| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of the master fix (2d2f9856) + cleanup script to staging.
generate() unconditionally lPush'd the invoice id to `${aid}:invoices`,
but it's also called from update() (tip/webhook changes, repeated bolt12
offer creation) re-using the existing id — so each update re-appended the
same id and the user's :invoices accumulated duplicate entries, rendering
the same offer/invoice multiple times. Guard the push with LPOS so it only
adds the id the first time.
Also add scripts/dedup-invoices.ts to scrub already-accumulated dupes
(dry-run by default, --apply to write, idempotent).
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master 7b71c96b, refining the zombie-listener watchdog staging
adopted in 7c34b431. Treating "waitanyinvoice blocked > 5min while cl is
alive" as a stall false-fires during normal quiet periods (~26x in 6h on
master 2026-06-07) and needlessly recycles the cl listen socket. Instead,
when blocked past the threshold, probe cl on the MAIN socket with
waitanyinvoice(pay_index, 2s server-side timeout): an invoice returned
means a real backlog the wedged listener missed -> recycle; a timeout
(CLN 904) means the listener is just idle -> leave it alone.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master a664af99. A unix-domain RPC socket has no TCP keepalive, so
if cl goes unresponsive while the socket stays open (zombie/half-open) an
awaited call hangs forever — nothing throws, the reconnect-on-error path
never fires, and only an app restart recovers. Wrap short-lived calls in a
30s Promise.race timeout (matching lib/health.ts); a timeout rejects with
ETIMEDOUT, now part of isSocketDied, so the client is dropped and the next
call reconnects. Long-polls, in-flight payments (pay/xpay/sendpay/keysend)
and bolt12 round-trips are exempt to avoid leaked debits.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master 8cb18821 (refs #80), adapted to staging's nwc event loop.
(a) pay_invoice: sendLightning runs xpay to completion and finalize() sets
p.ref = preimage, so return that directly instead of polling listpays and
racing a 20s timeout that returned a misleading "Payment timed out" even
when the payment settled. Keep a longer listpays loop as a fallback; on
still-in-flight return an explicit "check with lookup_invoice".
(b) rate limiting: raised per-pubkey cap 5 -> 60/min, and over-limit
requests now get a NIP-47 RATE_LIMITED reply instead of a silent drop that
looked like a lost reply. Every request path now always sends a 23195
reply (success, error, or handler-returned-nothing) so a client never
hangs. Pulled the reply/encrypt logic into a reply() helper.
The notify default-on part of 8cb18821 is intentionally NOT ported: master
10284aa1 reverted it to opt-in, which is already staging's behavior
(notify = String(notify) === "true").
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master 6372d918, adapted to staging's nwc relay loop (no
heartbeat here). NWC clients fetch the kind 13194 info event before they
talk to the wallet; without it they show "no info event" and an unknown
balance. 13194 is replaceable and meant to persist, so instead of blindly
republishing on every connect, query the relay (ensureInfo) and publish
only the server pubkeys whose event is actually missing. Re-check hourly
so a relay restart that leaves our socket intact self-heals; interval is
cleared on close/reconnect. Verified [email protected] Relay exposes ws/
readyState, on("eose") and unsubscribe.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
| |
Port of master ce9616f6. pay_invoice derived the amount solely from the
decoded invoice, so an amountless bolt11 yielded amount = round(undefined/
1000) = NaN and sendLightning rejected it with "Invalid amount". NIP-47
lets the client specify the amount (msat) in the request; fall back to it
when the invoice carries none, with a clear error if still missing.
Fixes #83.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
| |
Port of master 0154bb06. generate() stores `created` in milliseconds, but
NWC make_invoice returned it directly as created_at and computed
expires_at = created_at + expiry. NWC timestamps are Unix SECONDS
(NIP-47), so clients parsed a ~1.78e12 expires_at that overflowed their
date types (ptcpay's .NET DateTimeOffset, issue #84). Convert created to
seconds. Fixes #84.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master 183a3cb0 + 3db24424. scripts/analytics.ts streams live db
(SCAN, never KEYS) and reports a trailing window: MAU/WAU, new users,
custody total + top holders, network volume by type, most-active/highest-
volume users, bolt12 recipients; --csv flag for month-over-month trends.
scripts/analytics-geo.ts reports top countries from the request log.
lib/register.ts stamps `created` and stores `country` on new users so
reports get signup timing/geography from the db directly.
Note: these scripts were written against the master stack; on staging
balances live in TigerBeetle, so custody/holder figures may need the
balance source adjusted from the Redis balance: keys.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Port of master 8aba0bc2, adapted to this stack's tbRefund. When xpay's
response shape is unexpected (or a zero-amount invoice's decoded
amount_msat is undefined) the computed p.fee becomes NaN, and the refund
delta crashes the balance write, leaving finalize partial (user loses up
to maxfee). Validate the computed fee is a finite integer; if not, keep
the originally-reserved maxfee (zero refund, no balance corruption) and
warn. Same guard on the refund delta before tbRefund.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master 51107ed4 + fe767002 + cb0f1962 (combined) to the staging
stack. Track an outcome string through every exit path, handle the
xpay failed_parts branch explicitly (verify with listpays then finalize/
reverse/leave-pending), and dump abnormal xpay/keysend response shapes.
Leak guard re-reads the db record before firing: if ref is persisted the
payment is settled regardless of which path threw (finalize may set ref
then throw on the later tbRefund), so only a genuinely missing ref logs
"LEAKED DEBIT". Uses warn (no warnThrottled on this stack) and keeps the
existing retry_for: 20.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
| |
Port of master 9142b4db. Out-of-band emergency stop: nobal.sh (or a human)
creates a sentinel file on the host, bind-mounted at /locks in the
container; debit() checks for the file on every external send and refuses
with "External withdrawals temporarily disabled". /locks/ALL.locked is a
global override. Lives outside the db so a db-write vector can't disable
it. Internal sends are exempt.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
| |
Port of master 43427afe. The blacklist (send freeze) only matched the
current username, so a compromised account could shake the freeze by
renaming itself. Match on either the lowercased username OR the immutable
uid; add the uid to the blacklist set to freeze durably.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Port of master ca28c19f to the staging stack. The getPayment() pre-check
in the lightning listener/replay isn't atomic, so concurrent re-credits of
the same settlement each ran the credit path. Claim each lightning/bolt12
settlement with an atomic SET NX on credited:<ref> (ref = preimage) and
bail if already claimed. Guard sits before tbCredit and the pointer writes
so a duplicate bails cleanly. Also disable the /replay route that triggered
it.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
|
| |
|
|
|
|
|
|
| |
Guard refreshArkWallet against running when config.ark.arkPrivateKey is
unset, and add a comment explaining the ensureListenerAlive watchdog's
zombie-socket recovery.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Port of master bd1ece3 to staging. When cl restarts (e.g. autoheal after a
healthcheck timeout), the app's waitanyinvoice long-poll can hang on the now-dead
listen socket: listenerActive stays true forever, nothing restarts it, and
incoming lightning payments silently stop crediting until a manual app restart.
The per-call RPC timeout can't catch this — waitanyinvoice is meant to block.
ensureListenerAlive (run every 120s) now also detects a stalled listener — active
but blocked in waitanyinvoice for >5min — probes cl via the main socket (getinfo),
and only if cl is actually alive treats the listen socket as a zombie: bumps a
listenerEpoch, calls lnListen.reset() (new: drops the proxy socket so the next
call reconnects), and re-arms. The epoch guard makes the wedged invocation bow out
when its waitanyinvoice finally rejects, so we never run two listeners or
double-credit.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
| |
When lq.getAddressInfo throws on a connection error (ECONNREFUSED /
ETIMEDOUT / "Unable to connect"), the generic catch was reporting
"unrecognized address" — misleading for both users (who think their
address is wrong) and operators (who can't tell user error from
infrastructure problems in the logs). Map connection errors to a
specific "liquid temporarily unavailable" message instead.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two related defenses around the CLN unix socket connection:
- Global net.Socket.prototype.connect patch: when Bun's
net.createConnection is called on a non-existent unix socket, the
ENOENT it throws bypasses all JS error handlers and crashes the
process. existsSync-check the path first; if missing, emit a normal
'error' event so the caller can handle it gracefully.
- lightningProxy: existsSync gate before constructing LightningClient,
plus a client.on("error") listener that clears the cached client on
socket-died errors instead of letting the unhandled event take down
the process.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
| |
- fail loudly when an NWC request references an app whose user record no
longer exists (migrated/deleted account); previously crashed inside
handle() on `user.id`
- filter `revertedDuplicate` payments out of list_transactions so
clients never see the doubled records that were later reverted
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Apr 29 double-credit prevention, ported from master 0e368dab and adapted
to staging's hono/tigerbeetle structure. Two parallel callers of /confirm
(walletnotify, catchUp, bulk sweeps) could each read p.confirmed=false
and each run tbConfirm — doubling the pending→balance transfer.
tbConfirm itself isn't idempotent on its own, so the race materializes
as a real double-credit even with TB.
- routes/payments.ts: SET NX EX 60 lock around the confirmation block
keyed by `confirmlock:${txid}:${vout}`. Concurrent callers short-circuit
cleanly. Released on success and on every early-return.
- lib/payments.ts credit() + routes/payments.ts /confirm: mirror payment
+ invoice records and the `payment:<txid>:<vout>` / `payment:<hash>`
pointer keys into arc as soon as they're written. gf() will then find
the prior credit via arc fallback even if the main-db copy is wiped
by a future data-loss event.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The catch blocks for ln.xpay and ln.keysend reversed the user's debit
whenever the RPC call threw, regardless of the actual payment state in
CLN. If the RPC errored transiently while the payment was still pending,
the user's balance got refunded — and then the lightning payment could
still complete, leaving the user with both the refund and the outgoing
sats.
Both call sites now mirror the safe check() logic:
const failed = !pays.length || pays.every((p) => p.status === "failed");
if (completed) finalize ...
else if (failed) reverse ...
// else (pending) leave the debit; check() reconciles when CLN settles.
If listpays itself errors, log via warn() and skip the reverse — the
debit stays in the pending set and check() will retry.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
| |
- Add lib/changeid.ts for in-process account rekeying
- Block sends to addresses in redis blocked_addresses set, auto-rekey on attempt
- Add created param to credit() for accurate replay timestamps
- Add paid_at to replay() for correct payment dates
- Add /lnurl/proxy endpoint for server-side LNURL callback fetching
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
|
| | |
|
| | |
|
| |
|
|
| |
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
- Validate account ownership (aid) before allowing ark send/sync/vault
operations and in debit/sendOnchain as defense-in-depth
- Harden PIN check to reject missing/empty pin when user has one set
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Rewrite setup.sh: auto-installs Docker, downloads regtest snapshot,
pulls images, starts all services. Idempotent, works on fresh Ubuntu/macOS.
- Create coinos-base image (196MB) shared by all containers: bitcoin,
elements, lightning, strfry, kvrocks, electrs, tigerbeetle, chopsticks.
Total pull ~490MB, down from ~4.3GB.
- Move all images to ghcr.io/coinos/* namespace.
- Drop nutshell (mint/wallet), postgres, mon from default dev compose.
- Comment out ark stack (optional), guard ark.ts against empty config.
- Add missing sampledata: lightningc, strfry, kvrocks, valkey configs.
- Strip strfry binary (145MB debug symbols removed).
- Use single CLN version (v25.12.1) for all 3 lightning nodes.
- Switch valkey to alpine variant (110MB → 41MB).
- Update config.ts.sample with correct service URLs.
- Regtest snapshot hosted on GitHub Releases (~2MB).
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
Forward kind 445 Nostr events through push notifications so the
client can process messages instantly without relying on WebSockets.
Includes a 3KB size guard for the ~4KB push payload limit.
Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
| |
|
|
|
|
|
|
| |
Server now tracks MLS groups (synced from clients) and subscribes
to kind 445 events on group relays, sending push to all members
except the sender. Registry persisted to Redis across restarts.
Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
| |
|
|
|
|
|
|
| |
The listener was only watching the local strfry relay, but MLS
messages go to external relays like relay.coinos.io. Subscribe
to all configured relays so gift wraps are seen.
Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
| |
|
|
| |
Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
| |
|
|
|
|
|
|
|
| |
- Add lib/mls.ts: maintains Redis-backed index of MLS users with
background backfill and live subscription for kind 443 events
- Add /mls/users endpoint serving the directory to the UI
- Fix dmNotifications.ts: add setVapidDetails call so push delivery works
Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
| | |
|
