fix(azure_speech): 修复语音合成会话管理逻辑

修复 Azure TTS 助手中的会话管理问题，确保音频合成和播放仅在正确的会话上下文中执行。移除冗余的会话 ID 处理逻辑，并添加会话快照检查以防止跨会话音频数据干扰。

具体修改包括：
- 在合成开始和音频处理事件中添加会话 ID 一致性检查
- 移除 `effectiveSessionId` 计算逻辑，直接使用传入的 sessionid
- 在停止合成时清空会话快照
- 注释调试日志以减少控制台输出

local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureSpeechPlugin.swift

@@ -1116,7 +1116,7 @@ extension AzureSpeechPlugin: BleService.Callback {
             }
         }
         
-        print("分离音频数据 - 左声道: \(leftBuffer.count) bytes, 右声道: \(rightBuffer.count) bytes")
+            // print("分离音频数据 - 左声道: \(leftBuffer.count) bytes, 右声道: \(rightBuffer.count) bytes")
         
               // 安全解包版本
             guard let audioStream = azureAsrHelper.audioStream else {

local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureTtsHelper.swift

@@ -18,6 +18,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
     private var isInitialized = false
     private var speaking = false
     private var sessionid = ""  //语音播报任务id
+    private var synthesisSessionIdSnapshot = ""
 
     // 当前配置
     private var currentVoice = DEFAULT_VOICE
@@ -192,10 +193,10 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             return false
         }
 
-        let effectiveSessionId = sessionid.isEmpty ? self.sessionid : sessionid
-        if !sessionid.isEmpty {
-            self.sessionid = sessionid
-        }
+        // let effectiveSessionId = sessionid.isEmpty ? self.sessionid : sessionid
+        // if !sessionid.isEmpty {
+        self.sessionid = sessionid
+        // }
         // os_log("调用链: speakOnce 接收 session=%{public}@ effective=%{public}@ speaking=%{public}@ pendingTasks=%{public}d", log: log, type: .info, sessionid, effectiveSessionId, speaking.description, pendingTasks.count)
 
         // 清理文本
@@ -208,7 +209,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
         // 异步处理
         // os_log("调用链: speakOnce 入队合成 session=%{public}@ 文本长度=%{public}d", log: log, type: .info, effectiveSessionId, cleanedText.count)
         enqueueSynthesisTask {
-            self.performSynthesis(sessionid: effectiveSessionId, text: cleanedText)
+            self.performSynthesis(sessionid: sessionid, text: cleanedText)
         }
 
         return true
@@ -225,9 +226,10 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             ])
             return false
         }
-        let effectiveSessionId = sessionid.isEmpty ? self.sessionid : sessionid
-        if !sessionid.isEmpty {
-            self.sessionid = sessionid
+        // let effectiveSessionId = sessionid.isEmpty ? self.sessionid : sessionid
+        // if !sessionid.isEmpty {
+        if self.sessionid != sessionid {
+            return false
         }
         // os_log("调用链: speakStream 接收 session=%{public}@ effective=%{public}@ 输入长度=%{public}d bufferLen=%{public}d", log: log, type: .info, sessionid, effectiveSessionId, text.count, streamBuffer.count)
 
@@ -279,7 +281,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
 
             if !textToSpeak.isEmpty {
                 // os_log("调用链: speakStream 触发合成 session=%{public}@ speakLen=%{public}d remainLen=%{public}d", log: log, type: .info, effectiveSessionId, textToSpeak.count, streamBuffer.count)
-                return speakOnce(sessionid: effectiveSessionId, textToSpeak)
+                return speakOnce(sessionid: sessionid, textToSpeak)
             }
         }
 
@@ -289,7 +291,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             let startIndex = currentText.index(currentText.startIndex, offsetBy: maxCharsBeforeForcedSpeak)
             streamBuffer = String(currentText[startIndex...])
             // os_log("调用链: speakStream 强制分段合成 session=%{public}@ speakLen=%{public}d remainLen=%{public}d", log: log, type: .info, effectiveSessionId, textToSpeak.count, streamBuffer.count)
-            return speakOnce(sessionid: effectiveSessionId, textToSpeak)
+            return speakOnce(sessionid: sessionid, textToSpeak)
         }
 
         // os_log("调用链: speakStream 未触发合成 session=%{public}@ bufferLen=%{public}d", log: log, type: .debug, effectiveSessionId, streamBuffer.count)
@@ -307,6 +309,9 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             ])
             return false
         }
+        if self.sessionid != sessionid {
+            return false
+        }
         // os_log("调用链: flushStream session=%{public}@ bufferLen=%{public}d", log: log, type: .info, sessionid, streamBuffer.count)
 
         let remainingText = streamBuffer
@@ -326,8 +331,10 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
      * - Throws: 无（内部捕获 SDK/音频会话异常并转为 error 事件）
      */
     public func stop() -> Bool {
+        os_log("liwei------------- 停止语音合成 session=%{public}@", log: self.log, type: .info, sessionid)
         speaking = false
         sessionid = ""
+        synthesisSessionIdSnapshot = ""
         streamBuffer = ""
         lastSpokenText = nil
         
@@ -356,6 +363,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
         taskLock.unlock()
 
         do {
+           
             try synthesizer?.stopSpeaking()
             self.notifyEvent(eventType: .synthesisCanceled)
         } catch {
@@ -563,9 +571,10 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
         
         do {
             if (self.sessionid != sessionid) {
-                // os_log("调用链: performSynthesis 中止(会话变化) current=%{public}@ task=%{public}@", log: log, type: .info, self.sessionid, sessionid)
+                 os_log("liwei----------- 调用链: performSynthesis 中止(会话变化) current=%{public}@ task=%{public}@", log: log, type: .info, self.sessionid, sessionid)
                 return
             }
+            synthesisSessionIdSnapshot = sessionid
             // os_log("调用链: performSynthesis startSpeakingSsml session=%{public}@ ssmlLen=%{public}d", log: log, type: .info, sessionid, ssml.count)
             let result = try synthesizer?.startSpeakingSsml(ssml)
             
@@ -576,7 +585,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             speaking = false
             // 减少待处理文本计数
             pendingTextCount = max(0, pendingTextCount - text.count)
-            os_log("语音合成失败: %{public}@", log: log, type: .error, error.localizedDescription)
+            os_log("liwei-------------- 语音合成失败: %{public}@", log: log, type: .error, error.localizedDescription)
             notifyEvent(eventType: .error, params: [
                 "errorCode": "SYNTHESIS_FAILED",
                 "errorMessage": error.localizedDescription
@@ -653,6 +662,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             guard let self = self else { return }
             self.currentSynthesisBuffer = Data()
             if self.suppressStreamPlayback { return }
+            self.synthesisSessionIdSnapshot = self.sessionid
             if self.isUsingPushStreamCapture {
                 self.audioPlaybackQueue.async {
                     self.activeStreamSynthesisCount += 1
@@ -669,6 +679,9 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
         synthesizer?.addSynthesizingEventHandler { [weak self] _, event in
             if let audioData = event.result.audioData {
                 guard let self = self else { return }
+                if self.synthesisSessionIdSnapshot != self.sessionid {
+                    return
+                }
                 // os_log("调用链: 合成中 接收音频片段=%{public}dB session=%{public}@", log: self.log, type: .debug, audioData.count, self.sessionid)
                 if self.isUsingPushStreamCapture {
                     if self.pushStreamCallbackCount == 0, !audioData.isEmpty {
@@ -691,6 +704,18 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             if !self.isUsingPushStreamCapture {
                 self.speaking = false
             }
+            if self.synthesisSessionIdSnapshot != self.sessionid {
+                self.currentSynthesisBuffer = Data()
+                if self.isUsingPushStreamCapture {
+                    self.audioPlaybackQueue.async {
+                        self.activeStreamSynthesisCount = max(0, self.activeStreamSynthesisCount - 1)
+                        self.checkStreamPlaybackCompletedIfNeeded()
+                    }
+                }
+                self.pendingTextCount = max(0, self.pendingTextCount - 1)
+                self.notifyEvent(eventType: .synthesisCompleted)
+                return
+            }
             // 将当前合成的音频加入播放队列
             if !self.currentSynthesisBuffer.isEmpty {
                 // os_log("调用链: 合成完成 入队播放 数据长度=%{public}dB session=%{public}@", log: self.log, type: .info, self.currentSynthesisBuffer.count, self.sessionid)
@@ -771,6 +796,9 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             // os_log("调用链: 推送音频丢弃(无会话且未speaking) bytes=%{public}d", log: log, type: .debug, data.count)
             return
         }
+        if synthesisSessionIdSnapshot != sessionid {
+            return
+        }
         notifyAudioData(data)
         guard isUsingPushStreamCapture else {
             // os_log("调用链: 推送音频忽略(未启用pushStream) bytes=%{public}d", log: log, type: .debug, data.count)
@@ -790,6 +818,9 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
                 // os_log("调用链: 推送音频异步丢弃(无会话且未speaking) bytes=%{public}d", log: self.log, type: .debug, data.count)
                 return
             }
+            if self.synthesisSessionIdSnapshot != self.sessionid {
+                return
+            }
             let pcm = self.stripWavHeaderIfNeeded(data)
             if !pcm.isEmpty {
                 self.pcmPendingData.append(pcm)

pubspec.yaml

@@ -1,7 +1,7 @@
 name: voitrans
 description: "Voitrans - AI Voice Assistant."
 publish_to: "none"
-version: 1.0.21+71
+version: 1.0.22+71
 
 environment:
   sdk: ">=3.3.0 <4.0.0"