chore(deps): bump symphonia 0.5 → 0.6 + harden migration checksums against line-ending drift

CLAUDE.md

@@ -51,7 +51,7 @@ Rust / Tauri 2. Entry: `src-tauri/src/main.rs` → `lib.rs`.
 
 - **Commands** (`src-tauri/src/commands/`): organized by domain — `library.rs`, `playlist.rs`, `smart_playlists.rs`, `track.rs`, `browse.rs`, `player.rs`, `scan.rs`, `edit.rs`, `profile.rs`, `analysis.rs`, `deezer.rs`, `similar.rs`, `lyrics.rs`, `stats.rs`, `wrapped.rs`, `integration.rs`, `maintenance.rs`, `app_info.rs`, `radio.rs`, `mood_radio.rs`, `duplicates.rs`, `preferences.rs`, `share.rs`, `changelog.rs`, etc. All registered in `lib.rs` via `generate_handler![]`.
 - **External API clients** (crate-root modules): `deezer.rs` (public Deezer, no auth), `lastfm.rs` (`artist.getInfo`, user API key required). Both use `reqwest` with `rustls-tls`.
-- **Audio engine** (`src-tauri/src/audio/`): 3-thread lock-free architecture — `decoder.rs` (symphonia + rubato), `output.rs` (cpal callback on a dedicated thread, SPSC `rtrb` ring buffer), `state.rs` (`SharedPlayback` with atomics, no locks in hot path), `analytics.rs` (tokio task for `play_event` writes + auto-advance), `crossfade.rs`, `eq.rs`, `spectrum.rs`, `wasapi_exclusive.rs` (Windows-only opt-in), and `dsd/` (in-house DSF/DFF parser + DSD→PCM converter, Symphonia 0.5 doesn't decode DSD). Deep dive: [`docs/features/playback.md`](docs/features/playback.md).
+- **Audio engine** (`src-tauri/src/audio/`): 3-thread lock-free architecture — `decoder.rs` (symphonia + rubato), `output.rs` (cpal callback on a dedicated thread, SPSC `rtrb` ring buffer), `state.rs` (`SharedPlayback` with atomics, no locks in hot path), `analytics.rs` (tokio task for `play_event` writes + auto-advance), `crossfade.rs`, `eq.rs`, `spectrum.rs`, `wasapi_exclusive.rs` (Windows-only opt-in), and `dsd/` (in-house DSF/DFF parser + DSD→PCM converter, Symphonia doesn't decode DSD). Deep dive: [`docs/features/playback.md`](docs/features/playback.md).
 - **DLNA / UPnP MediaServer** (`src-tauri/src/dlna/`): worker thread → axum HTTP server + SSDP announcer. Opt-in (`app_setting['dlna.enabled']`, default OFF). See [`docs/features/dlna.md`](docs/features/dlna.md).
 - **OS media controls** (`media_controls.rs`): souvlaki bridge → SMTC / MPRIS / MediaRemote. Initialized post-window (needs HWND on Windows).
 - **Discord Rich Presence** (`discord_presence.rs`): named-pipe IPC, opt-in `app_setting['integrations.discord_rpc']` (default ON). See [`docs/features/integrations.md`](docs/features/integrations.md#discord-rich-presence).
@@ -68,7 +68,7 @@ These bite you if you ignore them — they're the contract the rest of the codeb
 - **Persistence**: per-profile settings live in `profile_setting` (key-value, typed). Pattern: `INSERT ... ON CONFLICT DO UPDATE`. App-wide settings live in `app_setting` with the same shape.
 - **Events**: backend emits Tauri events (`player:state`, `player:position`, `player:track-changed`, `player:queue-changed`, `player:error`, `player:ab-loop`, `player:spectrum`, `track:updated`, `library:rescanned`, `scan:progress`, `lyrics:updated`, …). Frontend listens via `listen()` from `@tauri-apps/api/event`.
 - **Audio callback is hot**: the cpal callback (and the WASAPI exclusive thread) MUST NOT allocate, lock, or log. Only `rtrb::Consumer` reads + `Atomic*` loads. All heavy work (EQ, ReplayGain, resampling, FFT, BLAKE3) runs on the decoder thread before samples reach the SPSC ring.
-- **Migrations are immutable once merged**: sqlx records a SHA-384 checksum in `_sqlx_migrations.checksum` at apply time, so editing a merged migration crashes every existing install at boot with `"migration <id> was previously applied but has been modified"` (no auto-recovery — user has to wipe their data dir). For any schema evolution, **create a new dated migration** `YYYYMMDDhhmmss_<slug>.sql`. Same rule for `migrations/app/`.
+- **Migrations are immutable once merged**: sqlx records a SHA-384 checksum in `_sqlx_migrations.checksum` at apply time, so editing a merged migration crashes every existing install at boot with `"migration <id> was previously applied but has been modified"`. For any schema evolution, **create a new dated migration** `YYYYMMDDhhmmss_<slug>.sql`. Same rule for `migrations/app/`. **Line-ending drift is a non-event** — [`db::migration_heal`](src-tauri/src/db/migration_heal.rs) reconciles stored checksums against the compiled-in migrator before each `Migrator::run`: when the stored hash matches the LF or CRLF variant of the same SQL (Windows `core.autocrlf=true` regression), it silently rewrites the row to the canonical hash and logs a warning. A real SQL change still panics, because neither LF nor CRLF normalization will rescue it.
 - **Virtual scroll everywhere**: TrackTable uses `@tanstack/react-virtual` for 6000+ track performance. Virtualized tables consume `usePageScroll()` for the scroll element instead of nesting their own `overflow-y-auto` — drives a single Spotify-style scrollbar.
 - **Multi-artist queries**: the scanner splits `"A, B"` on `", " / "; "` into individual `artist` rows linked via `track_artist`. Queries rebuild the display string via `GROUP_CONCAT` over `track_artist` ordered by `position`. `ArtistLink` accepts parallel `artist_name` + `artist_ids` strings so every contributor is individually clickable. New track queries must follow the same join pattern.
 - **Album grouping = `(canonical_title, album_artist_id)`**: [`scan.rs::upsert_album`](src-tauri/src/commands/scan.rs) keys on the album artist (Album Artist tag → `is_compilation` → primary artist fallback). `album.is_compilation` is sticky and `merge_implicit_compilations` collapses ≥ 3 distinct-artist same-title rows into "Various Artists" after every scan. `edit.rs` re-runs `upsert_album` with the OLD album's Album Artist / compilation flags so renames don't re-split. Deep dive: [`docs/features/library.md`](docs/features/library.md#album-grouping).

README.md

@@ -97,7 +97,7 @@ Once installed (any of the above), the in-app updater fetches future versions au
 | **Discord Rich Presence** | discord-rich-presence 1.1 (local IPC named pipe, no auth)                                                                          |
 | **Frontend**              | React 19, TypeScript, Vite 8, Tailwind CSS 4, Lucide icons, `@dnd-kit` (drag-and-drop), `@tanstack/react-virtual` (virtualization) |
 | **Backend**               | Rust, SQLite (sqlx), FTS5 contentless full-text search                                                                             |
-| **Audio**                 | symphonia 0.5 (decode), cpal 0.17 (output), rubato 2.0 (resample), rtrb 0.3 (SPSC ring)                                            |
+| **Audio**                 | symphonia 0.6 (decode), cpal 0.17 (output), rubato 2.0 (resample), rtrb 0.3 (SPSC ring)                                            |
 | **Metadata extraction**   | lofty 0.24 (tags, embedded art, POPM, INITIALKEY)                                                                                  |
 | **Imaging**               | image 0.25 + fast_image_resize 6 (SIMD thumbnails)                                                                                 |
 | **Filesystem watcher**    | notify 8 (debounced rescans of watched folders)                                                                                    |

docs/features/playback.md

@@ -4,8 +4,8 @@ The audio path lives in [`src-tauri/src/audio/`](../../src-tauri/src/audio). It
 
 ## Decoding & output
 
-- **Decoder** — [`symphonia 0.5`](https://crates.io/crates/symphonia) over MP3, FLAC, WAV, OGG Vorbis, AAC, ALAC (M4A). Source samples are converted to interleaved `f32`, channel-mapped (mono ↔ stereo, 5.1 → stereo Lo/Ro per ITU BS.775), then resampled to the device rate by [`rubato 2.0`](https://crates.io/crates/rubato) (`Fft<f32>` + `FixedSync::Input`, with a fast `Passthrough` variant when source rate already matches the device).
-- **DSD pipeline** — symphonia 0.5 doesn't decode 1-bit DSD, so DSF (Sony) and DFF (Philips) containers route through [`audio/dsd/`](../../src-tauri/src/audio/dsd/): a custom container parser reads the layout (DSD64 → DSD1024, mono / stereo / multichannel), and a 256-tap windowed-sinc FIR with a Blackman-Harris envelope decimates the bitstream by 64 to land DSD64 at 44.1 kHz, DSD128 at 88.2 kHz, etc. The resulting PCM joins the same channel-convert + resample + ring-buffer pipeline as symphonia output. `ActiveStream` carries a `StreamBackend` enum (Symphonia / Dsd) so seeking and decoder reset stay uniform from the engine's perspective. **Limitation**: real audiophile players use multi-stage halfband cascades for lower CPU at the same SNR; ours prioritises code clarity. DoP (DSD-over-PCM) is not yet wired — the converter always produces PCM.
+- **Decoder** — [`symphonia 0.6`](https://crates.io/crates/symphonia) over MP3, FLAC, WAV, OGG Vorbis, AAC, ALAC (M4A). Source samples are converted to interleaved `f32`, channel-mapped (mono ↔ stereo, 5.1 → stereo Lo/Ro per ITU BS.775), then resampled to the device rate by [`rubato 2.0`](https://crates.io/crates/rubato) (`Fft<f32>` + `FixedSync::Input`, with a fast `Passthrough` variant when source rate already matches the device).
+- **DSD pipeline** — symphonia doesn't decode 1-bit DSD, so DSF (Sony) and DFF (Philips) containers route through [`audio/dsd/`](../../src-tauri/src/audio/dsd/): a custom container parser reads the layout (DSD64 → DSD1024, mono / stereo / multichannel), and a 256-tap windowed-sinc FIR with a Blackman-Harris envelope decimates the bitstream by 64 to land DSD64 at 44.1 kHz, DSD128 at 88.2 kHz, etc. The resulting PCM joins the same channel-convert + resample + ring-buffer pipeline as symphonia output. `ActiveStream` carries a `StreamBackend` enum (Symphonia / Dsd) so seeking and decoder reset stay uniform from the engine's perspective. **Limitation**: real audiophile players use multi-stage halfband cascades for lower CPU at the same SNR; ours prioritises code clarity. DoP (DSD-over-PCM) is not yet wired — the converter always produces PCM.
 - **Output** — [`cpal 0.17`](https://crates.io/crates/cpal) on a dedicated thread because `cpal::Stream` is `!Send` on Windows. Samples cross the thread via an [`rtrb 0.3`](https://crates.io/crates/rtrb) SPSC ring (`RING_CAPACITY = 96 000` `f32`s ≈ 1 s @ 48 kHz stereo).
 - **Hot-path rules** — the cpal callback never allocates, locks or logs. It only reads the `rtrb::Consumer` and `Atomic*` fields in `SharedPlayback`.
 

src-tauri/Cargo.toml

@@ -94,7 +94,7 @@ blake3 = "1"
 # Audio playback: symphonia decodes, cpal outputs, rubato resamples,
 # rtrb is the SPSC ring between decoder thread and audio callback,
 # crossbeam-channel carries AudioCmd from tokio commands to the decoder.
-symphonia = { version = "0.5", default-features = false, features = [
+symphonia = { version = "0.6", default-features = false, features = [
     "mp3",
     "flac",
     "wav",
@@ -104,6 +104,10 @@ symphonia = { version = "0.5", default-features = false, features = [
     "alac",
     "isomp4",
     "pcm",
+    # Symphonia 0.6 split metadata behind feature flags; `all-meta` keeps
+    # ID3/APE/Vorbis tag detection during probe scoring (the scanner uses
+    # lofty for the actual tag read, so this is purely for probe accuracy).
+    "all-meta",
 ] }
 cpal = "0.17"
 rubato = "2.0"