-verification checklist for today's fleet incident — written
as the working record while the fixes land, so the deploy + verify steps
can be checked off against real nodes rather than memory. CHANGELOG
carries the user-facing notes for the release these fixes ship in.
Three launcher/bridge defects combined to make HTTPS dashboards look
broken while HTTP ones worked:
1. portAuth() looked the launch port up under the name the user clicks
('mempool-web', 'lnd', 'bitcoin-knots'…), but the signed catalog
declares those ports under the manifest id that owns them
(archy-mempool-web, lnd-ui, bitcoin-ui). The lookup missed,
portIsGateFronted answered false, and an HTTPS dashboard handed app
frames http:// URLs — blocked as mixed content: mempool and IndeeHub
'did not connect', bitcoin knots/core opened http:// in a new tab.
Resolution now follows launch aliases, then a port-wide catalog scan
that only answers when every declarer of that port agrees (a port
any app publishes as plain HTTP is never upgraded to https).
2. The signed-catalog cache was only warmed by the Store/Discover
views, so a user who went straight to My Apps launched apps with an
empty cache. Warmed at dashboard mount now — fetchAppCatalog()
already memoizes with a 1h TTL.
3. The NIP-07 bridge compared event.origin for strict equality with the
recorded (http) app URL and replied to the recorded URL as the
postMessage targetOrigin — both break the moment a frame is scheme-
upgraded (cached HSTS did exactly that): every nostr request was
silently dropped and replies to the stale origin threw. The bridge
now matches host+port (scheme deliberately ignored) and always
replies to event.origin — the frame's real origin.
Unit tests cover alias resolution (incl. bitcoin-knots→8334→https),
the conservative port-scan, and scheme-agnostic sender matching.
Converting Nginx Proxy Manager to a platform manifest (fc68c5b6) dropped
two things its image hard-requires, and the result was an endless
start/die loop — shorty-s watched it restart 3,176 times:
1. /etc/letsencrypt mount: NPM's s6 'prepare' service refuses to boot
without it ('ERROR: /etc/letsencrypt is not mounted!'). Mounted from
the same persistent app directory as before
(/var/lib/archipelago/nginx-proxy-manager/letsencrypt), so existing
certificates are preserved — no data moves, no migration.
2. NET_BIND_SERVICE: NPM's internal nginx listens on 80, 443 AND 81,
and the orchestrator runs --cap-drop=ALL. The legacy podman-run path
defaulted to the full capability set (and the legacy repair path in
package/config.rs always listed it), which is why this only broke
once the manifest became the source of truth.
The signed catalog embeds manifests with origin-wins semantics, so the
catalog carries the fix for every catalog-covered node — regenerate it
here (plus the generated store/launcher-port artifacts, which also pick
up drift from bf6ef964's retired apps). Catalog re-signing follows the
usual ceremony.
The HTTPS server block sent Strict-Transport-Security:
max-age=31536000; includeSubDomains. Browsers that visited HTTPS once
cached the policy and then silently upgraded the still-open HTTP
dashboard's fetches and frames to https — a scheme change is
cross-origin, so every /rpc/v1 call died 'No Access-Control-Allow-
Origin header' while the node was perfectly healthy (framework-pt
2026-09-01: the 'Failed to fetch' storm, dashboard 'not responding',
every app frame mixed-content-blocked).
Plain HTTP is a supported access mode BY DESIGN on this platform: the
node's certificate is optional and self-signed (Settings → Node
certificate, /ca.crt flow), and setup-node-ca.sh deliberately keeps
port 80 serving for devices that haven't installed the CA. So:
- port 80 sends no HSTS at all (with the rationale inline)
- port 443 sends max-age=0, which ACTIVELY DELETES the policy already
cached by affected browsers — leaving it absent would have kept every
stranded browser broken for a year
tests/lifecycle/bats/nginx-hsts.bats pins all three properties at the
gate: no live policy on :80, max-age=0 (never 31536000) on :443, and
no long-lived pin anywhere in the deployed config.
ssmithxandClaude Sonnet 5