axum does not set TCP_NODELAY. Right after AUTH the relay writes
AUTH_OK and immediately the first queued DELIVER; with Nagle enabled the
second write waited for the peer's delayed ACK, adding about 40 ms to
the path a woken-up app takes to fetch its messages.
After a network change (Wi-Fi to mobile data) the app reconnects right
away, while the relay only notices the old connection is dead once the
idle timeout expires. When that stale handler finally exited it removed
the device from the online map by address alone, which dropped the new,
healthy connection. From then on every message was stored as offline and
only a push wake-up could fetch it, even though the app was connected.
The online map is now its own type: a newly authenticated connection
takes over delivery for the address, and a closing connection removes
the entry only if it still points at itself. Tests cover the registry
and the full route after a stale connection ends.
The delivery-attempt cap reads and increments message_queue.attempt_count,
but the published init schema never created that column. On a fresh
database every queue flush failed, so queued messages were never handed
out after a reconnect.
The column is added in a new migration with IF NOT EXISTS, so installs
that already carry it from an older init schema migrate as a no-op.
The version was a hand-written constant that had to be kept in sync with
`version:` in pubspec.yaml. It was not: the constant still said 1.1.0 while
the shipped build was 1.1.2+22, so every abuse report carried a version that
no build ever had — and reports could not be traced back to a build.
AppInfo reads versionName / CFBundleShortVersionString, which Flutter derives
from the pubspec, so the two cannot drift apart again. Reports now carry the
build number as well; when the version is not known yet the field is omitted
rather than guessed.
Also adds the timestamp regression test that covers MessagingService.wireSentAt,
which shipped in 1.1.2 without its test.
Registration failures from the distributor were swallowed, so push looked
enabled while nothing arrived. Without wake-ups Android suspends the app
and messages only show up once the user opens it — which looks like broken
delivery even though the relay has them queued.
The failure is now remembered and surfaced in Settings, and the chat list
carries a dismissible line explaining the state.
A message sent into a dead socket ended up as failed — attachments even
stayed on "sending" forever — and only the user could send it again by
hand. Mobile NAT drops idle TCP silently, so neither side knew.
- pingInterval on the WebSocket, so the client notices a dead link itself
- outbox: after every reconnect, stuck messages are re-sent (batch of 30,
up to a day old); permanent errors are not retried
- sendAttachment gained the try/catch it was missing, otherwise the
attachment never reached the outbox
- de-duplicate by message id: a retry re-sends the SAME message, which
would otherwise show up twice on the other side
Messages arrived late or not at all, for two independent reasons.
Dead connections: neither side sent a ping, and after AUTH the server
waited for a frame without a timeout. Mobile NAT drops idle TCP silently,
so the client wrote into a void while the relay still held the device in
its online map — and therefore never sent a push wake-up either. The ping
is a protocol-level one rather than v1::Ping, because dart:io answers
those on its own; the fix works with clients already in the wild.
Poisoned queue: a client decrypts an envelope (ratchet advances), stores
it and acknowledges it. If that ACK is lost, a second decryption attempt
can only fail, so the envelope is never acknowledged — and was redelivered
on every connect until its 30-day TTL, without ever reaching the recipient.
attempt_count existed from the first migration but was never counted; it
now bounds redelivery.
Receipts reported DELIVERED_ONLINE whenever a write into the bounded mpsc
channel succeeded, which it does even for a long-dead socket. Online is now
claimed only for devices that recently sent something.
Push send failures were logged at debug while the relay runs at info, so a
rejected wake-up looked exactly like a working one.
Rate limiting and the per-IP WebSocket cap trusted X-Forwarded-For and
X-Real-Ip unconditionally. Onion traffic bypasses the reverse proxy, so a
client could set the header itself and slip past every limit. Forwarded
headers are now honoured only from peers inside SEPOT_TRUSTED_PROXIES;
everyone else is keyed by peer address, and onion clients share one bucket.
The queue flush also read the whole backlog into memory at once, and the
certificate authority kept the trust-root private key in the database even
though it is only ever needed offline.
Fixes reported by testers and found in an app review:
- First message right after install no longer fails. Before the app
finished registering with the relay, the outgoing message could crash on
a null delivery address. The profile field is now optional and the send
path throws a clear "still connecting, try again shortly" error while
registration completes in the background.
- The QR code and share link now appear only once the delivery address is
ready. Previously a link generated before the first connection lacked
routing, so the other side could not send the first message. The screen
now shows a connecting state and reactively swaps in the QR when ready.
- Incomplete contact links are rejected on import instead of failing at
first send, with a clear prompt to ask for a fresh link.
- Auto-lock no longer silences the app: after locking and unlocking,
messaging is restarted (it previously stayed dead until app restart).
- Group messages are marked failed on a non-per-member send error instead
of hanging in the "sending" state forever.
- Hardened the relay client against a rare duplicate connection when
toggling Tor or locking mid-connect, and fromShareString now rejects a
delivery address of the wrong length.
- Pin unifiedpush below 6.2.0 to drop the webcrypto dependency (its native
framework lacks LC_ENCRYPTION_INFO and blocked the iOS upload).
- Skip Android release lint (network dependency check hangs on the build
host and adds no value for a Flutter app).
Settings crashed with a TypeError: the profile card passed a
SepotFingerprint object where the localized string formatter expects
a String (hidden behind a `dynamic` parameter). Also fixes RangeError
in avatar initials (empty names, double spaces) and switches initials
to grapheme clusters so emoji don't get split.
New photo preview screen: a captured or picked photo no longer sends
immediately — it opens in a preview with freehand drawing (7 colours,
undo), draggable text overlay, 90° rotation and a caption field.
Nothing leaves the device until Send is tapped.
Incoming-content hardening: decryption and content processing are now
separated in the delivery path. Once an envelope is decrypted the
ratchet has advanced, so redelivery can never help — permanently
malformed content (FormatException/TypeError) is acked and dropped
instead of poisoning the queue on every reconnect. Attachment metadata
now validates base64 of the crypto fields up front, and a corrupt
inline thumbnail no longer breaks bubble rendering.
PNG without real transparency (screenshots, edited photos) is now
re-encoded as JPEG — several times smaller on the wire.
Tests: 116 (new widget tests for Settings and the photo preview,
AttachmentMeta wire validation).
Initial public release. Anonymous on-device identity (no phone number,
e-mail, or password), Signal-protocol cryptography (X3DH + PQXDH, Double
Ratchet, sealed sender) via libsignal, a metadata-minimizing relay,
group chats, disappearing messages, encrypted local storage, optional
Tor transport, and a UI localized into Czech, English, German, Russian,
and Italian.
Monorepo: Flutter client (app/), Rust relay + wire types (rust/),
protobuf definitions (proto/), self-host deploy (deploy/).
Licensed under AGPL-3.0 with an app-store distribution exception (§7).
Julius Joska