@@ -15,6 +15,59 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [1.9.9] - 2026-06-26 - "Workshop" (iOS creator / power tools + readiness gate)
+
+The final iOS TestFlight release before the v2.0.0 "Timebase" core rewrite — the
+analogue of Android v1.8.9. It brings the desktop creator/power-tools to touch and
+runs a full pre-freeze readiness pass (verification, security audit, gap analysis, a
+completeness-critic sweep). Everything is **additive / off-by-default**: the bridge
+gains only new forwarding fns over **existing** `rustynes-core` APIs (no signature or
+behaviour change), the deterministic core / chip crates are untouched, and the
+shipped / native / `no_std` / wasm core stays **byte-identical** — **AccuracyCoin
+holds 100% (139/139)**. Interim TestFlight.
+
+### Added
+
+- **Cheats** — a Game Genie editor (persistent live codes via the core engine,
+  applied on every PRG read) plus one-shot raw-RAM peek / poke of the `$0000–$1FFF`
+  work-RAM region, behind a new Cheats sheet.
+- **Read-only debugger inspector** (FOSS / TestFlight build only, gated off the
+  App-Store channel per ADR 0027) — live CPU registers, a disassembly window around
+  the PC, a hex memory view, and single-frame step, over new observational bridge
+  fns (`debug_cpu_state` / `debug_read_memory` / `debug_disassemble`).
+- **TAStudio piano-roll** — a touch frame-input table with deterministic scripted
+  playback and record-to-`.rnm` from power-on.
+- **Foreign movie import** — `.fm2` / `.bk2` / `.fcm` / `.fmv` / `.vmv` transcoded to
+  the native `.rnm` format via new bridge importers (bounds-checked: 16 MiB
+  per-ZIP-member cap, graceful errors on malformed input — never panics).
+- **Audio depth** — a host-side DSP stage in `rustynes-ios` (5-band EQ, per-channel
+  panning, Schroeder reverb, crossfeed), fully bypassable (disabled / flat / centered
+  is bit-for-bit the v1.9.8 passthrough; the cpal callback drains one sample per
+  output frame regardless of state).
+- **Symbol maps** — host-side `.sym` / `.mlb` / `.nl` label loading feeding the
+  debugger view.
+
+### Changed
+
+- **`rustynes-mobile` bridge** gains additive forwarding fns (cheats, foreign-movie
+  importers, observational debugger reads) — the first iOS release to extend the
+  shared bridge surface; all additive, so the Android build is unaffected and the
+  core stays byte-identical.
+
+### Fixed
+
+- File imports (ROM, palette, symbol map) and foreign-movie transcoding now run off
+  the main thread (`Task.detached`), so a large user-picked file no longer hangs the
+  UI; `poke_ram` clamps the address to the work-RAM mirror bridge-side; missed
+  EN/ES catalog strings and VoiceOver labels across the new creator-tool views added.
+
+### Notes
+
+- **Mobile ROM loading is iNES / NES 2.0 only.** The import picker no longer
+  advertises `.fds` / `.nsf` (the shared bridge has no FDS / NSF load path) — FDS
+  disk + NSF playback on mobile are a post-v2.0.0 carryover, alongside the 20-band EQ,
+  `.dbg` ca65/cc65 source maps, and a cheats database.
+
 ## [1.9.8] - 2026-06-26 - "Horizon" (iOS store-readiness)
 
 The iOS store-readiness release — the analogue of Android v1.8.8 "Atlas". A broad
 
@@ -21,7 +21,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.9.8"
+version = "1.9.9"
 edition = "2024"
 rust-version = "1.96"
 license = "MIT OR Apache-2.0"
 
@@ -39,8 +39,8 @@ android {
         applicationId = "com.doublegate.rustynes"
         minSdk = 26 // AAudio floor.
         targetSdk = 36 // Play mandate (Android 16) from 2026-08-31.
-        versionCode = 10908 // 1.9.8
-        versionName = "1.9.8"
+        versionCode = 10909 // 1.9.9
+        versionName = "1.9.9"
         // No abiFilters here — set per buildType so release ships arm64 only
         // while debug keeps x86_64 for the emulator.
         // PLAY_BUILD gates the freemium (demo timer + persistence locks + Billing).
 
@@ -72,6 +72,16 @@ void rustynes_ios_audio_push(RustyNesAudioSink *handle, const float *samples, si
 /* The negotiated device sample rate (request it from NesController::new). 0 if NULL. */
 uint32_t rustynes_ios_audio_sample_rate(RustyNesAudioSink *handle);
 
+/* Publish the live audio-depth DSP config (v1.9.9): EQ / pan / reverb / crossfeed,
+ * applied host-side after drain_audio. `eq` is up to 5 band gains (dB), `pan` up to
+ * 6 per-channel positions (-1..=1); extra entries ignored, missing ones neutral.
+ * `enabled == 0` (or a flat / centered config) is a bit-exact passthrough. The
+ * `eq` / `pan` pointers may be NULL (treated as neutral). No-op on a NULL handle. */
+void rustynes_ios_audio_set_depth(RustyNesAudioSink *handle, uint8_t enabled,
+                                  const float *eq, size_t eq_len,
+                                  const float *pan, size_t pan_len,
+                                  float reverb_mix, float reverb_room, float crossfeed);
+
 /* Pause / resume the output stream (scene background / audio interruption). */
 void rustynes_ios_audio_pause(RustyNesAudioSink *handle);
 void rustynes_ios_audio_resume(RustyNesAudioSink *handle);
 
@@ -24,6 +24,8 @@ use std::sync::Arc;
 
 use cpal::traits::{DeviceTrait, HostTrait, StreamTrait};
 
+use crate::audio_dsp::{AudioDepth, DepthConfig, DepthParams};
+
 /// A single-producer / single-consumer lock-free ring of mono `f32` samples.
 ///
 /// The producer ([`AudioSink::push`], called on the emu thread) only advances
@@ -122,6 +124,12 @@ pub struct AudioSink {
     ring: Arc<Ring>,
     stream: cpal::Stream,
     sample_rate: u32,
+    /// The live audio-depth (EQ / pan / reverb / crossfeed) configuration mailbox
+    /// (v1.9.9). The Swift Settings UI publishes a [`DepthConfig`] via
+    /// [`AudioSink::set_depth`]; the real-time callback snapshots it once per
+    /// buffer. Default / disabled = bit-exact passthrough, so the determinism
+    /// contract is untouched.
+    depth_params: Arc<DepthParams>,
 }
 
 impl AudioSink {
@@ -148,16 +156,46 @@ impl AudioSink {
         let ring = Arc::new(Ring::new((sample_rate as usize) / 4));
         let ring_cb = Arc::clone(&ring);
 
+        // The audio-depth DSP (v1.9.9): the callback owns a stateful processor
+        // (EQ biquad history + reverb delay lines, allocated once) and reads the
+        // live config from the shared `DepthParams` mailbox once per buffer.
+        let depth_params = Arc::new(DepthParams::new());
+        let depth_params_cb = Arc::clone(&depth_params);
+        let mut depth = AudioDepth::new(sample_rate);
+
         let stream = device
             .build_output_stream(
                 config,
                 move |data: &mut [f32], _: &cpal::OutputCallbackInfo| {
-                    // Expand the mono ring to the device's interleaved channels:
-                    // one queued sample fans out to every channel of a frame.
+                    // Mirror the live config in once per buffer (cheap; re-voices
+                    // only on an actual change), then check bypass once.
+                    depth.apply(&depth_params_cb.snapshot());
+                    let bypass = depth.is_bypass();
                     for frame in data.chunks_mut(channels) {
-                        let s = ring_cb.pop();
-                        for ch in frame.iter_mut() {
-                            *ch = s;
+                        // Exactly one queued mono sample is consumed per output
+                        // frame regardless of bypass, so toggling the DSP never
+                        // changes the drain rate.
+                        let mono = ring_cb.pop();
+                        if bypass {
+                            // Fan the mono value out to every channel (the
+                            // byte-identical pre-v1.9.9 behaviour).
+                            for ch in frame.iter_mut() {
+                                *ch = mono;
+                            }
+                        } else {
+                            let (l, r) = depth.process(mono);
+                            match frame {
+                                [] => {}
+                                [c0] => *c0 = 0.5 * (l + r),
+                                [c0, c1, rest @ ..] => {
+                                    *c0 = l;
+                                    *c1 = r;
+                                    // Any surround channels get the left image.
+                                    for ch in rest {
+                                        *ch = l;
+                                    }
+                                }
+                            }
                         }
                     }
                 },
@@ -171,6 +209,7 @@ impl AudioSink {
             ring,
             stream,
             sample_rate,
+            depth_params,
         })
     }
 
@@ -195,4 +234,11 @@ impl AudioSink {
     pub fn sample_rate(&self) -> u32 {
         self.sample_rate
     }
+
+    /// Publish a new audio-depth (EQ / pan / reverb / crossfeed) configuration
+    /// (v1.9.9). Lock-free: the real-time callback picks it up on its next buffer.
+    /// A default / disabled config is a bit-exact passthrough.
+    pub fn set_depth(&self, cfg: &DepthConfig) {
+        self.depth_params.store(cfg);
+    }
 }
Original file line number	Diff line number	Diff line change
`@@ -21,7 +21,7 @@ members = [`
`21`	`21`	`]`
`22`	`22`
`23`	`23`	`[workspace.package]`
`24`		`-version = "1.9.8"`
	`24`	`+version = "1.9.9"`
`25`	`25`	`edition = "2024"`
`26`	`26`	`rust-version = "1.96"`
`27`	`27`	`license = "MIT OR Apache-2.0"`