The six commits that built the live chat sync added no tests. Everything they
touch fails silently by nature — a filter that drops messages, a subscription
that stops being replayed, a group whose id never reaches the `#h` tag — so the
symptom is always "some messages didn't arrive", days later, on someone else's
phone. 46 tests, in four files.
**What is covered**
RelayPoolSubscriptionTest (13) — the pool's half of surviving a dropped socket.
A query is retained and replayed on reconnect; a closed one is forgotten and
stops the socket reconnecting for it; closing one of two leaves the other alone;
a negentropy exchange is never replayed (its rounds are stateful, so resuming
one reconciles against a conversation the relay is no longer having); an update
to a live subscription replaces what gets replayed, including when the send
itself fails; dropping a relay or closing the pool forgets what they carried;
replay is scoped to the relay that reconnected. Plus the semantic the whole
change rests on, asserted in both directions: a live subscription keeps
delivering after EOSE, a one-shot query still ends at it.
LiveSubscriptionReconcileTest (12) — the requirement this all exists for: the
group filter follows group membership with nobody calling a subscribe function.
Joining widens the filter *in place* rather than reopening (a reopen would drop
the live tail of every other group in that chunk); leaving drops one; leaving
everything closes the subscription; churn inside the debounce window collapses
to one update; a NIP-17 room never becomes a group subscription. Then the
collect loop: events stored against the relay they came from, an event after
EOSE still stored, a CLOSED reopened once the back-off elapses and not before,
and a rate-limited CLOSED waiting far longer — but still coming back.
Backgrounding closes and foregrounding rebuilds, reconnects, and queues the
catch-up.
LiveSubscriptionPlanTest (11) — the filter and planning rules, led by the one
most likely to be "tidied up" later: the gift wrap filter carries no `since`,
because NIP-59 randomizes created_at into the past and a `since` near the
present silently drops new messages.
RelayBackPressureTest (4) and ReconnectBackoffTest (6) — the two pure decisions.
Which CLOSED reasons mean "ease off", and the backoff arithmetic including the
exponent clamp: 2.0.pow(4000) is Infinity and Duration * Double throws on it, so
without it a socket failing long enough turned its reconnect loop into a crash
loop, at the point the network was least likely to recover unaided.
**Seams opened to get there**, each a readability win on its own terms:
- NostrSocketClientFactory becomes an interface with DefaultNostrSocketClientFactory
behind it, so the pool can be driven by a fake socket.
- RelayPool takes its CoroutineScope, so the replay a reconnect triggers can be
observed rather than raced.
- LiveSubscriptionManager depends on a new LiveSubscriptionTransport (4
methods) rather than RelaysSocketManager, which observes the active wallet in
its init and cannot be stood up in a test at all.
- Its pure planning helpers move to the companion as `internal`, and its
launches inherit the caller's dispatcher instead of pinning Dispatchers.IO.
SynchronizationViewModel already launches observe() on IO, so nothing moves —
but a coroutine that picks its own dispatcher cannot be driven by a test
scheduler.
- reconnectDelay is extracted to ReconnectBackoff.kt with jitter as a
parameter, so the arithmetic can be pinned without randomness.
- endsLiveSubscription names the live-subscription termination rule next to
isTerminalFor, which is the one-shot rule. Having both named makes the
difference between them reviewable rather than implicit.
kotlinx-coroutines-test is added to commonTest: the pool's bookkeeping is all
suspend functions and there is no runBlocking in a common source set.
The tests were checked by mutation, not just by passing — reintroducing a
`since`, making EOSE terminal, dropping the leftGroupAt filter and removing
retention from query() each produce failures.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
None of these are imported by any source file in composeApp, generated sources
included, and none is the kind of dependency that gets used without an import.
androidx.paging:paging-common no androidx.paging import anywhere; no DAO
androidx.paging:paging-compose returns PagingSource and nothing calls
collectAsLazyPagingItems, so the Room-Paging
integration that would need it is not in use
androidx.work:work-runtime-ktx no Worker, CoroutineWorker or WorkManager
reference, and no provider or initializer entry
in AndroidManifest.xml
okhttp3:okhttp-coroutines redundant rather than unused: quartz-android
1.14.0 depends on it and at 5.5.0, which was
already upgrading this declaration's 5.4.0. It
stays on the runtime classpath either way
com.ionspin.kotlin:bignum no com.ionspin import. Arrived in a10dc1a with
lightning-mobile support, not with
secp256k1-frost-kmp, though frost was the last
thing in the tree that used bignum at all
no.synth:kmp-zip no no.synth import
no.synth:kmp-zip-okio
The catalog entries and the pagingCommon, workRuntimeKtx and okhttp version refs
go with them, since each was referenced exactly once and nothing else in this
build points at them. lightning-kmp-app also declares work-runtime-ktx, but reads
it from its own catalog, so it is unaffected.
Deliberately kept:
app.cash.sqldelight looks unused by the same test -- no .kt file under
composeApp/src imports it -- but composeApp holds .sq schemas for three databases
(ChannelsDatabase, PaymentsDatabase, AppDatabase) and the generated sources import
it in 29 files. The runtime, the coroutines extensions and the platform drivers
all stay.
compose.desktop.currentOs and kotlinx-coroutines-swing in jvmMain are untouched,
but worth flagging: the `jvm()` target is commented out, so that source set and the
files under composeApp/src/jvmMain are not built. Whether they are dead or waiting
for the desktop target to come back is a decision about the target, not about a
dependency, so this commit leaves both alone.
Commented-out declarations elsewhere in the file -- room vs room3, kspIosX64,
kspJvm -- are left as the migration markers they are.
`:composeApp:assembleDebug` passes and the APK still packages libsecp256k1-jni.so
for all four ABIs.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Android Studio refused to sync the project after the lightning-kmp-app bump in 4e939c4:
Using multiple versions of the Android Gradle Plugin [9.1.1, 9.4.0] across Gradle builds is
not allowed.
Affected builds: [:, :lightning-kmp-app:lightning-kmp:bitcoin-kmp:secp256k1-kmp]
A composite build has to settle on a single AGP version across every build that applies the
plugin. Four builds in this tree do:
: 9.1.1 version catalog
:lightning-kmp-app 9.1.1 version catalog
:secp256k1-frost-kmp 9.1.1 version catalog
:lightning-kmp-app:lightning-kmp:bitcoin-kmp:secp256k1-kmp 9.4.0 hardcoded
lightning-kmp and bitcoin-kmp apply no android plugin at all, which is why the message names
only the two ends of the chain rather than everything between them.
secp256k1-kmp pins 9.4.0 twice -- a buildscript classpath entry, and a pluginManagement
resolutionStrategy that rewrites every requested com.android plugin to that version -- so it
cannot be talked out of it from here. The three catalogs move up to meet it. Pinning it down to
9.1.1 instead would mean editing a build four submodules deep and then committing in bitcoin-kmp
and lightning-kmp purely to carry the pointer, for no gain. Moving up is also the direction the
chain already set: bbba08b's f22c30a moved every build in it onto Gradle 9.7.1 for the same
reason, and 4e939c4 followed.
Submodule commits carried here:
lightning-kmp-app bbba08b -> ce1ed01 build: move to AGP 9.4.0, matching the version the
submodule chain pins
secp256k1-frost-kmp d3b294d -> 80a55b7 build: move to AGP 9.4.0 to match the rest of
mantra-kmp's composite
Both sit on branches (build/agp-9.4.0 and build/gate-ios-targets-on-macos) rather than their
masters, and neither is pushed yet, so a fresh clone cannot resolve these two pointers until
they are.
9.4.0 is a minor bump inside the AGP 9.x line, so the DSL is unchanged and every android.* flag
in gradle.properties -- newDsl, builtInKotlin, the r8 settings -- keeps its meaning.
`:composeApp:compileDebugKotlinAndroid` passes clean.
Note that this is only what the IDE needs to sync, not everything worth knowing about the
chain. bitcoin-kmp's settings.gradle.kts substitutes secp256k1-kmp, -jni-jvm and the three
-jni-jvm-{linux,darwin,mingw} coordinates, but not -jni-android. So the android app still
resolves secp256k1-kmp-jni-android from maven central at stock 0.24.0, without the
FROST/prefractal native modules the threshold branch exists to provide, and a call into those
from android would fail with UnsatisfiedLinkError. Fixing that needs a substitution rule in
bitcoin-kmp. Untouched here because it is a runtime concern, not a sync one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Add both libraries as git submodules and resolve them through Gradle
composite builds instead of remote publications, so local changes to
either library are picked up by the app build directly.
Submodules:
* secp256k1-frost-kmp (github.com/ac-cord-ac/secp256k1-frost-kmp) at
2478403 -- BIP-327 FROTH/FROST signing over secp256k1, KMP.
* lightning-kmp-app (github.com/kngako/lightning-kmp-app) at 6745d88 --
phoenix business logic on top of lightning-kmp-core. Both submodules
carry a local commit aligning them with this build's AGP 9.1.1 /
Kotlin 2.4.10 (AGP version must match across a composite build or the
android variants fail attribute matching).
settings.gradle.kts:
* includeBuild() both submodule checkouts.
* lightning-kmp-app's project stays named ':library' (compose-resources
derives its Res package from the project name), so an explicit
dependencySubstitution maps the coordinate the app declares,
fr.acinq.phoenix:lightning-kmp-app, onto that project.
* Drop the jitpack.io repository: it only served com.github.kngako,
which is no longer consumed as a binary.
composeApp/build.gradle.kts:
* commonMain gains ac.cord.auxiliary:library:1.0.0 (secp256k1-frost-kmp,
substituted by the composite build).
* commonMain replaces com.github.kngako.lightning-kmp-app:library (via
JitPack) with fr.acinq.phoenix:lightning-kmp-app:1.0.0 (substituted by
the composite build). lightning-kmp-core still comes from Maven
Central.
* Remove the RestoreJitPackClassifier component-metadata rule and the
javax.inject import: it only existed to repair the artifact
classifiers JitPack drops from the apple metadata variants, which no
longer applies.
gradle/libs.versions.toml:
* Remove the now-unused lightningKmpApp version and the
com.github.kngako lightning-kmp-app catalog entry.
Verified:
./gradlew :composeApp:compileCommonMainKotlinMetadata :composeApp:compileDebugKotlinAndroid
The shared lightning/phoenix logic now lives in kngako/lightning-kmp-app,
which publishes a single `:library` module. JitPack serves a repository's
submodules under <GROUP>.<ARTIFACT>, hence the
com.github.kngako.lightning-kmp-app:library coordinate. The repository is
content-filtered to com.github.kngako so it is not consulted for anything
else -- an unfiltered jitpack.io entry gets asked about every dependency
that misses in mavenCentral, and each miss is a remote round trip.
The version is a commit rather than master-SNAPSHOT. For a -SNAPSHOT
version JitPack advertises a unique-snapshot maven-metadata.xml
(timestamp=<sha>, buildNumber=1) while serving the files under their
literal -SNAPSHOT names, so gradle derives library-<target>-master-<sha>-1
and gets a 404 on every artifact. Pinning a commit sidesteps the snapshot
machinery entirely and is reproducible; it needs bumping when the fork
moves.
That leaves the classifier. JitPack rewrites the version inside a
published .module file and drops the classifier while doing so, so both
the sources and the host-specific metadata variants of each apple target
come back naming library-<target>-<ver>.jar -- a file that does not
exist, next to the -sources.jar and -metadata.jar that do. The klib and
the aar are named without a classifier and so survive the rewrite, which
is why the android compilation resolves this dependency perfectly well
and only the metadata compilations fail. Every shared ios source set
resolves through those, so a component metadata rule puts the -metadata
classifier back. It is scoped to the two apple modules and to their
metadata variant by name, so it cannot disturb the klib artifacts.
Verified: :composeApp:compileCommonMainKotlinMetadata and
:composeApp:compileDebugKotlinAndroid both pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
