rewrite avsynth: pure ObjC shim with C FFI, no msg_send!/objc_exception

Author: willwade

Cargo.toml

@@ -19,7 +19,7 @@ sherpaonnx = ["sherpa-onnx", "serde_json"]
 tungstenite = ["dep:tungstenite"]
 uuid = ["dep:uuid"]
 url = ["dep:url"]
-avsynth = ["objc", "objc-foundation", "objc_exception"]
+avsynth = []
 sapi = ["windows"]
 
 [dependencies]
@@ -34,15 +34,13 @@ anyhow = "1"
 tungstenite = { version = "0.29.0", features = ["rustls-tls-webpki-roots"], optional = true }
 uuid = { version = "1.23.2", features = ["v4"], optional = true }
 url = { version = "2.5.8", optional = true }
-objc = { version = "0.2", optional = true }
-objc-foundation = { version = "0.1", optional = true }
-objc_exception = { version = "0.1", optional = true }
 
 [target.'cfg(target_os = "windows")'.dependencies]
 windows = { version = "0.61", features = ["Win32_Media_Speech", "Win32_System_Com", "Win32_System_Ole"], optional = true }
 
 [build-dependencies]
 cbindgen = "0.28"
+cc = "1"
 
 [lints.clippy]
 all = "warn"

build.rs

@@ -4,7 +4,7 @@ fn main() {
     let crate_dir = env::var("CARGO_MANIFEST_DIR").unwrap();
     let config = cbindgen::Config::from_file("cbindgen.toml").unwrap_or_default();
     match cbindgen::Builder::new()
-        .with_crate(crate_dir)
+        .with_crate(&crate_dir)
         .with_config(config)
         .generate()
     {
@@ -15,4 +15,15 @@ fn main() {
             eprintln!("cbindgen warning: {e}");
         }
     }
+
+    if env::var("TARGET").unwrap_or_default().contains("apple") {
+        cc::Build::new()
+            .file("extern/avsynth_shim.m")
+            .compiler("clang")
+            .flag("-fobjc-arc")
+            .compile("avsynth_shim");
+        println!("cargo:rustc-link-lib=framework=AVFAudio");
+        println!("cargo:rustc-link-lib=framework=Foundation");
+        println!("cargo:rustc-link-lib=objc");
+    }
 }

extern/avsynth_shim.m

@@ -0,0 +1,99 @@
+#import <Foundation/Foundation.h>
+#import <AVFAudio/AVSpeechSynthesis.h>
+
+void* avsynth_create(void) {
+    @try {
+        AVSpeechSynthesizer *synth = [[AVSpeechSynthesizer alloc] init];
+        return (void *)CFBridgingRetain(synth);
+    } @catch (NSException *e) {
+        return NULL;
+    }
+}
+
+void avsynth_destroy(void *handle) {
+    if (!handle) return;
+    @try {
+        AVSpeechSynthesizer *synth = (__bridge_transfer AVSpeechSynthesizer *)handle;
+        synth = nil;
+    } @catch (NSException *e) {}
+}
+
+void avsynth_speak(void *handle, const char *text, const char *voice_id,
+                   float rate, float pitch, float volume) {
+    if (!handle || !text) return;
+    @try {
+        AVSpeechSynthesizer *synth = (__bridge AVSpeechSynthesizer *)handle;
+        NSString *nsText = [NSString stringWithUTF8String:text];
+        if (!nsText) return;
+        AVSpeechUtterance *utterance = [[AVSpeechUtterance alloc] initWithString:nsText];
+        utterance.rate = rate < 0.1f ? 0.1f : (rate > 10.0f ? 10.0f : rate);
+        utterance.pitchMultiplier = pitch < 0.5f ? 0.5f : (pitch > 2.0f ? 2.0f : pitch);
+        utterance.volume = volume < 0.0f ? 0.0f : (volume > 1.0f ? 1.0f : volume);
+        if (voice_id && voice_id[0] != '\0') {
+            NSString *nsVoiceId = [NSString stringWithUTF8String:voice_id];
+            AVSpeechSynthesisVoice *voice = [AVSpeechSynthesisVoice voiceWithIdentifier:nsVoiceId];
+            if (voice) utterance.voice = voice;
+        }
+        [synth speakUtterance:utterance];
+    } @catch (NSException *e) {}
+}
+
+void avsynth_stop(void *handle) {
+    if (!handle) return;
+    @try {
+        AVSpeechSynthesizer *synth = (__bridge AVSpeechSynthesizer *)handle;
+        [synth stopSpeakingAtBoundary:AVSpeechBoundaryImmediate];
+    } @catch (NSException *e) {}
+}
+
+void avsynth_pause(void *handle) {
+    if (!handle) return;
+    @try {
+        AVSpeechSynthesizer *synth = (__bridge AVSpeechSynthesizer *)handle;
+        [synth pauseSpeakingAtBoundary:AVSpeechBoundaryImmediate];
+    } @catch (NSException *e) {}
+}
+
+void avsynth_resume(void *handle) {
+    if (!handle) return;
+    @try {
+        AVSpeechSynthesizer *synth = (__bridge AVSpeechSynthesizer *)handle;
+        [synth continueSpeaking];
+    } @catch (NSException *e) {}
+}
+
+int avsynth_voice_count(void *handle) {
+    if (!handle) return 0;
+    @try {
+        NSArray<AVSpeechSynthesisVoice *> *voices = [AVSpeechSynthesisVoice speechVoices];
+        return (int)[voices count];
+    } @catch (NSException *e) {
+        return 0;
+    }
+}
+
+int avsynth_get_voice(void *handle, int index,
+                      char *id_buf, int id_buf_len,
+                      char *name_buf, int name_buf_len,
+                      char *lang_buf, int lang_buf_len) {
+    if (!handle) return -1;
+    @try {
+        NSArray<AVSpeechSynthesisVoice *> *voices = [AVSpeechSynthesisVoice speechVoices];
+        if (index < 0 || index >= (int)[voices count]) return -1;
+        AVSpeechSynthesisVoice *voice = voices[index];
+
+        NSString *identifier = voice.identifier ?: @"";
+        NSString *name = voice.name ?: @"";
+        NSString *language = voice.language ?: @"";
+
+        strncpy(id_buf, [identifier UTF8String], id_buf_len - 1);
+        id_buf[id_buf_len - 1] = '\0';
+        strncpy(name_buf, [name UTF8String], name_buf_len - 1);
+        name_buf[name_buf_len - 1] = '\0';
+        strncpy(lang_buf, [language UTF8String], lang_buf_len - 1);
+        lang_buf[lang_buf_len - 1] = '\0';
+        return 0;
+    } @catch (NSException *e) {
+        return -1;
+    }
+}