Support SF3 soundfonts, update TinySoundFont, add tests and CI

.github/workflows/test.yaml

@@ -0,0 +1,45 @@
+name: Test
+on:
+  pull_request:
+  push:
+    branches: [master]
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  linux:
+    name: Linux tests
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Java
+        uses: actions/setup-java@v4
+        with:
+          distribution: 'zulu'
+          java-version: '17'
+
+      - name: Setup Gradle
+        uses: gradle/actions/setup-gradle@v4
+
+      - name: Cache Kotlin/Native toolchain
+        uses: actions/cache@v4
+        with:
+          path: ~/.konan
+          key: konan-${{ runner.os }}-${{ hashFiles('gradle/libs.versions.toml') }}
+
+      # The shared tests (src/commonTest) built for linuxX64 and run natively: same TinySoundFont
+      # and stb_vorbis code as every other target, without an Android emulator.
+      - name: Run tests
+        run: ./gradlew :soundfont:linuxX64Test
+
+      - name: Upload test reports
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: test-reports
+          path: soundfont/build/reports/tests/

README.md

@@ -4,7 +4,7 @@
 [![GitHub License](https://img.shields.io/badge/license-Apache%20License%202.0-blue.svg?style=flat)](http://www.apache.org/licenses/LICENSE-2.0)
 
 Library for [sample-based synthesis](https://en.wikipedia.org/wiki/Sample-based_synthesis "Sample-based synthesis") to generate audio data. 
-The library allows to load `.sf2` files from memory. Is basically a KMP wrapper around [TinySoundFont](https://github.com/schellingb/TinySoundFont) library.
+The library allows to load `.sf2` and `.sf3` files from memory. Is basically a KMP wrapper around [TinySoundFont](https://github.com/schellingb/TinySoundFont) library.
 
 The goal of this library is to allow operations on .sf2 files and generating audio in Kotlin Multiplatform projects. 
 The library has `batteries-included` so no configuration is required.
@@ -35,7 +35,7 @@ mikrosoundfont = { module = "io.github.lemcoder.mikrosoundfont:soundfont", versi
 ___
 ### Usage
 ```
-val sfBuffer   // .sf2 file bytes
+val sfBuffer   // .sf2 or .sf3 file bytes
 
 MikroSoundFont.load(sfBuffer).use { soundFont ->
     // .. use soundFont to generate audio data
@@ -44,6 +44,19 @@ MikroSoundFont.load(sfBuffer).use { soundFont ->
 
 A `SoundFont` holds its samples in native memory until `close()` is called, so close it (or load it in `use { }`) when you are done with it.
 
+#### SF3 files
+`.sf3` files store their samples compressed with Ogg Vorbis. `load` decodes all of them up front, so:
+* loading takes noticeably longer than an `.sf2` of the same size (seconds for a large font on a phone), so load off the main thread;
+* the loaded `SoundFont` takes several times the file size in memory, as the samples are kept decoded.
+
+Decoding uses [stb_vorbis](https://github.com/nothings/stb), which is not hardened against malformed data, so only load `.sf3` files you trust.
+
+## Testing
+
+The tests in `soundfont/src/commonTest` load an `.sf2` and an `.sf3` font and render audio through the native library:
+* `./gradlew :soundfont:linuxX64Test` or `:soundfont:macosArm64Test` runs them on the host (CI runs the Linux ones on every pull request);
+* `./gradlew :soundfont:connectedAndroidDeviceTest` runs them on a connected Android device or emulator.
+
 ## License
 
 This project is licensed under the Apache 2.0 License - see the [LICENSE.md](LICENSE.md) file for details

gradle/libs.versions.toml

@@ -10,6 +10,7 @@ maven-publish = "0.37.0"
 konanplugin = "2.0.0-alpha05"
 
 [libraries]
+androidx-test-runner = { module = "androidx.test:runner", version = "1.6.2" }
 kotlin-test = { module = "org.jetbrains.kotlin:kotlin-test", version.ref = "kotlin" }
 
 [plugins]

native/include/tsf/tsfCInterop.h

@@ -21,7 +21,7 @@
 
    LICENSE (MIT)
 
-   Copyright (C) 2017-2023 Bernhard Schelling
+   Copyright (C) 2017-2025 Bernhard Schelling
    Based on SFZero, Copyright (C) 2012 Steve Folta (https://github.com/stevefolta/SFZero)
 
    Permission is hereby granted, free of charge, to any person obtaining a copy of this
@@ -185,6 +185,7 @@ TSFDEF void tsf_render_float(tsf* f, float* buffer, int samples, int flag_mixing
 //   pitch_wheel: pitch wheel position 0 to 16383 (default 8192 unpitched)
 //   pitch_range: range of the pitch wheel in semitones (default 2.0, total +/- 2 semitones)
 //   tuning: tuning of all playing voices in semitones (default 0.0, standard (A440) tuning)
+//   flag_sustain: 0 to end notes that were held sustained and disable holding sustain otherwise enable it
 //   (tsf_set_preset_number and set_bank_preset return 0 if preset does not exist, otherwise 1)
 //   (tsf_channel_set_... return 0 if a new channel needed allocation and that failed, otherwise 1)
 TSFDEF int tsf_channel_set_presetindex(tsf* f, int channel, int preset_index);
@@ -196,6 +197,7 @@ TSFDEF int tsf_channel_set_volume(tsf* f, int channel, float volume);
 TSFDEF int tsf_channel_set_pitchwheel(tsf* f, int channel, int pitch_wheel);
 TSFDEF int tsf_channel_set_pitchrange(tsf* f, int channel, float pitch_range);
 TSFDEF int tsf_channel_set_tuning(tsf* f, int channel, float tuning);
+TSFDEF int tsf_channel_set_sustain(tsf* f, int channel, int flag_sustain);
 
 // Start or stop playing notes on a channel (needs channel preset to be set)
 //   channel: channel number