The send loop was hardcoded to 960 samples (20ms/Opus24k), causing
DEGRADED (Opus 6k, 40ms) and CATASTROPHIC (Codec2 1200, 40ms) to
fail — the encoder needed 1920 samples but only got 960.
Changes:
- capture_buf, ring read threshold, and timestamp increment are now
computed from profile.frame_duration_ms (960 for 20ms, 1920 for 40ms)
- decode_buf sized to MAX_FRAME_SAMPLES (1920) to handle any incoming codec
- recv codec switch now uses correct QualityProfile per codec (was
inheriting original profile's frame_duration_ms, breaking cross-codec)
- added ComfortNoise guard on recv path
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The CallDecoder now inspects each incoming packet's codec_id and
automatically switches the audio decoder if it differs from the
current profile. This enables cross-codec interop where one client
sends Opus and the other sends Codec2 — previously the receiver
would try to decode with the wrong codec, producing garbled audio.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Enables debugging Codec2 by allowing forced codec selection from CLI.
Supports: good, degraded, catastrophic, codec2-3200, codec2-1200.
Frame size, timing, and jitter buffer are all adjusted dynamically
based on the selected profile.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Relay identity:
- Stored in ~/.wzp/relay-identity (hex-encoded 32-byte seed)
- Generated on first run, reused on restart
- Fingerprint stays consistent across relay restarts
Linux build script (scripts/build-linux-notify.sh):
- Fire and forget: Hetzner VM → build all binaries → upload to rustypaste → ntfy notify → destroy VM
- Builds: wzp-relay, wzp-client, wzp-client-audio, wzp-web, wzp-bench
- Packages as tar.gz, uploads to rustypaste
- --keep flag to preserve VM
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Relay recognizes SNI "ping" and returns immediately — no handshake,
no stream accept, no timeout error logs. Client closes after QUIC
connect for RTT measurement.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Ping was a static JNI method that loaded the .so before nativeInit,
crashing jemalloc. Now ping is an instance method on WzpEngine:
- Engine is created once (nativeInit), reused for both ping and call
- pingRelay() uses same tokio runtime pattern as startCall()
- Auto-pings all servers on app launch (after engine init)
- No process restart needed
- TOFU fingerprints saved on first successful ping
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Same fix as Android: Box::new([0i16; 16384]) allocates 32KB on the
stack before moving to heap. Use vec![].into_boxed_slice() for
direct heap allocation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The jni crate emits VERBOSE logs for every JNI method lookup (~10 lines
per call, 100+ calls/sec on audio threads). This floods logcat, consumes
CPU, and triggers system kills. Filter to only show INFO+ for our crates
and WARN+ for everything else.
Also fix build script: clean full Rust target to ensure libc++_shared.so
is always copied by cargo-ndk.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds nativeWriteAudioDirect / nativeReadAudioDirect JNI functions
that accept a DirectByteBuffer instead of ShortArray. The buffer's
native memory is accessed directly by Rust via pointer — no
GetShortArrayRegion / SetShortArrayRegion, no GC-managed array
copies on the audio hot path.
This fixes SIGBUS crashes on Android 16 where ART's concurrent
mark-compact GC crashes when flipping thread roots during JNI
array operations on MAX_PRIORITY audio threads.
Old ShortArray methods kept for backward compatibility.
AudioPipeline switched to use Direct variants.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- AudioRing: use vec![].into_boxed_slice() instead of Box::new([]) to
avoid 32KB stack allocation that crashes scudo on Android
- JNI bridge: wrap tracing_subscriber init in catch_unwind to survive
sharded_slab allocation failures on some devices
- Engine: per-step encode profiling (avg_agc_us, avg_opus_us, avg_fec_us,
avg_send_us) logged every 5 seconds in send stats
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds average microsecond timings for each encode step:
- avg_agc_us: AGC processing
- avg_opus_us: Opus encoding
- avg_fec_us: FEC encode + repair generation
- avg_send_us: QUIC send_media
- avg_total_us: sum of above
Logged every 5 seconds in send stats. Resets each interval.
Use to identify which step is bottlenecking the encode loop
on devices where fps drops below 50.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Same fix as Android (4af7c5f): writer never touches read_pos,
reader self-corrects when lapped. Power-of-2 capacity (16384),
bitmask indexing, overflow/underrun counters.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Rust tracing subscriber was never initialized — all info!/warn!/error!
calls in the engine went to /dev/null. This meant our send/recv health
logging was invisible and we couldn't confirm the congestion fix was
active.
Now initializes tracing-android layer on first nativeInit(), routing
all Rust logs to logcat under tag "wzp_android". Also expanded logcat
filter in DebugReporter to capture engine-level log lines.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
send_datagram() returns Err(Blocked) when the QUIC congestion window
is full. This is transient — the window reopens once ACKs arrive.
Previously, all send paths treated this as fatal (break/return),
which killed the send task and cascaded via tokio::select! to kill
the entire call.
Now: log warning, drop the packet, continue. Brief audio glitch
(20-100ms) instead of complete call death. FEC on the receiver
side recovers most dropped packets.
Fixed in:
- CLI run_live send task (continue + error counter)
- CLI run_file_mode send paths (2 locations)
- Desktop engine send task
Also hardened recv tasks: transient errors (non-closed/reset)
are survived instead of causing exit.
Matches the fix applied to Android client (engine.rs).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Root cause: send_media() returns Err(Blocked) when QUIC congestion
window is full. The send task treated ANY send error as fatal (break),
killing the entire call. Now send errors drop the packet and continue.
Also hardened recv task to survive transient errors and added health
logging (recv gap tracking, periodic stats) to both send and recv.
Relay: added comprehensive debug logging — recv gaps, lock contention,
forward latency, send errors — all per-participant with 5s stats.
Other changes:
- AEC toggle in Settings (persisted, applied on next call)
- Debug report: records call audio (WAV), RMS histogram (CSV), logcat,
stats. Emailed as zip via Android share intent after call ends.
- Replaced LinearProgressIndicator with Box (compose version compat)
- FileProvider for sharing debug zip attachments
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- New desktop/ directory with Tauri v2 + Vite + TypeScript
- Rust backend: CallEngine wrapping wzp-client audio + transport
- Web frontend: connect screen, in-call screen with participants,
mic/speaker mute, keyboard shortcuts (m/s/q)
- Dark theme UI, settings persistence via localStorage
- Platform-aware --os-aec: warns on Windows/Linux (not yet implemented)
- Workspace updated to include desktop/src-tauri
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mic mute: the send loop now zeros the capture buffer when muted instead
of relying on write_audio() to skip writes. Previously stale ring data
and AGC amplification of near-silence caused crackling artifacts.
AEC: attach Android's hardware AcousticEchoCanceler to the AudioRecord
session. Also attach NoiseSuppressor when available. Both are released
on capture stop.
Room UI: deduplicate participants by fingerprint so ghost entries from
stale relay state don't show duplicate names.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
UI: deduplicate room participants by fingerprint so ghost entries from
stale relay state don't show duplicates.
Engine: after select! ends, call close_now() + connection.closed() with
500ms timeout to wait for the relay to acknowledge the CONNECTION_CLOSE.
Previously the close frame was queued but the runtime died before quinn
could retransmit if the first packet was lost.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
shutdown_background() killed the tokio runtime before quinn could send the
CONNECTION_CLOSE frame on the wire, so the relay never knew the client left.
Now use shutdown_timeout(500ms) to give quinn time to flush the close frame,
matching the desktop client pattern (which uses 2s timeout).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace Mutex-based CPAL callbacks with atomic SPSC ring buffers
- Proper async send/recv loops (no block_on), 20ms playout tick
- Add signal task for RoomUpdate presence display
- Add --alias, --raw-room flags and key persistence (~/.wzp/identity)
- Add SetAlias signal variant and relay-side handling
- Graceful Ctrl+C shutdown with force-quit on second press
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
stop_call() now calls close_now() on the stored transport handle before
killing the tokio runtime. This sends a QUIC CONNECTION_CLOSE frame so
the relay's recv loop breaks immediately, triggering leave() + RoomUpdate
broadcast. Previously the runtime was killed first, so transport.close()
never ran and the relay kept stale participants until idle timeout.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Settings now uses draft state — changes only persist on explicit Save
- Back button discards unsaved changes
- Added applyServers() for batch server updates
- Added missing alias field to CallOffer in featherchat tests
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add SettingsScreen with identity (alias, key backup/restore), audio defaults,
server management, network prefs, and default room
- SettingsRepository persists all settings via SharedPreferences
- Auto-generate random display names on first launch (e.g. "Swift Wolf")
- Thread alias through CallOffer → relay handshake → RoomUpdate broadcast
- Derive caller fingerprint from identity key in relay handshake (fixes null
fingerprints when --auth-url is not set)
- Persist identity seed for stable fingerprints across reconnects
- Add alias field to SignalMessage::CallOffer (serde default for backward compat)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add RoomUpdate signal message to wzp-proto with participant count + list
- Add RoomParticipant struct (fingerprint + optional alias)
- Store fingerprint/alias in relay Participant struct
- Broadcast RoomUpdate to all room members on join and leave
- Add signal recv task in Android engine to handle RoomUpdate
- Surface room_participant_count + room_participants in CallStats JSON
- Show "X in room" with participant names in Android in-call UI
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Wire CallService foreground service for background calls (microphone type)
- Add Voice Volume + Mic Gain sliders (-20 to +20 dB) applied in Kotlin
- Connect AudioRouteManager for real speaker toggle via AudioManager
- Feed quinn QUIC RTT into PathMonitor, display Loss/RTT/Jitter from live data
- Nuclear teardown between calls — recreate engine + audio pipeline each call
- Fix re-entrant teardown loop from CallService notification callback
- Park audio threads as daemons to avoid libcrypto TLS destructor crash on exit
- Remove duplicate wakelocks from Activity (service owns them now)
- Strip AEC + denoise from capture path, keep AGC only (incremental approach)
- Fix .so copy target: libwzp_android.so not libwzp.so
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Wire AutoGainControl on both capture (mic → encode) and playout
(decode → speaker) paths to normalize volume levels
- Add server list with add/remove custom server dialog
- Add IPv4/IPv6 preference toggle for DNS resolution
- Resolve DNS hostnames to IP in Kotlin before passing to Rust engine
- Revert to IP addresses for default servers (DNS still broken on QUIC)
AGC confirmed working — voice levels noticeably improved in testing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- AudioPipeline: Kotlin AudioRecord/AudioTrack on JVM threads, PCM
shuttled to Rust via lock-free ring buffers + JNI
- FEC: RaptorQ fountain codes on encode (5 frames/block, 20% repair
ratio for GOOD profile), decoder feeds repair symbols for recovery
- Real audio level meter from mic RMS (replaces fake animation)
- Room name editable in UI (default: "android")
- Relay changed to pangolin.manko.yoga:4433
- Stats overlay shows FEC recovered count
- CallState now synced from polled stats (fixes "Connecting" stuck bug)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
pthread_create crashes on Android due to static bionic __init_tcb stubs
in the Rust std prebuilt rlibs. This is unfixable without rebuilding std.
Solution: run the entire call (QUIC connect, handshake, media send/recv)
on a single tokio current_thread runtime. The JNI startCall() now blocks,
so Kotlin dispatches it to Dispatchers.IO (JVM thread, not pthread).
Audio pipeline temporarily simplified to silence frames — will restore
once threading is solved (either via Java Thread or rebuilding std).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The getauxval override (dlsym wrapper) fixes SIGSEGV in
init_have_lse_atomics at library load time. The current_thread
tokio runtime avoids SEGV_ACCERR in pthread_create/__init_tcb.
Both fixes are required together.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Multi-thread tokio runtime crashes with SEGV_ACCERR in __init_tcb
during pthread_create on Android (static bionic stubs from CRT).
Switch to current_thread runtime which runs network I/O on the
calling thread without spawning additional OS threads.
Also: clean up build.rs — use only libc++_shared.so (dynamic),
remove getauxval_fix.c hack, remove static c++/c++abi linking.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
compiler-rt's init_have_lse_atomics calls getauxval(AT_HWCAP) at
library load time. The static getauxval from the CRT reads from
__libc_auxv which is NULL in shared libraries → SIGSEGV at 0x0.
Fix: compile getauxval_fix.c that provides a getauxval() which uses
dlsym(RTLD_DEFAULT) to find the real bionic getauxval at runtime.
Also switch to libc++_shared.so (bundled in APK) to avoid pulling
in static libc stubs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Set cpp_link_stdlib(None) to suppress cc crate's automatic linking
- Explicitly link both c++_static and c++abi with NDK sysroot search path
- Fixes RTTI vtable symbol (_ZTVN10__cxxabiv117__class_type_infoE) error
- Verified: only liblog.so remains as dynamic dependency
Closes#001
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previous fix linked c++_static but not c++abi. Android NDK splits the
static C++ runtime into two archives: libc++_static.a (STL) and
libc++abi.a (RTTI/exceptions). Without c++abi, dlopen fails on
_ZTVN10__cxxabiv117__class_type_infoE.
Now using cpp_link_stdlib(None) to suppress cc crate auto-linking, then
explicitly linking both c++_static and c++abi via cargo:rustc-link-lib.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The app crashed immediately when loading libwzp_android.so because the
cc crate's default dynamic linking produced a runtime dependency on
libc++_shared.so, which was never packaged into the APK.
Adding .cpp_link_stdlib(Some("c++_static")) to build.rs bakes the C++
runtime into libwzp_android.so directly, eliminating the missing .so.
See issues/001-libc++-shared-crash.md for full diagnosis and logcat trace.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
