Closes DEC-032. Restores the canonical Android ABI set
{arm64-v8a, armeabi-v7a, x86_64} per SDD-073 item 4 / SDD-118 item 3.
Root cause was the audiopus_sys + cmake-rs + NDK toolchain-file gap:
cargo-ndk 4.x sets ANDROID_ABI / ANDROID_PLATFORM as env vars per
invocation, but upstream cmake-rs 0.x does not forward them to the
child cmake invocation as -D variables, so armeabi-v7a and x86_64
configure steps fell through to the toolchain-file default and
failed to build.
Fix:
- Cargo.toml: add a workspace [patch.crates-io] stanza pinning the
cmake crate to fork pr2502/cmake-rs @ commit
bdad5edc569d82151922c5c6c4685b1563f12aa1 (branch android-build),
which carries cmake-rs PR #257
(https://github.com/rust-lang/cmake-rs/pull/257). The patch is a
9-line addition that forwards ANDROID_ABI and ANDROID_PLATFORM
from the env to the child cmake as -D variables.
- Cargo.lock: regenerated by 'cargo update -p cmake'; the lone
cmake entry now points at the fork rev.
- apps/chanora_flutter/android/app/build.gradle.kts: restore
abiFilters to {arm64-v8a, armeabi-v7a, x86_64}; remove the
TODO(x86_64/armv7 follow-up) comment.
- docs/governance/product-decision-register.md: mark DEC-032 as
Resolved (2026-05-18) with the resolution mechanism, update the
§3 / §7 rows, and append a 0.9.8.1 change-history entry.
Verification (host: Linux):
cargo update -p cmake -> pulled fork rev
cargo check --workspace --all-targets -> PASS
cargo test --workspace -> PASS (no regressions)
cargo ndk --platform 28 -t arm64-v8a build -p chanora_bridge -> PASS
cargo ndk --platform 28 -t armeabi-v7a build -p chanora_bridge -> PASS
cargo ndk --platform 28 -t x86_64 build -p chanora_bridge -> PASS
Upstream tracking: re-evaluate the [patch.crates-io] override once
cmake-rs PR #257 merges and a fresh cmake release lands on
crates.io; at that point switch to a plain dep bump and remove the
override.
Author SysRS-310 ratifying the macOS minimum runtime baseline at 10.15
(Catalina) at the SysRS layer, parallel to SysRS-286 (iOS 13.0) and
SysRS-288 (Android API 28).
Closes the Wave 1.5 traceability-audit deferred-but-optional follow-up.
ID allocation: SysRS-310 (not SysRS-290, which is already allocated to
MVP single-active-server-connection scope); monotonic numbering
preserved.
Cross-references SAD-087, SDD-119, SysRS-286, SysRS-288. Verification:
Review + Platform Test on a macOS 10.15 system. Raising the baseline
(e.g., to 11.0 / Big Sur) requires a DEC entry.
Trace: SysRS v0.9.11, SAD-087, SDD-119.
Resolve the macOS half of the SDD-119 item 3 single-source-of-truth follow-up. The chanora_bridge cdylib macOS deployment-target floor ('10.15') was previously hard-coded at 7 sites across Podfile and chanora_bridge.podspec; this commit collapses them to a single Ruby constant declaration in a new SoT file.
Selected Option B (Ruby constant) over Option A (.xcconfig — rejected because the podspec prepare_command runs before any xcconfig is applied) and Option C (versioned text file — rejected as overkill given both consumers are already Ruby). Realizes SAD-087(a)'s 'single macOS build-configuration location' mandate.
Out of scope: the pbxproj 'MACOSX_DEPLOYMENT_TARGET = 10.15' lines (565/666/717) belong to the PBXProject default config and are independently overridden to '11.0' at the Runner PBXNativeTarget level; they are the Runner app's floor, not the bridge cdylib's floor. The iOS half (IPHONEOS_DEPLOYMENT_TARGET=13.0) remains an open follow-up.
Files: new apps/chanora_flutter/macos/macos_deployment_target.rb (MACOS_BRIDGE_DEPLOYMENT_TARGET = '10.15'.freeze); Podfile + chanora_bridge.podspec require_relative the constant and consume it at 7 sites; docs/architecture/sdd.md SDD-119 item 3 rewritten + Notes bullet updated + v0.9.17 changelog entry.
Canonical baseline lives at crates/chanora_audio/benches/baselines/x86_64-unknown-linux-gnu.json
and is updated only via the bench-baseline-update.yml workflow.
SDD-120 amendment v0.9.15 → v0.9.16 reflecting commit 3a7750a discovery.
The two post-processor tools (emit_baseline, compare_baseline) were originally
specified in SDD-120 §2 as residing in crates/chanora_audio/benches/, with
[[bin]] declarations in Cargo.toml. The implementation discovered that Cargo's
dependency resolver only routes [dev-dependencies] to [[test]], [[bench]], and
[[example]] targets — NOT to [[bin]] targets under src/bin/ or to ad-hoc paths.
Routing serde_json (required by both binaries) as a regular [dependencies] entry
to support [[bin]] placement would force it into the production cdylib build,
contradicting SDD-120 §10 release-artifact isolation: "no production telemetry
export" and "no benchmark surface in shipping artifacts." Confirmed by the user's
question about whether benchmark crates ship in the release artifact (they must
not).
The correct Cargo-idiomatic placement is examples/. Cargo auto-discovers files
under examples/ as example targets; they receive [dev-dependencies] routing; and
they are excluded from cargo build --release and from flutter build --release
artifacts by Cargo design.
Amendment scope (SDD-120 only):
- §1 item 4: [[bin]] → [[example]]; cargo run --bin → cargo run --example.
- §2: added item 5 with the rationale clause + Cargo dependency resolver
explanation.
- §5 item 1: path benches/ → examples/; invocation flag --bin → --example.
- §5 item 4: dev-dependencies routing clause updated.
- §6 step 6/8: --bin → --example.
- §7 step 5: --bin → --example.
- §10 item 1: release-artifact-isolation bullet extended to cite the
[dev-dependencies] Cargo design mechanism that enforces it.
- §11 verification matrix unchanged.
- "Allocated to" line + "Software units" list updated to reference examples/
paths.
No semantic change to SDD-120: same harness, same metrics, same workflows, same
out-of-scope deferrals. The implementation at 3a7750a already lives at the
corrected paths; this amendment brings the SDD text into agreement with the
code.
Implementation of SDD-120 §1-§8:
Bench harness (crates/chanora_audio/benches/):
- common.rs: deterministic synthetic audio (440 Hz sine, no RNG).
- realtime_capture.rs: bench_capture_alloc_count (dhat) +
bench_capture_callback_wall_clock (criterion).
- opus_codec.rs: bench_opus_encode_latency + bench_opus_decode_latency
(direct audiopus, not AudioHandler — SDD-120 §3 item 4).
- resampler.rs: bench_resampler_throughput across 44.1->48 /
16->48 / 48->48 passthrough.
CI tooling (crates/chanora_audio/examples/):
- emit_baseline.rs: aggregates criterion estimates.json outputs
into the SRS-217 baseline schema.
- compare_baseline.rs: applies SRS-219 tolerance, renders markdown
table with 🟢/🟡/🔴 markers + yellow simpler-form realization per
SDD-120 §8.
Deviation from SDD-120 §2 / §5 / §7 placement: these tools live
under examples/, not benches/ or src/bin/. Rationale: they must
consume serde_json (a dev-only dep — production builds must not
pull it). Cargo only resolves dev-dependencies for [[test]],
[[bench]], and [[example]] targets; [[bin]] targets under
src/bin/ see only regular [dependencies]. examples/ keeps the
binaries out of the production dep tree while still giving them
cargo run --example invocation. An SDD-120 amendment should
reflect this.
Workflows (.github/workflows/):
- bench-advisory.yml: PR + push triggers; runs benches; posts a
sticky PR comment via actions/github-script@v7; job status is
always success (SRS-218 clause 4 — non-blocking).
- bench-baseline-update.yml: workflow_dispatch only; runs benches;
opens PR via peter-evans/create-pull-request@v6 (sole writer of
the SAD-089 baseline JSON).
Cargo.toml additions ([dev-dependencies] only — verified excluded
from --release builds): criterion 0.5, dhat 0.3, serde_json 1.
Source-code seam: minimal pub-but-#[doc(hidden)] bench_seam module
in chanora_audio (engine.rs + lib.rs re-export) so the criterion
bench harness can construct a CaptureState and drive
CaptureState::ingest without re-implementing the engine (SDD-120
§3). Non-iOS targets only — CaptureState itself is iOS-gated.
Initial baseline seed: crates/chanora_audio/benches/baselines/
x86_64-unknown-linux-gnu.json = {}. compare_baseline handles the
missing-baseline case gracefully and emits a 'no red markers'
report; the first manual dispatch of bench-baseline-update.yml
after merge establishes the real values.
Out of scope per SDD-120 §10: production telemetry export,
build-failing hard CI gate, multi-host benchmarking, IDE
integration, Dart-side bridge round-trip bench.
Verification:
- cargo check --workspace --all-targets: PASS.
- cargo bench --bench realtime_capture --no-run: PASS.
- cargo bench --bench opus_codec --no-run: PASS.
- cargo bench --bench resampler --no-run: PASS.
- cargo build --example emit_baseline --example compare_baseline
-p chanora_audio: PASS.
- cargo test --workspace: 106 passed, 0 failed, 3 ignored — no
regression from prior count.
The capture cpal callback (CaptureState::ingest) ran two heap
allocations per callback on the realtime audio thread:
1. engine.rs:1196-1202 — fresh `mono: Vec<f32>` for the downmix
output, once per cpal callback (50–100 Hz).
2. engine.rs:1217-1218 — `pcm_accum.drain(..FRAME_SAMPLES).collect()`
building a fresh Vec<f32> of 960 samples per Opus frame.
Both sites mirror the pattern already fixed for the output side at
engine.rs:1389-1397, where allocating per callback on glibc malloc
was correlated with user-perceptible audio popping. The output-side
fix replaced the per-callback allocation with a pre-allocated
`scratch` Vec that is cleared and resized in place; this commit
applies the same template to the capture side.
Changes:
- Add `mono_scratch: Vec<f32>` and `frame_scratch: Vec<f32>` to
CaptureState. Initialised with Vec::with_capacity(4096) and
Vec::with_capacity(FRAME_SAMPLES=960) respectively in
CaptureState::new.
- Replace the downmix Vec construction with in-place push into
`self.mono_scratch`; `clear()` retains capacity across callbacks.
- Replace the drain().collect() with `self.frame_scratch.extend(
self.pcm_accum.drain(..FRAME_SAMPLES))`; same capacity-retention.
- The resampler call uses std::mem::take to swap the scratch buffer
out for the duration of the &mut self call, then moves it back —
the backing allocation is preserved across callbacks.
Algorithm semantics are unchanged: same downmix arithmetic, same
clamp loop, same Opus encode call sequence. Only the storage
strategy differs.
Out of scope (intentionally not touched):
- Android audio path (android_voice_unit.rs, mobile_voice_backend.rs):
researcher constraint C-4 — the Android cpal data path is mid-
migration and being replaced.
- Output callback at engine.rs:1409+: the only obvious per-callback
allocation there (`scratch`) was already fixed; a fuller audit
is a separate scope decision.
- The `scratch` buffer at engine.rs:1389-1397 — already correct.
Verification:
- cargo check --workspace --all-targets: passes.
- cargo test --workspace: 106 passed / 0 failed / 3 ignored.
- cargo clippy --workspace --all-targets: no new lints introduced;
the one warning inside the edited region (clamp-like pattern at
line 1266) was pre-existing on the copied clamp loop.
Dart consumer for the Android permission state pipeline. New
AndroidPermissionsService listens on the app.chanora/android_permissions
MethodChannel and exposes a ValueListenable for the UI. The voice-join
flow in main.dart calls ensureRecordAudio() before rust.voiceJoin and
clamps to listen-only via setHardMute on denial. A non-modal banner
above the VoiceBar surfaces the Grant / Open Settings action depending
on whether the state is Denied or PermanentlyDenied. On non-Android
hosts the service short-circuits to granted; the banner is never built.
Also adds the BackIntentService Dart consumer (back_intent_policy +
back_intent_service) which the Kotlin BackIntentBridge invokes via
MethodChannel for deterministic route-pop ordering.
Trace: SDD-028, SDD-106, SRS-163, SRS-209.
Android P0 platform shell:
- ChanoraApplication: early System.loadLibrary("c++_shared") +
System.loadLibrary("chanora_bridge") so JNI is hot before
MainActivity.onCreate.
- MainActivity: configureFlutterEngine + onResume/onDestroy wiring
for BackIntentBridge and AndroidPermissionRequester; publishes
permission-state changes through both the MethodChannel (Dart UI)
and the JNI hook (Rust audio engine).
- AndroidVoiceForegroundService: microphone-type foreground service
with notification channel chanora.voice.session per SDD-107.
- AndroidPermissionRequester: RECORD_AUDIO state machine with
persisted "has-ever-requested" flag so PermanentlyDenied is
correctly distinguished from never-asked across cold launches.
- BackIntentBridge: API 33+ OnBackInvokedCallback + pre-33
OnBackPressedDispatcher with deterministic Dart-side policy.
- MethodChannels: centralized constants for app.chanora/*.
- build.gradle.kts: SDD-118 Gradle automation that auto-builds the
Rust cdylib via cargo-ndk with per-ABI Exec tasks, minimal-env
isolation, CMAKE_TOOLCHAIN_FILE pinning, libc++_shared.so staging,
release-inspection assertion. abiFilters temporarily reduced to
arm64-v8a only per DEC-032 (multi-ABI restoration pending).
- AndroidManifest.xml: INTERNET, RECORD_AUDIO, FOREGROUND_SERVICE,
FOREGROUND_SERVICE_MICROPHONE, POST_NOTIFICATIONS,
MODIFY_AUDIO_SETTINGS, BLUETOOTH_CONNECT permissions; service
declaration with foregroundServiceType=microphone.
- proguard-rules.pro: keep rules for JNI native methods + Flutter
plugin entry points + FRB bindings.
Trace: SDD-073, SDD-105, SDD-106, SDD-107, SDD-108, SDD-110, SDD-118,
SRS-111, SRS-119, SRS-163, SRS-187, SRS-209, SRS-215.
Per SDD-106 §5 add BridgeEvent::PermissionState{permission, state}
with the PermissionStateKind enum (Granted, Denied, PermanentlyDenied,
Unknown). The Kotlin side publishes mid-session permission changes
through a new JNI entry point Java_app_chanora_chanora_1flutter
_MainActivity_publishPermissionState routed by the new
permission_jni.rs module; the Rust audio engine subscribes and
authoritatively clamps the transmit gate (see SDD-106 §6).
Adds crates/chanora_bridge/build.rs to emit
cargo:rustc-link-lib=dylib=c++_shared on Android so libchanora_bridge
.so carries DT_NEEDED libc++_shared.so; this is required by Android
API 24+ per-library linker namespaces to resolve __cxa_pure_virtual
and friends at System.loadLibrary time.
Includes the FRB-regenerated Dart counterparts so each commit is
independently buildable.
Trace: SDD-105, SDD-106 §5, SDD-118 item 6 (extended).
Per SDD-106 §6 add a permission-state clamp to TransmitModeSelector.
When RECORD_AUDIO is Denied or PermanentlyDenied the transmit gate
is forced false regardless of PTT or voice-activity state; on
Granted the clamp releases and normal transmit decisions resume.
The clamp takes precedence over PTT and hard_mute in the decision
ordering documented inline.
Three new tests cover the clamp behavior, the release-on-grant
transition, and the non-RECORD_AUDIO ignore path.
Trace: SDD-106 §6, SRS-209.
Replace the prior one-shot android_engage_voice_communication call
with a ModeStack-mediated acquire/release pair. AudioEngine snapshots
the system audio mode on first acquire via android_get_audio_mode()
and restores it on last release via android_set_audio_mode(prior).
MODE_IN_COMMUNICATION (3) is engaged across the voice-session lifetime
per SDD-108.
Includes the Android AudioManager getMode/setMode JNI helpers
(placed in chanora_audio::engine alongside the existing JNI surface)
and the small ptt.rs touch needed for the SDD-108 ID-tag on the
existing tests.
Trace: SDD-108, SDD-115.
Introduce ModeStack, a pure-Rust refcount-composable wrapper for
Android audio-mode acquire/release with prior-mode snapshot. Per
SDD-108 §1/§2 the engine snapshots the system audio mode on first
acquire and restores it on last release; composed acquires are
no-ops while the mode is held.
ModeStack is panic-free; release-on-zero returns AlreadyReleased
rather than panicking. Six SWE4-UV-045-tagged unit tests cover the
acquire/release semantics on the host target.
Trace: SDD-108, SWE4-UV-045.
Debug and Profile had ENABLE_OUTGOING_NETWORK_CONNECTIONS = NO which
blocked outbound TCP. All three configs now have:
- ENABLE_HARDENED_RUNTIME = YES
- ENABLE_OUTGOING_NETWORK_CONNECTIONS = YES
This was the root cause of the 'Operation not permitted' connection
failure on macOS Sequoia.
macOS:
- Transparent title bar with hidden title, full-size content view
- macOS: inline Row header (no AppBar) with 56px traffic-light pad
- Other platforms: standard Material AppBar unchanged
- App name 'Chanora' in CFBundleName/CFBundleDisplayName (iOS + macOS)
- NSLocalNetworkUsageDescription added to both platforms
Linux:
- tools/build-linux.sh: builds Rust .so + Flutter bundle + tarball
- Verifies GTK3, libopus dev headers, Rust target
- Copies libchanora_bridge.so into bundle/lib/
Required for local network privacy prompt on macOS 15+ and iOS 14+.
App appears in System Settings → Local Network after connecting to a
LAN server. Internet-hosted servers only need network.client entitlement.
When macOS denies network access (PermissionDenied), show a localized
dialog explaining how to grant permission in System Settings, with
an 'Open System Settings' button that opens directly to the Local
Network privacy pane.
Also adds author info (Edison Jwa) to About dialog and moves
diagnostics button from AppBar into About dialog.
The ring-buffer architecture (rc.8+73..+74) was making playback
strictly worse. Diagnostic data at +74 conclusively showed:
* Producer task ran perfectly at 50 Hz (250 ticks per 5 s).
* AudioHandler returned silence on 65-84% of fill_buffer calls
even when window_peak_f32 reached 0.98 (full-scale audio).
* Ring buffer never accumulated beyond 30 ms because consumer
(VPIO render callback at 43.5 Hz, ~1440 samples per call)
drained samples faster than the 50 Hz producer could push
them, in net effect.
The producer drained AudioHandler at 50 Hz \u2014 slightly faster
than iOS VPIO actually consumes audio. Each fill_buffer call
asked for 20 ms but adjacent Opus packets hadn't arrived yet, so
fill_buffer returned mostly silence. Linux/SDL's same pattern
works because SDL calls fill_buffer at EXACTLY the device
callback rate (50 Hz = 20 ms per buffer); the rates match.
Fix: revert to direct fill_buffer call from the render callback
(the SDL pattern in tsclientlib's own reference example at
tsclientlib/examples/audio_utils/ts_to_audio.rs). The render
callback now:
1. Resizes scratch_stereo Vec to 2 * num_frames f32 if needed
2. Zeros the live slice (fill_buffer is additive, not clearing)
3. Locks AudioHandler, calls fill_buffer(scratch_stereo)
4. Downmixes L+R -> mono i16 with master gain into out[]
5. Applies output_muted bypass
6. Tracks peak_out + audio/silence ratios for diagnostic
The closure owns scratch_stereo across callbacks for stable
allocation. Same memory model as Linux/SDL.
Removed:
* tokio::spawn producer task
* rtrb dep + RingBuffer<i16> + Producer/Consumer split
* tokio::sync::oneshot shutdown channel
* producer_shutdown_tx field on IosVoiceUnit struct
* RING_BUFFER_SAMPLES / PRODUCER_TICK_MS constants
* Producer-side diagnostic counters
Diagnostic kept: cb / num_frames / frames_changes /
callbacks_with_audio / callbacks_with_silence / peak_out_i16 /
gain. Logged every 100 callbacks.
The choppy / clicks symptom is independent of the buffer
architecture \u2014 it's whatever AudioHandler is doing on iOS
that's different from Linux. Next investigation step is to
either (a) switch from VPIO to RemoteIO unit (lose Apple's
voice processing entirely), or (b) understand why AudioHandler
returns silence so often on iOS-arrival packet timing patterns.
Build counter 74 -> 75.
External reviewer correctly identified that the +73 ring-buffer
commit didn't fix the symptom but the architecture is still
right. We need to distinguish two possible causes:
(a) Producer task isn't running (or running too rarely) so
ring stays underfilled.
(b) Producer IS running but fill_buffer returns zeros most of
the time (AudioHandler stuck in buffering_samples state
or no packets reaching it).
The +73 render-side diagnostic was insufficient: we logged
underruns + peak_out_i16 but not what the producer was
actually pushing. This commit adds producer-side metrics
rolled up every 5 s (250 ticks at 20 ms):
Producer task:
producer_ticks : timer firings (= ~250 per 5 s window;
fewer = tokio scheduler stalled)
produced_chunks : pushes into ring (= ticks - drops)
fill_buffer_calls : AudioHandler queries
fill_buffer_zero_returns : ticks where scratch came back all
zeros (no decoded content to play)
ring_full_drops : ticks where ring was full and we
skipped the push
ring_min/max_samples : depth envelope across window
ring_min/max_ms : same in milliseconds
window_peak_f32 : max scratch sample across window
window_rms_f32 : RMS of all scratch samples across
window
Render callback (per-100-callback as before, plus new fields):
ring_avail_before : Consumer::slots() before this read
(= how many samples were sitting in
the ring at callback entry)
read_frames : samples successfully popped
zero_filled : samples zero-filled because ring
was empty (= num_frames - read_frames)
underruns / underrun_samples / peak_out_i16 / clip_count_i16
: as before
Reading the next iteration's log:
If producer_ticks << 250 per 5 s window:
tokio scheduler isn't running the task fast enough.
Move producer to its own dedicated runtime, or use
std::thread + std::sync::mpsc + std::thread::sleep
instead of tokio.
If producer_ticks ~= 250 AND fill_buffer_zero_returns is
high (most ticks return silence):
AudioHandler isn't decoding packets fast enough OR is
stuck buffering. Bug is upstream in protocol layer
packet delivery or AudioHandler's jitter state machine.
The ring buffer architecture cannot fix this.
If producer_ticks ~= 250 AND fill_buffer_zero_returns is
low AND ring_min_ms stays >100ms AND underruns are low BUT
consumer's peak_out_i16 is still 0:
Something is wrong between push and pop. Lock-free
ring corruption, or wrong stride.
Pure diagnostic. No behavioural change beyond the logging.
Producer scratch envelope scan is O(scratch.len()) = 1920
samples per 20 ms tick = ~96k iterations/sec on the audio
producer thread \u2014 negligible CPU.
Build counter 73 -> 74.
User confirmed at +72 the symptom is 'voice + constant clicks +
choppy fragments'. The diagnostic data conclusively pointed to
iOS VPIO render-callback timing as the cause:
* frames_changes=60+ per 100 callbacks at cb>=1800
iOS keeps switching num_frames between 960 and 1104
on roughly 60% of callbacks
* peak_out_i16 is sensible (2500-16870, never clipping)
when AudioHandler returns content
* input_was_zero=true on most callbacks during active speech
AudioHandler keeps entering buffering_samples state
The cause: previous render callback called fill_buffer
synchronously every iOS audio thread invocation. With iOS
calling at irregular rates with irregular sizes, AudioHandler's
jitter buffer (sized around 20 ms Opus frames) cannot satisfy
arbitrary-sized requests and falls back to returning silence
(&[] empty slice) on misaligned reads. The silent gaps in the
middle of the output buffer create discontinuities = audible
clicks; the missing-tail content produces choppy fragments.
Fix (architectural): decouple the AudioHandler decoder from the
VPIO render callback via a lock-free SPSC ring buffer.
Producer (tokio task, 50 Hz):
every 20 ms:
fill_buffer(scratch_stereo_f32, 1920 = 20 ms stereo)
downmix L+R -> mono i16 (960 samples)
ring_buffer.push_slice(mono_i16)
Consumer (VPIO render callback, iOS audio thread):
every callback:
pop num_frames samples from ring buffer into out
zero-fill tail on underrun
Why it works:
* Producer always asks AudioHandler for a stable 20 ms chunk
(perfectly aligned with internal Opus frame size). No more
buffering_samples false-triggers.
* Consumer pulls whatever iOS asks for whenever iOS schedules
it; ring buffer's 200 ms depth absorbs the callback jitter.
* This is the standard pattern every production VoIP audio
engine uses (WebRTC, Discord, FaceTime) to bridge bursty
Opus decoders to bursty platform audio callbacks.
Implementation:
* New dep: rtrb 0.3.4 (RustAudio realtime-safe SPSC ring
buffer, 6.8M downloads, lock-free push/pop with no
allocation on the audio thread).
* RING_BUFFER_SAMPLES = 9600 (200 ms mono i16 at 48 kHz).
Sized for 10x producer ticks of headroom.
* PRODUCER_TICK_MS = 20 (matches Opus 50 Hz packet rate).
set_missed_tick_behavior(Skip) to avoid burst catch-up on
runtime stalls.
* Producer task spawned in IosVoiceUnit::start, shutdown
via tokio::oneshot when IosVoiceUnit drops.
* Render callback is now just: pop into out, zero-fill tail,
apply mute then gain.
* Gain applied CONSUMER-side so user volume changes take
effect within one callback (<= 200 ms latency).
* Underrun diagnostics: count underrun callbacks + total
zero-filled samples, log every 100 callbacks.
Threading + safety:
* rtrb is lock-free SPSC. Audio thread never blocks.
* Producer can block briefly on Arc<Mutex<AudioHandler>>
contention with the inbound forwarder (handle_packet), but
not with the audio thread.
* Producer task is owned by tokio runtime; explicit shutdown
channel ensures it exits when the engine stops.
Build verify:
* Linux host: cargo check clean in 4.06s (downloads rtrb 0.3.4).
* iOS Mac: cargo check clean in 2.14s.
Build counter 72 -> 73.
External code review pushed back on the 'iPhone speaker hardware
distortion' hypothesis and pointed out we need more than just
peak measurements. The reviewer's checklist:
* peak_i16
* rms_i16
* num_clipped_samples (abs >= 32767)
* zero_fill_count / underrun_count
* callback_frame_count variability
* decoded_packet_duration_ms
* input_was_zero
* actual ASBD / actual sample rate
The format diagnostics at +71 already showed iOS honoured
48 kHz Int16 mono on both buses and that .default mode +
.defaultToSpeaker routed to the speaker correctly with
outputVolume=0.45. So format + route are confirmed correct.
The remaining mystery is WHY 'loud but distorted' \u2014 we need
sample-level metrics to isolate where in the pipeline the
breakage occurs.
This commit instruments the VPIO render callback with:
* num_frames + frames_changes : detects iOS re-negotiating
buffer size between callbacks
(which would imply jitter the
fixed scratch_stereo Vec can't
absorb cleanly).
* peak_stereo + rms_stereo : characterises AudioHandler's
output BEFORE our downmix.
Distinguishes 'real audio
arriving' from 'silence'.
* peak_out_i16 + clip_count : measures what we hand VPIO.
clip_count > 0 means we're
clipping at our boundary even
with gain=1.0 \u2014 indicates
upstream is over-driven.
* input_was_zero : explicit silence/no-talker
indicator separate from peak=0
which could mean tiny content
rounded to 0.
Reviewer's preferred diagnostic path is to dump PCM to file
and play with ffplay externally; that's iOS-impractical
without a shared filesystem path the user can extract via
Files.app. Instead we sample the same metrics in-callback at
~2 Hz which gives us the same information at run time.
Pure diagnostic. No behavioural change. Counters live in the
FnMut closure so the audio thread cost is one branch +
counter increment per callback, plus a one-pass RMS sum +
peak scan every 100 callbacks.
Build counter 71 -> 72.
Reviewer also recommended a headphone test in parallel \u2014
that will be done by the user (out-of-band) at the next
test cycle to determine whether the symptom changes when
audio leaves the speaker path.
Per external review (helpful checklist from ChatGPT-style analysis
pointing out we never verified that iOS actually accepted our
preferred sample rate / channels / format): preferredSampleRate
and preferredIOBufferDuration are HINTS, not guarantees. iOS may
substitute its own values if the hardware can't satisfy our
preference. If VPIO is running at 44.1 kHz Float32 stereo while
our render callback writes 48 kHz Int16 mono into the buffer,
the symptoms would match what user reports (broken playback,
pitch shifted, severe distortion) and our previous diagnostics
wouldn't catch it because they only sampled signal-level metrics.
This commit adds two diagnostic emissions to verify:
1. AppDelegate.swift::activateAudioSession: after setActive
succeeds, log the ACTUAL session state \u2014 category, mode,
sampleRate, ioBufferDuration, current route (inputs +
outputs), outputVolume. Lets us see whether iOS honoured our
.default + .defaultToSpeaker setup and which physical route
it picked at launch.
2. ios_voice_unit.rs::IosVoiceUnit::start: after unit.start()
succeeds, log the actual OUTPUT and INPUT stream formats
VPIO accepted (sample_rate, channels, sample_format, flags).
If these differ from our requested 48 kHz Int16 mono, we
have a format-substitution problem.
Three possible outcomes from the next test:
* Both diagnostics confirm 48 kHz Int16 mono on both buses and
the session sampleRate=48000 -> format is correct; the
playback breakage is somewhere else (e.g. AudioHandler
jitter buffer behaviour, route binding, or hardware mixer).
* Session sampleRate != 48000 -> we need to insert a sample
rate converter or pin AVAudioSession's
setPreferredSampleRate(48000) explicitly in Swift before
setActive.
* VPIO substituted Float32 for our Int16 request -> our render
callback is writing i16 magnitudes into a Float32 buffer
which would explain the distortion. Fix: write Float32
directly using data::Interleaved<f32> instead of i16.
Build counter 70 -> 71. Pure diagnostic; no behavioural
change.
User report after the 8x boost commit (e85a6d3): playback STILL
broken, but now the diagnostic clearly shows the actual problem.
Render-callback peak_out_i16 SATURATES at 32767 on speech peaks
(cb=600, 1100, 1200, 2400) because the 8x boost amplifies an
already-loud signal into hard clipping. Quiet content reaches
audible level but loud peaks are catastrophically distorted.
The 8x boost was treating the wrong cause.
Real root cause (researched online after user prompted: 'this is
iOS a popular platform, there must be solutions'): iOS has TWO
independent audio channels:
In-call channel (.voiceChat / .videoChat modes)
* Routes through the phone-call audio path.
* Aggressively ducks non-voice content to the earpiece.
* Volume controlled by a separate in-call hardware
register, not the side buttons when not actively on a
phone call.
Media channel (.default mode)
* Routes through the standard media playback path.
* No automatic ducking.
* Volume controlled by the side volume buttons normally.
With AVAudioSession mode .voiceChat, iOS sends our output
through the in-call channel which plays at 'earpiece-level'
loudness on the speaker too. Signal is technically present but
buried under the speaker's noise floor. With mode .default +
.defaultToSpeaker option, output routes via media channel and
plays at normal loudness.
Both Twilio (video-quickstart-ios) and Daily.co (patched WebRTC
module) document the same workaround and use VPIO for AEC while
keeping the session mode at .default for loud playback:
github.com/twilio/video-quickstart-ios/issues/522
stackoverflow.com/questions/79834998 (Daily.co)
The user also noticed 'tx/rx almost no changes even receiving
packages' \u2014 likely a misinterpretation of the frames counter
not advancing as fast as expected during quiet voice; AudioHandler
returns silence when its jitter buffer is in buffering_samples
state which doesn't fire 'decode failed' but also doesn't
increment frames_received. The real issue is still the playback
ducking; the counter behaviour is a downstream symptom.
Changes:
1. AppDelegate.swift: AVAudioSession mode .voiceChat -> .default
with options [.defaultToSpeaker, .allowBluetoothHFP,
.allowBluetoothA2DP]. VPIO continues to do its job (AEC, NS,
AGC on the mic side); only the playback routing changes.
The earlier 'speaker selector silent under .default' bug
does NOT apply because we no longer use cpal RemoteIO \u2014
VPIO honours overrideOutputAudioPort under any mode.
2. ios_voice_unit.rs: revert the 8x output boost from e85a6d3.
With media-channel routing, signal levels are correct and
no software amplification is needed. Render callback restored
to plain (l+r)*0.5*gain downmix.
3. ios_voice_unit.rs: revert the BypassVoiceProcessing toggle
from c16318c. The VPIO chain stays enabled so we keep
capture-side AEC/AGC/NS for free \u2014 the playback breakage
it was trying to fix was the wrong layer all along.
4. ios_voice_unit.rs: drop the diagnostic render-callback log
line. Production-clean code; can be re-enabled by reverting
the diff in the closure if future debugging needs it.
Build counter 69 -> 70.
User-pasted log at +66 (https://pb.hit.moe/q8heratf.txt) shows
conclusive data over a 110-second continuous talker session:
Average peak_stereo_f32: ~0.005-0.010
Loud peak (one moment): ~0.234
peak_out_i16: ~150-300 (out of 32767)
The signal arriving at our render callback from
AudioHandler::fill_buffer is consistently at -40 dB FS for
normal human speech. The Opus decode path in tsclientlib is
correct (Channels::Stereo decoder, no attenuation in fill_buffer,
queue.volume defaults to 1.0). The remote (official TS3 client)
is simply transmitting voice at the level desktop TS3 clients
typically do \u2014 well below speaker-ready amplitude.
On Linux/macOS/Windows our cpal+SDL output paths play that
signal through OS audio mixers that apply additional system-
volume amplification, reaching the user's ears at sensible
loudness. iOS's VPIO output is NOT amplified by the system
mixer \u2014 it goes nearly raw to the speaker, so the same -40
dB signal is barely audible. Musicbot (which encodes near full
scale at ~-6 dB) plays fine; human voice does not.
Fix: apply a fixed 8x (+18 dB) iOS output boost on top of the
existing user-controllable output_gain. A -40 dB signal becomes
-22 dB (normal speakerphone level). User's volume slider
continues to function in a useful 0-2x range on top.
effective_gain = user_gain * IOS_OUTPUT_BOOST
Hard-clip at \u00b11.0 in the mono downmix prevents loud signals
(musicbot at peak 0.5 -> 4.0 -> clamped to 1.0) from
overflowing i16 wrap-around. Musicbot may distort on extreme
sustained content but voice remains intelligible at all
levels. Distortion ceiling matches the cpal-side FromF32 for
i16 conversion in engine.rs.
This is the same pattern Discord / Zoom / FaceTime iOS clients
apply: an internal output normalization on top of the user-
facing volume slider, calibrated so received voice is audible
at default settings.
Build counter 68 -> 69.
User report at +67 (.voiceChat mode): playback still 'broken'.
Even with VPIO's Apple-documented session-mode pairing,
its output-side gating chain (echo subtraction + adaptive
noise suppression) chops quiet inter-phoneme content of human
voice. Musicbot signal (loud, ~continuous) survives because it
stays above the gating threshold; speech does not.
Fix: set kAUVoiceIOProperty_BypassVoiceProcessing = 1 on the
unit immediately after EnableIO (before stream format / callbacks
/ initialize). This disables ALL VPIO voice processing \u2014 the
unit becomes effectively a vanilla RemoteIO with mic + speaker
buses. Raw samples pass through both directions.
Trade-off:
* Lost: Apple's hardware AEC + AGC + NS on the mic path. User
reports current capture is clean already, suggesting their
test environment (headset? non-speakerphone?) doesn't need
AEC. If echo loops back when speakerphone is engaged, we'll
re-evaluate \u2014 either re-enable VPIO selectively for
echo-prone routes or ship software AEC (DEC-007).
* Gained: playback is no longer gated. Quiet inter-phoneme
speech content reaches the speaker.
Property setter:
* Constant: kAUVoiceIOProperty_BypassVoiceProcessing = 2100
* Scope: Global, Element: Input (1) per WebRTC's reference iOS
ADM (voice_processing_audio_unit.mm).
* Value: u32 = 1 (= bypass).
* Soft-fail with warn log if the property is rejected on an
exotic iOS version (the unit still works, just with VPIO
defaults).
Build counter 67 -> 68.
User report at +66: capture (mic -> remote) is clean, but local
playback (remote -> speaker) is 'broken and poor', particularly
for human voice. Musicbot audio (loud, near-continuous) plays
correctly; human voice (peaks ~-6 dB, average ~-40 dB, classic
20 dB peak-to-average ratio) sounds gated out so most inter-
phoneme content is unintelligible.
Diagnostic at +66 (render callback peak sampling every 100
callbacks during a 60-second talker session) showed peak_stereo
values in the 0.005-0.01 range with occasional 0.13-0.49 spikes
\u2014 i.e. the signal is REAL and reaching the device, but VPIO's
output-side voice processing chain is gating the average-level
content.
Root cause: AVAudioSession mode .default + VPIO is a mismatched
pairing. Under .default mode the VPIO unit's internal AGC/NS
thresholds are tuned wrong for telephony-style speech and treat
quiet inter-phoneme content as noise to gate out.
Fix: switch back to mode .voiceChat which is Apple's documented
pair for VoiceProcessingIO. WebRTC's reference iOS audio device
manager (chromium googlesource voice_processing_audio_unit.mm)
also uses this pair. VPIO under .voiceChat tunes its processing
chain for speech and passes quiet content through cleanly.
The original 'speaker/receiver toggle is silent under .voiceChat'
bug was caused by cpal's RemoteIO unit binding to a stale
physical transducer at construction time, not by .voiceChat
itself. After migrating to VPIO at commits 1-4 (af686ca through
e7c3ffa) the route binding is correct under either mode because
VPIO natively re-binds on overrideOutputAudioPort \u2014 it IS the
canonical voice unit. So .default lost its only benefit and we
revert to the Apple-documented pairing.
Category options unchanged: .allowBluetoothHFP +
.allowBluetoothA2DP \u2014 BT headsets still permitted in both
directions regardless of mode.
Build counter 66 -> 67.
User reports capture-side audio (mic -> remote) is clean but
local playback (remote -> speaker via VPIO render callback) is
'broken and poor' at +65. The pipeline appears correct on paper:
fill_buffer -> downmix (L+R)*0.5 -> gain -> clamp -> i16 -> VPIO.
No errors logged. To stop guessing, add structured logging
inside the render callback so the next test cycle yields data
about what's actually flowing through.
Diagnostic emitted every 100th callback (~2 s at iOS's typical
20-50 Hz callback rate):
ios VPIO render callback diagnostic sample
cb=<counter>
num_frames=<N> VPIO buffer size in mono samples.
Expected ~960 (20ms) or ~1104 (23ms).
Outliers point at format mismatch.
peak_stereo_f32=<f32> Peak |sample| of AudioHandler's
output BEFORE gain + downmix.
0.0 = handler is producing silence
(jitter underrun, no audio).
~1.0 = full-scale content reaching
the callback as expected.
peak_out_i16=<i16> Peak |sample| of the downmixed mono
i16 we write to VPIO. Zero with
non-zero peak_stereo = downmix bug.
Near 32767 = clipping pressure.
gain=<f32> Current master output gain.
What we'll be able to diagnose from a 5-second talker session:
* peak_stereo_f32 = 0 throughout
-> AudioHandler isn't producing samples. Inbound forwarder
may not be feeding it, or jitter buffer is stuck in
buffering_samples state. NOT a render-callback bug.
* peak_stereo_f32 oscillating, peak_out_i16 = 0
-> Downmix or i16 cast is broken. Math bug in the loop.
* num_frames wildly different from ~960-1104
-> StreamFormat got rejected and VPIO is delivering a
different rate. Format-pinning fight with the session.
* peak_stereo_f32 normal AND peak_out_i16 normal AND user
still says 'broken'
-> The signal reaches the device cleanly but iOS's VPIO
output processing (AEC residual subtraction, AGC
compression, NS gate) is mangling it after our callback
returns. That's a VPIO-config problem, not a render-
callback problem; fix is to disable specific VPIO
voice-processing properties on the unit before
initialize().
Pure diagnostic commit. No behavioural change beyond a
warn-rate-limited info log line every ~2 seconds. Cost in the
audio thread is one branch + counter increment + (every 100th)
a tracing macro invocation.
Build counter 65 -> 66.
Two related fixes for the version-display work landed at 97a6ba6:
PROBLEM 1: build counter stuck at +59 on the device.
The user reported the About dialog showed v1.0.0-rc.8+59 even
though pubspec.yaml has been bumping (60 -> 61 -> 62 -> 63 -> 64).
Root cause: Xcode caches ios/Flutter/Generated.xcconfig (which
holds FLUTTER_BUILD_NUMBER) and doesn't regenerate it on Cmd+R
unless inputs change. 'flutter pub get' also doesn't rewrite it.
Only 'flutter build ios' or deleting the file forces regen.
Fix (operational, not code-side): the Generated.xcconfig file on
the Mac was deleted + regenerated and is now at
FLUTTER_BUILD_NUMBER=64, so the next Xcode Run picks up the
correct value. Going forward, if the build counter ever lags
again the workaround is:
rm ios/Flutter/Generated.xcconfig && flutter pub get
before running from Xcode. We may add this to a build-doc note
or a Makefile target during the rc.8 wrap-up.
PROBLEM 2: 'v1.0.0-rc.8' was being displayed as 'v1.0.0.8'.
CFBundleShortVersionString on iOS rejects non-numeric
characters, so Flutter strips the '-rc.8' suffix to '.8' when
populating Info.plist. package_info_plus.version reflects that
mangled value. CFBundleVersion (the build counter) is passed
through intact, so the issue affects only the semver half.
Fix: split _kAppVersion resolution. Hardcode the semver baseline
as _kSemverBaseline = 'v1.0.0-rc.8' (kept in sync with the git
tag + pubspec semver portion; bumped once per release-candidate
cycle, not per test build). Use package_info_plus for the
'+<buildNumber>' suffix only, where CFBundleVersion survives the
sanitiser unchanged. Final display becomes
'v1.0.0-rc.8+<n>' (e.g. 'v1.0.0-rc.8+65').
flutter analyze: clean.
Build counter 64 -> 65. After Xcode Clean Build Folder + Run the
About dialog should now read 'v1.0.0-rc.8+65' on the iPhone.
Replace the silence-emitting render callback from commit 1 with
real playback that drives AudioHandler::fill_buffer and downmixes
its 48 kHz stereo f32 output to the i16 mono buffer VPIO expects.
Pipeline per render callback (mirrors the cpal-output + sdl_output
contracts so the platform-neutral playback path is preserved):
1. Lock the shared Arc<Mutex<AudioHandler>>, ask fill_buffer to
populate a stereo-f32 scratch slice of length 2*num_frames.
AudioHandler runs Opus decode + per-client jitter buffer + mix
internally. Same primitive every other platform calls.
2. If output_muted is true, zero the i16 output buffer and return.
We still ran fill_buffer in step 1 so the jitter buffer drains
while muted — preventing unbounded growth — which matches the
cpal/SDL backend contract.
3. Downmix stereo -> mono with master gain:
mono_f32 = (l + r) * 0.5 * gain
i16_out = (mono_f32.clamp(-1.0, 1.0) * i16::MAX) as i16
The 0.5 average preserves total signal energy with 3 dB
headroom against sum-of-correlated-peaks clipping. Multiply by
gain after the downmix saves one mul per sample. Hard-clip on
the i16 cast is acceptable because the upstream stereo signal
is already in [-1.0, 1.0] from the f32 mix; only gain >1.0
creates clipping pressure and that path is identical to every
other backend's i16 conversion.
Closure ownership:
* scratch_stereo: Vec<f32> moved into the FnMut closure. First
callback grows it to 2*num_frames; subsequent callbacks reuse
the backing allocation. The audio thread never hits the
allocator on steady-state callbacks.
* handler_for_render / output_gain_for_render / output_muted_for_render
are Arc clones taken before the closure literal.
Public API change: AudioEngine -> IosVoiceUnit::start parameters
that were previously underscored (commit 1 placeholder) are now
all consumed by the wiring. Signature is unchanged, just the
binder names lose the leading underscore. engine.rs call-site
is unaffected.
Build verify on Mac (target aarch64-apple-ios): cargo check
clean in 0.33s, no errors, no warnings.
Build counter 63 -> 64 — About dialog shows v1.0.0-rc.8+64.
Replace the no-op input callback from commit 1 with a real
capture pipeline that mirrors the cpal-side CaptureState in
engine.rs but is type-specialised for the i16 mono samples VPIO
delivers natively.
New IosCaptureState struct (private to ios_voice_unit.rs) owns:
* OpusEncoder configured for VoIP at 48 kHz mono (32 kbps,
complexity 10, inband FEC, packet-loss-perc 5 — identical
tuning to try_open_capture in engine.rs).
* pcm_accum: Vec<i16> with capacity 2*FRAME_SAMPLES_MONO, growing
if a VPIO callback ever delivers more than ~40 ms.
* opus_out: [u8; MAX_OPUS_FRAME] scratch.
* Cloned Arc<AtomicBool> transmit gate + Arc<AtomicU32> frames-sent
counter shared with AudioEngine.
ingest_i16 flow:
1. If PTT gate is off -> clear accumulator + return (matches cpal
behaviour, no pop on PTT-release edge).
2. Apply mic_gain. Fast-path when gain==1.0 skips the multiply +
saturate loop entirely; otherwise saturating mul-then-cast
keeps the signal in the i16 envelope.
3. Drain complete 20 ms / 960-sample frames from the accumulator,
encode via encoder.encode (i16 path, no float conversion
needed since VPIO already gave us i16), build OutPacket with
AudioData::C2S { codec: OpusVoice }, try_send on voice_out_tx.
4. Frame buffer is stack-allocated [i16; FRAME_SAMPLES_MONO] —
no per-callback heap allocation on the realtime audio thread.
VPIO setup changes in IosVoiceUnit::start:
* NEW: explicit kAudioOutputUnitProperty_EnableIO (=2003) with
value 1 on (Scope::Input, Element::Input) BEFORE the stream
format setters. VPIO's input element is OFF by default; without
this toggle no audio flows in and the input callback never
fires. Commit 1's comment claiming set_input_callback handles
this was wrong; coreaudio-rs's set_input_callback only installs
the kAudioOutputUnitProperty_SetInputCallback property, not
the EnableIO toggle.
* Apple's documented sequence (now matched):
1. AudioComponentInstanceNew -> AudioUnit::new_uninitialized
2. EnableIO on element 1 -> set_property(2003, ...)
3. Stream format both elems -> set_stream_format x2
4. Install callbacks -> set_input_callback + set_render_callback
5. AudioUnitInitialize -> unit.initialize
6. AudioOutputUnitStart -> unit.start
* set_input_callback closure now moves the IosCaptureState in
by value and calls ingest_i16 with args.data.buffer (the
&mut [i16] coreaudio-rs delivers after running AudioUnitRender
internally to pull the mic samples into a pre-allocated
AudioBufferList).
What this commit does NOT do:
* Output render callback is still a silence-emitting stub.
Commit 4 lands the AudioHandler::fill_buffer + i16 downmix.
* Route-change handling — commit 5.
Build verify on Mac (target aarch64-apple-ios): cargo check
clean in 1.18s, no errors, no warnings.
Build counter 62 -> 63 — About dialog shows v1.0.0-rc.8+63.
Two follow-ups after the iOS Rust build went green at 5de6ecc:
1. cpal-side framing constants (SAMPLE_RATE / FRAME_SAMPLES /
MAX_OPUS_FRAME) are dead code in the current iOS commit
because the VPIO callbacks are still no-op stubs and don't
reach the constants yet (commits 3 + 4 will). They are
genuinely live on every other platform via the cpal capture
pipeline. Mark each with #[allow(dead_code)] and add a
comment pointing at the commits that will reactivate them on
iOS, instead of cfg-gating per-platform (the constants are
framing invariants of the engine itself, not per-backend
details).
2. ios/Podfile.lock regenerated on the Mac via 'pod install'
to register package_info_plus (0.4.5) which landed in
97a6ba6. Without this regen the Xcode build fails with
'The sandbox is not in sync with the Podfile.lock' because
Xcode's CocoaPods integration check sees a new plugin in
pubspec.yaml that has no matching Pod entry. Five pods now
in the lockfile: Flutter, audio_session, chanora_bridge,
connectivity_plus, package_info_plus.
Build counter 61 -> 62 — the About dialog will display
v1.0.0-rc.8+62 so the user can confirm the build under test
matches this commit (the previous build said +61).
iOS build of chanora_audio (target aarch64-apple-ios) failed with
four compilation errors after commit 2 landed. Root causes were
all simple symbol-path / cfg-gating mistakes from the skeleton
commit; the underlying design is unchanged.
1. ios_voice_unit.rs: wrong import path for OutPacket. The
chanora_protocol crate re-exports it at the crate root
(`pub use ...::OutPacket` in lib.rs line 52), not from a
`voice` submodule (which doesn't exist).
- use chanora_protocol::voice::OutPacket;
+ use chanora_protocol::OutPacket;
2. ios_voice_unit.rs: LinearPcmFlags lives in
`coreaudio::audio_unit::audio_format`, not in
`stream_format` (the doc page lists it under StreamFormat but
the actual module path is the upstream Apple naming).
- use coreaudio::audio_unit::stream_format::LinearPcmFlags;
+ use coreaudio::audio_unit::audio_format::LinearPcmFlags;
3. ios_voice_unit.rs: `Ordering` import unused (commit 1
skeleton callbacks don't load atomics yet — that comes in
commits 3 + 4). Remove from the std::sync::atomic import to
silence the unused_imports warning.
4. engine.rs::AudioEngine::stop(): the existing body unconditionally
touched self._input_stream and self._output_stream, but commit
2 cfg-gated those fields away on iOS (and added an iOS-only
_ios_voice_unit field in their place). Split the field drop
logic with the same target_os = "ios" cfg so each platform
only touches the fields it actually has.
Also clean up two pre-existing warnings exposed by the iOS cfg
gating:
5. engine.rs: `tracing::{error, warn}` were imported
unconditionally but are only used inside cpal log lines.
Cfg-gate the import to not(target_os = "ios").
6. engine.rs: `AudioData`, `CodecType`, `OutAudio` from
chanora_protocol are only referenced in the Opus encoder feed
inside CaptureState — cpal-side only. Cfg-gate to
not(target_os = "ios"); keep `InboundVoice` + `OutPacket`
on the unconditional path because the inbound forwarder + (in
commit 3) the iOS capture pipeline both reference them.
Also fix a stray duplicate `#[cfg(not(target_os = "ios"))]`
attribute that landed on line 18 in commit 2.
Build verify (Linux host): cargo check -p chanora_audio clean
in 0.53s. iOS-side check pending on Mac.
Build counter bumped 60 -> 61 — the About dialog will display
v1.0.0-rc.8+61 so the user can confirm the build under test
matches this commit.
Add package_info_plus 8.3.1 (Flutter Community Plus, BSD-3, ~3M
downloads) and resolve the displayed version string from the
platform manifest at app init. The string shown in the About
dialog is now formatted as 'v<version>+<build>' (e.g.
'v1.0.0-rc.8+60') where both halves come from the SAME
pubspec.yaml 'version:' field that Flutter uses to populate iOS's
CFBundleShortVersionString + CFBundleVersion and Android's
versionName + versionCode.
Motivation: during the iOS VoiceProcessingIO audio rollout the
user is rebuilding repeatedly from Xcode. Without a build-number
suffix there is no way to confirm from inside the app which
commit produced the build under test — they all read
'v1.0.0-rc.8'. Every test commit going forward bumps the +<build>
field in pubspec.yaml so the About dialog uniquely identifies the
build.
Implementation:
* pubspec.yaml: version 1.0.0-rc.8+59 -> +60 (this commit is a
test build). Add package_info_plus: ^8.0.0.
* main.dart: _kAppVersion changes from 'const String' to
'String' (resolved at runtime). Populated in main() before
runApp() via new _resolveAppVersion() helper that calls
PackageInfo.fromPlatform() and formats 'v<info.version>+<info.buildNumber>'.
Falls back to the hardcoded 'v1.0.0-rc.8' baseline if the
platform call fails (extremely unlikely; the platform channel
is a constant lookup).
The consumer site at the About dialog (l10n.aboutVersion(_kAppVersion))
is unchanged — the variable still resolves to a String. flutter
analyze: clean.
Split AudioEngine::start_with_gate into two backends:
* start_with_gate_cpal — non-iOS path, the existing cpal + (SDL on
Linux) flow, renamed verbatim, no
behavioural change.
* start_with_gate_ios — iOS path, constructs a single
IosVoiceUnit (VoiceProcessingIO via
coreaudio-rs) for combined mic + speaker.
Spawns the same inbound forwarder task
that pumps Opus packets into AudioHandler.
The public entry point start_with_gate dispatches at the top via
cfg(target_os = "ios") so callers stay backend-agnostic.
Struct field changes:
* _input_stream : cfg-gated to not(ios)
* _output_stream : cfg-gated to not(ios), keeps the
Linux=SdlOutput / else=cpal::Stream split
* _ios_voice_unit: new field, cfg-gated to ios, owns the VPIO
AudioUnit for the engine's lifetime.
Module-level cfg-gating:
* All cpal-only helpers (try_open_capture, build_input_stream,
build_output_stream, CaptureState + impl, ToF32 / FromF32 traits
and impls, PlaybackResampleState) are now wrapped with
#[cfg(not(target_os = "ios"))]. Same for the audiopus
encoder + cpal trait imports — iOS doesn't pull libopus into the
engine yet (commit 3 will, once the VPIO input callback wires
into CaptureState).
Behaviour on iOS for THIS commit:
* AudioEngine starts cleanly, IosVoiceUnit::start succeeds (VPIO
unit allocates + initialises + starts).
* Mic capture is dropped (the input callback is a no-op stub).
* Output emits silence (the render callback fills the buffer with
zeros).
* Inbound forwarder still runs and pushes Opus packets into
AudioHandler — they accumulate in the jitter buffer but no
fill_buffer drain happens (commit 4 fixes that), so the buffer
will grow up to MAX_BUFFER_TIME (~0.5 s) and then tsclientlib
starts dropping the oldest frames. This is fine for now — the
point of this commit is verifying the AudioUnit constructs +
starts cleanly on the device. Audible silence is the expected
state until commits 3/4 land.
Build verify (Linux host): cargo check -p chanora_audio clean in
0.53s. iOS-side compile happens on the Mac via the Xcode build
the user will trigger next.
Skeleton scaffolding for the iOS VoiceProcessingIO backend that
will replace cpal on iOS. This commit lands the dependency + the
module + a constructable AudioUnit that emits silence and drops
input; nothing in engine.rs is wired up yet (that is commit 2).
Compilation contract for this commit:
* Linux / Windows / macOS / Android builds unaffected (the new
module is target_os='ios' gated, the new dep is in
'[target."cfg(target_os = \"ios\")"]').
* iOS build pulls in coreaudio-rs 0.14, constructs a VPIO unit,
pins stream format to 48 kHz Int16 mono on both buses, installs
no-op input + silence-emitting render callbacks, initializes,
and starts. No audio is actually moved until commits 3/4.
Why VPIO and not RemoteIO via cpal: cpal's iOS backend opens
RemoteIO with no control over stream format / buffer size /
channels and produces a mono-only output element that stays bound
to the route present at construction time. End-user symptom on
iPhone 16 Pro iOS 18.7.8: tapping Speaker in the picker flips
AVAudioSession.currentRoute.outputs to Speaker (confirmed in our
diagnostic logs from commit da631a2) but audio keeps coming out
the receiver because the AudioUnit's output binding is stale.
Every production iOS VoIP client (Mumble iOS, Linphone /
mediastreamer2, Signal-iOS, Jitsi Meet iOS, the WebRTC reference
impl) avoids RemoteIO and uses VoiceProcessingIO instead. VPIO is
Apple's recommended voice unit; it ships hardware AEC + AGC + NS
and re-binds the physical transducer correctly on route changes
because it IS the canonical voice unit on iOS — FaceTime's audio
path runs through it.
coreaudio-rs 0.14 (RustAudio org, 8.6M downloads, same maintainers
as cpal) gives us a safe wrapper around the AudioUnit C API on
iOS. Uses objc2-* crates underneath so links cleanly into iOS
builds. ios_voice_unit.rs sits next to sdl_output.rs as the iOS
sibling of the Linux SDL2 output path.
Build verify (Linux host): `cargo check -p chanora_audio`
finished clean in 16.09s. iOS build verification happens in
commit 2 when the module is exercised; for this commit the module
compiles in isolation but is dead code on iOS too (no caller).
Two changes that together address the user-reported 'speaker selector
not working' AND 'audio quality bad' symptoms on iPhone:
1. AppDelegate.swift: AVAudioSession mode .voiceChat -> .default
.voiceChat binds the underlying AudioUnit's output element to a
SINGLE physical transducer (the receiver/earpiece) at session-
configure time. overrideOutputAudioPort updates AVAudioSession's
route metadata so currentRoute.outputs reports Speaker, but the
AudioUnit's output binding is stale and audio keeps routing to
the original transducer. Net: tapping Speaker in the picker
flipped the route in our log but produced no audible change.
.voiceChat also enables iOS's telephony processing chain (forced
mono output, aggressive AGC, heavy noise gating) which explains
the 'garbled / watery / metallic' quality complaints.
.default mode uses iOS's standard audio graph: stereo output, no
AGC, no telephony post-processing, AudioUnit re-binds live when
the route changes. Same mode Music.app and most non-telephony
apps use. Trade-off: we lose iOS hardware AEC. If users report
speakerphone echo we'll add software AEC (DEC-030).
Category options unchanged \u2014 .allowBluetoothHFP +
.allowBluetoothA2DP still permit BT headsets in both directions.
2. engine.rs::build_output_stream: mono device downmix fix
The mixing path at dev_channels==1 previously wrote only the L
channel of AudioHandler's stereo output into the single mono
device channel and discarded R entirely. Anything panned right
in the stereo voice mix was silently lost \u2014 on .voiceChat
speakerphone (forced mono device) this manifested as quiet
remote speakers being inaudible. Fix: when dev_channels==1,
output = (L + R) * 0.5 instead of just L. The dev_channels>=2
branch is unchanged.
With change #1 iOS will typically expose stereo so this branch
is rarely hit, but the fix is correct for any genuinely-mono
sink (some BT car-audio profiles, USB mono headsets).
Add structured NSLog instrumentation to AppDelegate.swift and debugPrint
chains in voice_compact.dart::_AudioOutputPickerSheetState so we can
correlate user picker taps with what iOS actually does to the route.
Three diagnostic streams:
* 'chanora.session[<tag>]' from Swift — full session snapshot (category,
mode, sampleRate, ioBufferDuration, current route inputs+outputs,
preferredInput) emitted on every setActive and every
AVAudioSession.routeChangeNotification with the reason decoded
(override / routeConfigurationChange / newDeviceAvailable / etc).
* 'chanora.route[<tag>]' from Dart — current route's inputs+outputs
emitted before/after every overrideOutputAudioPort or
setPreferredInput call, plus a delayed re-check at +250 ms to detect
silent reverts.
* Existing 'chanora: ...' debugPrint lines from the picker now include
the OK case (override returned, setPreferredInput returned) so we see
a positive signal in the log when the API didn't throw.
Used to root-cause the 'speaker selector not working' issue: the
hypothesis is that cpal's RemoteIO AudioUnit reacts to its own format
configuration notifications by triggering routeConfigurationChange
that reverts our Dart-side override. The logs will confirm or deny
this — if we see 'chanora.session[routeChange.override] out=Speaker'
followed by 'chanora.session[routeChange.routeConfigurationChange]
out=Receiver' within a few hundred ms, that's the smoking gun.
Pure diagnostic commit. No behavioural change. Logs are NSLog +
debugPrint so they appear in Xcode console / 'flutter logs' / the
device log via Console.app or 'devicectl device log'.
User report: 'speakerphone (built-in mic input)' logs printed
successfully but audio output didn't actually switch to speaker.
No exception thrown by either AVAudioSession call.
Root cause:
Previous _selectSpeaker called:
1. await overrideOutputAudioPort(.speaker)
2. await setPreferredInput(builtInMic)
Apple-documented behavior in .voiceChat mode: when
setPreferredInput is called, iOS recalculates the entire route
based on the natural input/output pairing for the selected port.
Built-in mic's natural output pairing is the receiver/earpiece
(matches the 'I'm talking on a phone' UX of .voiceChat). So step 2
caused iOS to SILENTLY REVERT the speaker override from step 1
and route output back through the earpiece.
Net: await chain returned without exception (both calls 'succeeded'
in API terms), debugPrint logged the success message, but the user
heard the call audio coming out of the earpiece, not the speaker.
Same issue affected _selectReceiver (after the speaker bug fix
was reverted): redundant setPreferredInput(builtInMic) call risked
the same recalc race.
Fix:
* _selectSpeaker: only call overrideOutputAudioPort(.speaker).
No setPreferredInput. The override alone is sufficient \u2014 the
input stays on whatever the system was already using (built-in
mic by default, or BT/wired if connected).
* _selectReceiver: only call overrideOutputAudioPort(.none).
No setPreferredInput. Removing the speaker override naturally
returns to .voiceChat's default route (receiver).
* _selectInput (BT / wired / USB): unchanged. These inputs PAIR
their own output device, so .none + setPreferredInput is the
correct combo (user hears audio through the same device they
speak into).
flutter build ios --release --no-codesign: 21.2 s, Runner.app
30.4 MB.
User report: 'speaker change not work' \u2014 selecting Speaker or
iPhone receiver in the audio output picker had no audible effect.
Root cause:
AVAudioSession was configured with category options
[.defaultToSpeaker, .allowBluetoothHFP, .allowBluetoothA2DP] +
mode .voiceChat. The .defaultToSpeaker flag tells iOS 'this app's
baseline output route is the speakerphone, even though .voiceChat
mode would normally route to the receiver.'
When the user picked Speaker:
overrideOutputAudioPort(.speaker) <- already at speaker baseline; no-op
When the user picked iPhone receiver:
overrideOutputAudioPort(.none) <- removes speaker OVERRIDE,
restores baseline = .defaultToSpeaker
= speakerphone. Receiver
row silently mapped to speaker.
So both rows produced the same audible state. The picker UI changed
the selected radio but the route didn't actually move.
Fix:
1. AppDelegate.swift: drop .defaultToSpeaker from options. With
pure .voiceChat mode (no .defaultToSpeaker), the baseline is
the receiver/earpiece. overrideOutputAudioPort then works as
documented:
Default = receiver
overrideOutputAudioPort(.speaker) -> speakerphone
overrideOutputAudioPort(.none) -> back to receiver
BT/AirPods connected -> automatic
Wired headphones plugged in -> automatic
2. voice_compact.dart: replace 'catch (_) {/* ignore */}' silent
swallow with debugPrint logging of (a) the actual exception
and (b) which route was selected. So if iOS rejects an
override (e.g. wired headphones plugged in), we can see WHY
in the device log instead of a silent picker no-op.
3. _selectSpeaker also now sets preferredInput to the built-in
mic so input + output stay consistent. Previously the
speakerphone override could leave the mic still routed to a
previously-selected BT input \u2014 user hears self through
speaker but server hears nothing.
flutter build ios --release --no-codesign: 21.9 s, Runner.app
30.4 MB.
User report: 'sound heard are too poor' on iPhone iOS \u2014 garbled/
robotic + stutters/dropouts.
The Opus encoder ran with audiopus defaults: 'auto' bitrate that
drops to ~6 kbps during silence (sounds watery on speech resume),
inband FEC disabled (single packet loss = silent gap), no packet-
loss percentage hint (encoder can't budget bits for redundancy).
On lossy mobile networks (cell, WiFi roaming) this combination
sounds noticeably worse than the same Opus stream from a desktop
client. Garbled = silence-bitrate transitions; stutters = packet
loss without FEC.
Fix: tune the encoder once at construction with values derived
from RFC 6716 \u00a77.1 (Opus VoIP recommendations), Discord's voice
client tuning, and the Mumble defaults.
* set_bitrate(32_000) \u2014 sweet spot for mono speech. Below
24 kbps starts to sound watery;
above 64 kbps wastes bandwidth.
Discord uses 64 kbps; Mumble
defaults to 40 kbps; we pick
32 kbps as a conservative VoIP
value that survives ~100 kbps
uplinks comfortably.
* set_complexity(10) \u2014 max quality. The CPU cost on a
modern iPhone (A14+) or any
desktop is negligible (~0.5 % of
a single core for 48 kHz mono).
* set_inband_fec(true) \u2014 Opus inserts a low-bitrate copy
of the previous frame inside the
current packet so single-packet
loss can be reconstructed from
the next packet. Essential on
lossy mobile. The decoder side
(tsclientlib's AudioHandler)
auto-handles FEC frames; no
receiver-side change needed.
* set_packet_loss_perc(5) \u2014 tells the encoder to budget
bits for 5 % expected loss.
Higher values trade audio
quality for resilience.
Each setter is wrapped in a soft-fail: if an exotic libopus build
rejects one of these, we log + continue with the still-functional
encoder rather than aborting the audio engine. info!-log a one-
liner per-engine-start summarising the tuned values so a future
diagnostic export can correlate audio reports with the active
configuration.
Application::Voip mode was already set (engine.rs:630, unchanged);
the new calls layer on top of that mode's defaults.
cargo test -p chanora_audio --release --lib: 32 passed.
flutter build ios --release --no-codesign: 17.2 s, Runner.app
30.4 MB.
Two user-reported issues addressed:
1. 'when user join the server aka default channel, but there are
no ptt button also can not talk'
Root cause: TS3 servers auto-place a newly-connected client into
the server's default channel. We did not detect that. The UI
gated all voice controls (PTT button, mic/headset AppBar icons,
voice modal entry point) on _inChannel which was only flipped
true by an explicit voice_join() call from the user. So after
connect the user saw themselves in the default channel via the
channel tree but had no way to talk.
Fix: in chanora_core::Session::connect(), after the initial
snapshot resolves, call find_own_in(&snap) to determine whether
the server placed us in a channel. If yes, voice_selector
.set_in_channel(true) + emit SessionEvent::VoiceState
{ in_channel: true } + ensure_audio_running. This treats the
server-side default channel placement identically to a user-
driven voice_join: the UI receives a VoiceState(true) event and
renders all voice controls.
Tolerates audio engine startup failure the same way voice_join
does \u2014 server-side we are in the channel regardless; if mic
permission / device init fails, the UI gains the controls and
emits SessionEvent::AudioStopped so the user can resolve the
underlying issue.
Drops the inner lock before calling the public helpers because
set_in_channel + emit_voice_state + ensure_audio_running all
re-lock self.inner.
2. 'save bookmark cause a crash framework.dart line 6268
_dependents.isEmpty is not true'
Root cause: the showDialog-with-inline-TextEditingController-
dispose anti-pattern. _onAddCurrentBookmark and
_askChannelPassword both constructed a TextEditingController in
the surrounding async function, passed it to a dialog's
TextField via the dialog builder, then called ctl.dispose()
synchronously after returned.
On iOS the dialog route pop animation is still mid-flight when
showDialog's Future resolves. The inline dispose tore the
controller out from under EditableText while EditableText still
held InheritedWidget dependencies on the dialog route (theme,
localizations, default text style). When the dialog route's
InheritedElement then deactivated as part of the pop animation,
the framework's debug-only assertion _dependents.isEmpty tripped
because the disposed dialog tree had not finished detaching its
dependents yet.
Fix: hoist both dialogs into dedicated StatefulWidgets
(_BookmarkNameDialog and _ChannelPasswordDialog) that own their
own TextEditingController. The State.dispose() runs as part of
the dialog's normal unmount lifecycle, AFTER the pop animation
completes and all InheritedWidget dependencies have been cleared.
No race possible.
Bonus: dialog builders now use the dialog's own ctx for
AppL10n.of(...), Theme.of(...), and Navigator.of(...) calls
uniformly, rather than capturing the outer _HomePageState
context's l10n in a closure. That avoids a secondary leak where
the dialog widget tree held references back to the outer
route's InheritedElements through closure capture.
Added onSubmitted: -> pop(_ctl.text) on both fields so iOS
hardware-keyboard 'return' submits the dialog (small UX win
discovered while restructuring).
iOS build: flutter build ios --release --no-codesign 20.3 s,
Runner.app 30.4 MB. flutter analyze: 6 pre-existing Radio
deprecation infos (unchanged). cargo test -p chanora_core --release
--lib: 13 passed.
User reported: the 'output device' picker only listed AirPlay
destinations (other iPhones / AirPlay speakers / AppleTV) and not
the speaker / iPhone receiver / AirPods / wired headset choices.
Root cause: audio_router 1.1.1's iOS path uses AVRoutePickerView,
which is Apple's **AirPlay** picker UI \u2014 by design it only lists
AirPlay-eligible output destinations, NOT the input/output route
choices we need (speaker vs receiver vs Bluetooth HFP vs wired).
AVRoutePickerView is the right UI for 'cast audio elsewhere'; for
'pick how I hear / talk' (VoIP) the right primitive is direct
AVAudioSession calls.
Fix: replace audio_router with audio_session 0.2.3 (Ryan Heise,
verified publisher, 865k downloads, MIT). audio_session exposes:
* AVAudioSession.availableInputs \u2014 enumerate every real input
port: builtInMic, bluetoothHfp, bluetoothA2dp, headsetMic
(wired), usbAudio, carAudio, airPlay.
* AVAudioSession.currentRoute \u2014 .inputs + .outputs of the
active route.
* AVAudioSession.setPreferredInput(port) \u2014 switch the input
(HFP / wired / USB / car audio also move output to themselves).
* AVAudioSession.overrideOutputAudioPort(.speaker | .none) \u2014
toggle built-in speakerphone vs receiver/earpiece.
* AVAudioSession.routeChangeStream \u2014 live notifications when
the user plugs / unplugs / connects a device while the picker
is open.
This is exactly the same primitive Discord, WhatsApp, FaceTime
use for their VoIP audio chooser. No native UI plugin needed.
New widgets in voice_compact.dart:
* _AudioOutputTile: shows the active output port name (Speaker /
iPhone / AirPods / 'Phil's Wired Headset' / etc.) with the
matching icon. Subscribes to routeChangeStream for live
updates. Tap opens _AudioOutputPickerSheet.
* _AudioOutputPickerSheet: bottom sheet with 'Choose audio' title
and a Discord-style list:
- Speaker (volume_up)
- iPhone (phone_in_talk; the receiver/earpiece)
- <BT name> (bluetooth_audio)
- <Wired headset> (headset)
- <USB / Car> (usb / directions_car)
Selected row is highlighted + has a check mark. Tap routes:
- Speaker -> overrideOutputAudioPort(.speaker)
- iPhone -> overrideOutputAudioPort(.none) + setPreferredInput(builtInMic)
- External -> overrideOutputAudioPort(.none) + setPreferredInput(port)
* _PickerRow: shared row widget with selected/check styling.
AppDelegate.swift is unchanged: the manual AVAudioSession
.setCategory(playAndRecord / .voiceChat) we already do at launch
(0466000 / 4ee2b38) is fully compatible with audio_session \u2014 the
plugin only adds Dart-side accessors over the same underlying
AVAudioSession singleton.
Removed l10n keys not used anymore (audioRouteUsb was already gone).
Kept audioRouteSpeaker / Receiver / Bluetooth / WiredHeadset /
CarAudio / Airplay / Unknown \u2014 all still used by the new picker.
flutter analyze: 6 pre-existing Radio deprecation infos (unchanged).
flutter build ios --release --no-codesign: 54.9 s, Runner.app
30.4 MB (+200 KB vs audio_router build).
User reported 'package:flutter/src/widgets/framework.dart line 6268
_dependents.isEmpty is not true' assertion AND the persistent
keyboard lag.
Root cause analysis:
framework.dart:6268 is the assertion in
InheritedElement.debugDeactivated() that fires when an
InheritedElement is being deactivated while descendants still
depend on it. This fires in debug builds; release builds skip it.
The previous tap-outside-to-dismiss implementation used:
return GestureDetector(
behavior: HitTestBehavior.opaque,
onTap: () => FocusScope.of(context).unfocus(),
child: Column(...),
);
The FocusScope.of(context) call subscribes this GestureDetector's
Element to the _FocusScopeMarker InheritedWidget on every build.
During the modal-sheet-pop / Navigator-deactivate sequence, the
ancestor _FocusScopeMarker InheritedElement can deactivate before
the descendant GestureDetector clears its dependency, tripping the
debug assertion.
This was the same root cause as the keyboard lag: an outer
GestureDetector in the gesture arena ahead of the TextField's own
recognizer ALWAYS interferes, either via the opaque-vs-translucent
arena race (causing lag) or via the InheritedWidget subscription
race (causing the assert).
Fix: remove the outer GestureDetector entirely. Use the built-in
TextField.onTapOutside callback added in Flutter 3.10+ instead:
TextField(
...
onTapOutside: _onTapOutside,
)
void _onTapOutside(PointerDownEvent _) {
FocusManager.instance.primaryFocus?.unfocus();
}
Why this is strictly better:
* FocusManager.instance is a global singleton with NO
BuildContext dependency. No InheritedWidget is subscribed; no
_dependents map grows; no race possible on dispose.
* TextField.onTapOutside uses the framework's TapRegion machinery
internally. Taps inside the field's TapRegion don't fire the
callback; only true outside-region taps do. Zero gesture-arena
interference with the TextField's own tap recognizer \u2014
keyboard appears synchronously on the first tap with no lag.
* iOS's default _EditableTextTapOutsideAction at
flutter/widgets/editable_text.dart:6748-6755 is intentionally a
no-op for touch on mobile (Apple convention: dismiss via Done
or swipe-down). Our explicit onTapOutside override is the
correct way to opt into tap-outside-to-dismiss on mobile
without fighting iOS conventions or the framework's gesture
arena.
Wired onTapOutside on all three connect form fields (host /
nickname / password).
flutter build ios --release --no-codesign: 20.5 s, Runner.app
30.2 MB (unchanged). flutter analyze: 6 pre-existing Radio
deprecation infos (unchanged).
User reported on iPhone iOS 18: 'still a bit lag and stuck' after
the _kickFocus removal (1ceb47f). Web research (flutter/flutter
keyboard performance issues) and code review of GestureDetector
hit-test semantics identified the remaining race.
Root cause:
The outer GestureDetector wrapping the connect form Column was
using HitTestBehavior.translucent. Translucent semantics dispatch
the pointer event to BOTH the GestureDetector AND any descendant
hit-test target. So when the user tapped a TextField, two things
fired simultaneously:
1. GestureDetector.onTap -> FocusScope.of(context).unfocus()
This drove the keyboard *down* via the platform TextInput.hide
side effect of clearing focus.
2. TextField's own TapGestureRecognizer -> EditableText.attach
This drove the keyboard *up* via TextInput.show.
The two CAAnimations on iOS 18 raced each other inside the same
UIKit transaction, producing:
* 500-1000 ms of visible 'thinking' before the keyboard appeared
(one full slide-down + one full slide-up).
* Occasional 'stuck' state where the keyboard never came back up
because UIResponder.becomeFirstResponder was called before
resignFirstResponder finished.
Fix: switch to HitTestBehavior.opaque. With opaque:
* The GestureDetector still receives hit-test results for its
entire bounds (so taps on empty padding between fields still
reach onTap).
* But the gesture arena routes a tap that lands on a TextField
to that TextField's recognizer ONLY \u2014 the outer
GestureDetector loses the arena and its onTap does not fire.
* Net: tapping a field is exactly equivalent to having no outer
GestureDetector at all (no race, no lag, no stuck). Tapping
empty space still dismisses the keyboard cleanly.
This is the canonical pattern that several Stack Overflow answers
and the GestureDetector dartdoc recommend for 'tap-outside-to-
dismiss-keyboard'. translucent is for cases where you want both
the outer and inner to react simultaneously (rare).
flutter build ios --release --no-codesign: 29.0 s, Runner.app
30.2 MB.
User reported: 'amount need wait 500ms - 1s if i click input field
-> then keyboard popup'. The 79f8360 _kickFocus workaround was the
source of the lag.
Root cause of the lag:
void _kickFocus(FocusNode node) {
if (node.hasFocus) node.unfocus();
Future.microtask(() { // <-- this microtask
if (!mounted) return;
node.requestFocus();
});
}
The Future.microtask deferral forces EditableText's attach-to-
TextInput path to wait one frame past the user's pointer-up. iOS
26's keyboard slide-up animation then dovetails into that extra
frame in a way that adds another 200-800 ms before the keyboard
actually appears on screen. Net latency: ~500-1000 ms.
Fix: remove _kickFocus entirely. Rely on:
1. TextField's native onTap path (no onTap override = no
deferral, no microtask hop, no SystemChannels race).
2. The tap-outside-to-unfocus GestureDetector wrapping the
connect form Column (7c62d14) which already guarantees the
FocusNode is in the unfocused state when the user taps any
field, because any prior keyboard dismissal (tap outside / tap
a sibling field) goes through FocusScope.of(context).unfocus().
This means the FocusNode is always in a clean false state when a
TextField gets tapped, so EditableText's own attach path can fire
synchronously on the first frame and the keyboard appears
instantly.
The full _kickFocus implementation is retained as a code comment
above the connect-form's build() for documentation and quick
re-introduction should iOS regress again. The flutter/flutter#181474
issue (the underlying iOS 26 bug) remains open, so the comment
documents the canonical workaround if needed.
flutter analyze: 6 pre-existing Radio.groupValue deprecation infos
(unchanged). flutter build ios --release --no-codesign: 20.4 s,
Runner.app 30.2 MB.
Three user-reported issues addressed at once.
1. Audio output displaying as Unknown on iOS
The route tile only set _device from currentDeviceStream events,
which fire on route *changes*. On first sheet open with no route
change yet, _device was null \u2192 _deviceLabel fell through to
audioRouteUnknown.
Fix: query AudioRouterPlatform.instance.getCurrentDevice() in
initState before attaching the stream listener. Plugin returns the
current AVAudioSession route synchronously (well, via Future) so
the tile renders Speaker / iPhone receiver / AirPods / etc.
immediately on first open. Errors swallowed \u2014 the stream remains
authoritative for subsequent updates.
2. 'Adjust mode & release tail' too deep (chip \u2192 modal \u2192 button \u2192 dialog)
Inlined the mode radio buttons and release-tail slider directly
into the voice modal sheet. Dropped the OutlinedButton 'Adjust'
trigger and the nested VoiceSettingsDialog dispatch entirely on
mobile.
Modal sheet is now a single-screen control panel:
Title 'Voice'
--------
Audio output: <current route> > (iOS/Android only)
--------
Transmit mode
\u25c9 PTT
\u25cb Continuous
\u25cb Voice activity (Coming soon) (disabled)
--------
Release tail 200 ms
[\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u25cf\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501] (0\u20131000 ms, step 50)
Bound key: F (desktop only)
--------
Level meter
TX/RX frame counts
PTT capability badge (desktop only)
VoiceSettingsDialog is retained for the wide-mode VoiceBar
'configure' button (desktop entrypoint) and the PTT-bind flow, so
desktop UX is unaffected.
New widgets: _VoiceSheetBody (StatefulWidget with local _mode +
_tail), _ModeRow (RadioListTile-shaped row with optional disabled
state for VoiceActivity). New API on showVoiceDetailsSheet:
onModeChanged + onReleaseTailChanged callbacks (replace
onAdjustVoiceSettings). Wiring in main.dart writes through to
rust.setTransmitMode / rust.setReleaseTailMs and mirrors _state.
l10n: dropped voiceAdjustSettings (en + zh). Added voiceBoundKeyLabel
(en + zh) for the desktop-only bound-key row.
3. iOS first-tap-keyboard regression (flutter/flutter#181474)
The 79f8360 _kickFocus workaround (unfocus + microtask refocus on
every TextField.onTap) was kept, but extended with a tap-outside-
to-unfocus GestureDetector wrapping the connect form Column. This
guarantees the FocusNode is in the unfocused state when the next
field tap arrives, so the focus transition is always false\u2192true
on first tap.
GestureDetector(HitTestBehavior.translucent, onTap: unfocus) is the
canonical pattern recommended in the flutter/flutter#181474 thread
+ several older iOS keyboard issues. Translucent behaviour means it
catches taps on the column padding / empty regions without
swallowing taps on the TextFields themselves (those have
onTap: _kickFocus already).
flutter analyze: 6 pre-existing Radio.groupValue deprecation infos
in voice_settings.dart (unchanged). flutter build ios --release
--no-codesign: 27.9 s, Runner.app 30.2 MB (unchanged).
Awaiting iPhone retest to confirm all three fixes.
User reported: 'input field still need to click twice' on iPhone +
'a bit lag after keyboard pop up'. The previous Listener-based
approach (020be77 \u2192 23bd1c7) actively made both symptoms worse.
Root cause is a confirmed open Flutter framework bug:
flutter/flutter#181474 \u2014 [iPadOS] Keyboard is dismissed, but the
TextField keeps focus, causing subsequent taps not to trigger
keyboard presentation.
Open, P2, triaged-text-input, platform-ios, e: OS-version specific.
Reported on iPadOS 26.2 + Flutter 3.38.1 in Jan 2026 by
Crazymuyang.
Reproduction matches our symptom exactly: iOS 26 dismisses the soft
keyboard (e.g. on tap-outside, or in some fresh-launch states), but
the EditableText's FocusNode keeps hasFocus = true. Because the
node is already focused, the next user tap is a no-op from the
focus system's perspective, so the platform TextInput channel is
never re-opened and iOS keeps the soft keyboard hidden until a
second tap finally triggers an explicit re-focus path.
Why our previous attempts failed:
* 020be77 wrapped each TextField in a Listener that called
requestFocus pre-arena. That's racing the wrong layer \u2014 it
doesn't help when the bug is 'node is already focused, so
requestFocus is a no-op'.
* 23bd1c7 added SystemChannels.textInput.invokeMethod('TextInput.show')
to the same Listener. This forced the keyboard up but raced
EditableText's own attach path, producing the post-attach
typing lag the user reported.
Fix: the community-recommended workaround in the issue thread \u2014
on every TextField tap, **unfocus first, then re-request focus on
the next microtask**. This forces a real false\u2192true focus-change
transition that re-opens TextInput on the first tap. No platform
channel races, no gesture-arena fighting, no Listener wrappers.
void _kickFocus(FocusNode node) {
if (node.hasFocus) node.unfocus();
Future.microtask(() {
if (!mounted) return;
node.requestFocus();
});
}
TextField(
onTap: () => _kickFocus(_hostFocus),
...
)
Wired into all three connect-form fields (host / nickname /
password). Removed the now-redundant _focusOnTap Listener helper.
No-op on hosts where #181474 doesn't reproduce \u2014 the unfocus call
is a no-op when the node isn't focused, and the microtask
requestFocus is what TextField would have done anyway via its own
TapGestureRecognizer.
flutter analyze: 6 pre-existing Radio.groupValue deprecation infos
in voice_settings.dart (unchanged). flutter build ios --release
--no-codesign: 27.6 s, Runner.app 30.2 MB.
User reported: 'input field still need to click twice' on iPhone.
The 020be77 Listener + requestFocus() approach was insufficient.
Root cause analysis:
* Flutter's EditableText opens the platform TextInput method
channel (which is what slides the iOS soft keyboard up) only
after a TapGestureRecognizer wins the gesture arena.
* Our outer Listener calls node.requestFocus() pre-arena. That
flags the FocusNode as focused in Flutter's focus tree, but
does NOT open the TextInput channel \u2014 so iOS keeps the OS
keyboard hidden until EditableText's own tap recognizer wins
on a second tap.
* requestFocus on its own is therefore a no-op for the user
visually: the cursor + caret appear briefly but the keyboard
stays down.
Fix: in the Listener.onPointerDown handler, additionally invoke
'TextInput.show' on SystemChannels.textInput. This is the same
private platform RPC EditableText calls internally on attach;
forcing it ourselves slides the keyboard up regardless of arena
state.
Belt-and-braces with the existing requestFocus() guarantees
keyboard-on-first-tap on iPhone / iPad and is harmless on:
* Android (the platform ignores the redundant show call when
the keyboard is already up),
* Linux / Windows / macOS desktop (no soft keyboard exists; the
method channel handler returns success without doing
anything).
No new imports needed \u2014 SystemChannels is already in
package:flutter/services.dart (imported for FilteringTextInputFormatter).
flutter build ios --release --no-codesign: 20.7 s, Runner.app
30.2 MB.
Two user-reported issues addressed:
1. 'on supported devices such as iphone or android user should be
able to select audio device such as speaker or airpods or phone'
2. 'the bottom folder is duplicated with the app bar settings'
Issue 1 \u2014 audio output route picker:
Added the audio_router 1.1.1 plugin (MIT, supports iOS + Android)
which renders the platform-native picker:
* iOS: Apple AVRoutePickerView system sheet \u2014 the same UI as
Control Center's audio chooser. Lists Speaker / iPhone receiver /
AirPods / connected Bluetooth devices / AirPlay / CarPlay.
System manages the device list; we don't have to track route
changes manually.
* Android (post-rc.8 when we wire the platform): Material Design 3
dialog backed by AudioManager.setCommunicationDevice() with
SCO Bluetooth + USB headsets filtered for VoIP.
The picker prerequisite documented by the plugin (audio session
must be playAndRecord/voiceChat before the picker fires) is already
satisfied by our AppDelegate.swift configuration from commit 0466000.
The new _AudioOutputTile widget in voice_compact.dart subscribes to
AudioRouter.currentDeviceStream so the row label + icon auto-update
when the user plugs in headphones, connects AirPods, etc. \u2014 no
manual KVO observation needed.
Tile is mobile-only (Platform.isIOS || Platform.isAndroid). Desktop
hosts continue to use the system mixer; the tile is hidden.
Plugin caveat: the published enum AudioSourceType has no .usb
variant (despite README mentioning USB support). We map only the
seven real enum cases: builtinSpeaker / builtinReceiver / bluetooth /
wiredHeadset / carAudio / airplay / unknown.
Issue 2 \u2014 collapse voice controls into the single modal sheet:
The AppBar gear icon (Icons.tune) that opened VoiceSettingsDialog
was removed. It duplicated the configuration entry point that the
status chip \u2192 modal-sheet path already provides, and the user found
that duplication confusing on a phone-narrow screen where AppBar
real estate is precious.
The voice modal sheet (showVoiceDetailsSheet) is now the **single**
voice-controls surface on mobile, with layout (top to bottom):
1. Audio output route picker tile (iOS / Android only) \u2014 new.
2. Mode + bind / release-tail recap (display only).
3. 'Adjust mode & release tail' OutlinedButton that closes the
sheet and opens the same VoiceSettingsDialog the gear icon
used to open. One config form, not two.
4. Mic level meter.
5. TX / RX frame counts + mic state.
6. PTT capability badge (desktop only).
Sheet title renamed from 'Voice settings' (which collided with the
gear-icon tooltip) to 'Voice'. New l10n keys: voiceSheetTitle,
voiceAdjustSettings, audioOutputLabel, audioRoute{Speaker,Receiver,
Bluetooth,WiredHeadset,CarAudio,Airplay,Unknown}. en + zh translated.
Build: flutter build ios --release --no-codesign clean, 50.5 s,
Runner.app 30.2 MB (+200 KB from audio_router). flutter analyze
clean (6 pre-existing Radio.groupValue deprecation infos in
voice_settings.dart, unchanged).
The server-host TextField on the connect form accepts a hostname or
hostname:port pair (e.g. kr.teamspeak.app:9987). Two improvements:
1. keyboardType: TextInputType.url surfaces '.', '/', ':' on the
primary on-screen keyboard plane so the user does not have to
switch to the symbols pane mid-address. Matches iOS Safari's
URL bar.
2. textCapitalization.none + autocorrect/enableSuggestions=false
prevents iOS from auto-capitalising the first letter or
'correcting' 'kr.teamspeak.app' to something else.
3. inputFormatters belt-and-braces:
* deny whitespace (handles tab-indented paste)
* lowercase pipeline (handles uppercase paste)
4. Visual affordances: prefix dns icon + 'host[:port]' hint.
Nickname field intentionally unchanged \u2014 may contain unicode,
mixed case, spaces.
flutter build ios --release --no-codesign: 30.7 s, Runner.app
30.0 MB. flutter analyze clean (6 pre-existing deprecation
warnings on Radio.groupValue/onChanged).
Xcode warning on iOS SDK 26+:
'allowBluetooth' was deprecated in iOS 8.0: renamed to
'AVAudioSession.CategoryOptions.allowBluetoothHFP'
The flag was renamed in iOS 8 (a decade ago) but the old name has
been kept as a soft-deprecated alias. iOS 26 SDK finally emits the
warning, and -Werror builds would fail on it. Same semantics:
permit HFP-profile Bluetooth headsets as input + output. Kept
.allowBluetoothA2DP alongside for higher-quality output-only A2DP
devices.
flutter build ios --release --no-codesign: 10.7 s, Runner.app
30.0 MB.
From iPhone log:
chanora_flutter: AVAudioSession setup failed:
Error Domain=NSOSStatusErrorDomain Code=561017449
'Session activation failed'
Error code 561017449 = AVAudioSessionErrorCodeCannotStartPlaying
(ASCII '!cat' big-endian). iOS 17+ refuses setActive(true) calls
made before the app's scene is foregrounded: the audio policy
server denies the activation because the app is not yet considered
the foreground priority owner. didFinishLaunchingWithOptions runs
BEFORE the scene becomes .active, so synchronous activation there
hits this race on cold launch.
Symptom flow:
1. App cold-launch -> AppDelegate.didFinishLaunching fires
2. setActive(true) -> Error 561017449
3. Audio session is left inactive
4. cpal's later attempts to open RemoteIO see an inactive
session and reject with StreamConfigNotSupported
5. voice_join fails at ensure_audio_running
6. user sees the audio failure manifested as missing mute /
continuous / PTT buttons (now fixed in f1f81a3 to be
lenient; this commit also unblocks the underlying audio).
Fix: split the AVAudioSession configuration into two phases:
* setCategory at didFinishLaunching (always safe).
* setActive(true) deferred to UIApplication.didBecomeActive Notification, which fires after the cold-launch settle and
on every resume-from-background. Repeated setActive while
already-active is a no-op per docs.
This is the canonical iOS voice-app pattern (Discord, Zoom,
FaceTime, Flutter's package all follow it). Documented
in commit body comments.
flutter build ios --release --no-codesign: 13.2 s, Runner.app
30.0 MB.
Two distinct fixes prompted by user reports from the iPhone build:
1. 'strange text on Chanora (RFLOWED BY)' \u2014 the Flutter debug
overlay's 'OVERFLOWED BY N PIXELS' strip was appearing next to
the AppBar title because the title Row ('Chanora' + channel
pill) plus 5-6 trailing IconButton actions exceeded a typical
iPhone AppBar width. User saw the strip clipped to '...RFLOWED
BY...' since only its end fit on screen.
_AppBarTitle now drops the 'Chanora' label on narrow widths
(<840 dp). Title shows only the channel pill when in voice
channel; the user already knows they're in Chanora because
they just opened it. Wide widths (tablet/desktop, >= 840 dp)
keep the full 'Chanora \u00b7 #channel-pill' title because there's
room. Eliminates the overflow.
Note: the OVERFLOWED-BY strip only renders in debug builds
anyway; release builds suppress the overlay. But the
underlying Row overflow was a real layout bug worth fixing.
2. First-tap TextField still failed on iPhone after the earlier
FocusNode + TextField.onTap fix. Root cause: TextField.onTap
fires AFTER the gesture-arena resolves, so if the enclosing
SingleChildScrollView wins the arena (which it does on iOS
for the very first tap), the focus request never fires.
Wrap each connect-form TextField in a Listener with
HitTestBehavior.translucent and onPointerDown: requestFocus.
Listener fires synchronously on PointerDownEvent BEFORE arena
resolution, so even if the scrollable would have won the arena
we have already grabbed focus. Translucent means the pointer
ALSO propagates down to the TextField so its normal touch
handling still runs (text selection / cursor placement).
_focusOnTap helper added; wraps all three TextFields
(host, nick, password).
flutter analyze: clean (6 pre-existing Radio.groupValue infos).
flutter build ios --release --no-codesign: 18.4 s, Runner.app
30.0 MB.
The podspec's prepare_command only fires on `pod install`. Once a
framework was generated, Rust source changes were silently ignored
because Xcode kept re-bundling the stale framework into Runner.app.
Manifested today as: ran `cargo build --release --target
aarch64-apple-ios` to pick up the lenient voice_join fix, ran
`flutter build ios`, but Runner.app/Frameworks/chanora_bridge.
framework/chanora_bridge was still the framework from the previous
pod install (16:10) not the just-built 18:30 dylib.
Add an explicit `script_phase` to the podspec that re-runs:
1. cargo build --release --target aarch64-apple-ios -p chanora_bridge
2. cp dylib into Frameworks/chanora_bridge.framework/chanora_bridge
3. install_name_tool -id @rpath/...
on every Xcode 'Build', not just on pod install. The script short-
circuits when the framework's binary mtime is newer than the cargo
output (fast no-op on incremental builds where Rust didn't change).
Side effect: every Xcode build now invokes cargo, which can take
~5 s on a warm cache and ~1 min cold. This is the right trade-off
because the previous behavior silently shipped stale Rust code.
User report from iPhone: 'mute / continuous / PTT buttons missing'
with NO error popup. Root cause: voice_join's audio-engine startup
was failing silently, and the failure propagated out as a hard
error \u2014 which means SessionEvent::VoiceState(true) was never sent
to Dart even though the server-side channel move had already
succeeded. Dart's _inChannel stayed false; every control gated on
_inChannel disappeared while the channel tree continued to show
the user as joined.
This commit makes voice_join lenient on audio-engine failures so
the UI state matches the server-side reality, and also gives iOS
a writable log file so diagnostics from device builds are
recoverable for the first time.
core/chanora_core/src/lib.rs::voice_join
* ensure_audio_running's error is now logged + emitted as
SessionEvent::AudioStopped, but does NOT abort voice_join.
The server move at step 1 already succeeded; failing the
Dart-visible promise here would leave the UI in a phantom
'in-channel visually but no controls' state. After this
commit:
- mic / headset / settings appear in the AppBar
- PTT button appears at the bottom
- status chip shows live audio state ('Mic on/off')
- if audio actually failed (mic permission denied,
no input device, CoreAudio rejecting stream config)
the user can retry by switching modes / channels;
BridgeEvent::AudioStopped wires _audioStarted=false
in Dart so audio-stats poll is honest about the
engine state.
crates/chanora_bridge/src/api.rs::log_file_path
* iOS now writes the log to /home/milkice/Documents/chanora.log
(Documents is the standard user-visible iOS sandbox dir).
* Android remains None pending the bridge JNI init wiring
a writable path (P1 follow-up).
apps/chanora_flutter/ios/Runner/Info.plist
* Adds UIFileSharingEnabled + LSSupportsOpeningDocumentsInPlace
so the Documents directory shows up under 'On My iPhone \u2192
Chanora' in the Files.app. The user can now copy chanora.log
out for support without needing Xcode \u2192 Devices and
Simulators \u2192 Download Container.
Workspace tests: 78/0/1 unchanged.
flutter build ios --release --no-codesign: 22.7 s clean
(Runner.app 29.9 MB).
Restructure the narrow / mobile body layout around three principles
(Hoober thumb-zone research validated, Material 3 components):
1. Channel tree gets ~85% of the screen height.
2. Live voice state is always visible in a 2-line status chip
above the PTT button.
3. PTT button is wide, bottom-anchored (thumb-natural lower zone).
4. Frequent toggles (mic mute, headset mute, voice settings) live
in the AppBar so they don't compete with the channel tree.
5. Non-essential live readouts (level meter, TX/RX, capability
badge) live in a modal sheet opened by tapping the status chip
-- progressive disclosure.
apps/chanora_flutter/lib/widgets/voice_compact.dart (new file)
* VoiceStatusChip: 2-line live readout. Line 1 = mode + bind hint;
Line 2 = release-tail + 'Mic on/off'. Tap opens
showVoiceDetailsSheet.
* VoicePttButton: 56 dp wide bottom-anchored Push to Talk button.
Same touch-and-hold gestures as the previous _PttHoldButton.
* showVoiceDetailsSheet: modal bottom sheet with mode recap +
bind/tail hint + level meter + TX/RX counts + PTT capability
badge (desktop-only).
apps/chanora_flutter/lib/main.dart
* AppBar gains mic-mute, headset-mute, voice-settings icons when
in voice channel AND MediaQuery width < 840 dp (mobile only).
Wide mode keeps these controls inside the existing VoiceBar
widget unchanged.
* AppBar title becomes a Row of 'app name + channel chip' when
in voice channel.
* Narrow-mode body restructured: Expanded(channelTree) +
VoiceStatusChip + VoicePttButton (latter only when in voice
channel AND PTT mode). The old narrow-mode VoiceBar is gone;
wide-mode VoiceBar is unchanged.
* New _onOpenVoiceDetailsSheet handler bridges the chip-tap to
showVoiceDetailsSheet.
* New top-level helper _isTouchOnlyPttHost mirrors the helpers
in widgets/voice_bar.dart and widgets/voice_settings.dart so
the AppBar + narrow-mode chip can branch consistently.
apps/chanora_flutter/lib/l10n/app_en.arb
apps/chanora_flutter/lib/l10n/app_zh.arb
apps/chanora_flutter/lib/l10n/generated/* (regenerated)
* New string voicePttHoldHint = 'Hold the button' / '按住按钮'.
Surfaced in line 1 of VoiceStatusChip on touch-only hosts and
in the modal sheet's PTT line where the desktop equivalent
would name a bound key.
Wide-mode (>= 840 dp) layout intentionally unchanged so the
signed-off rc.8 desktop verification still applies.
flutter analyze: clean (6 pre-existing Radio.groupValue infos).
Local cargo + flutter analyze pass; Mac was offline at commit
time so iOS device build verification is pending the next sync.
Edison Jwa