Fix HLS transcoding audio with non-MP3 input

The buffer splitting I had in place was slightly off, with the result of
horrible sounding audio when transcoding from AC3.  Feeding entire single
frames of audio to the encoder is the trick, and finally, we are there.

Fixes #10488

Author: Beirdo

mythtv/programs/mythtranscode/transcode.cpp

@@ -133,39 +133,61 @@ class AudioReencodeBuffer : public AudioOutput
 
         QMutexLocker locker(&m_bufferMutex);
         AudioBuffer *ab;
+        unsigned char *buf = (unsigned char *)buffer;
+        int savedlen = 0;
+        int firstlen = 0;
 
         if (m_audioFrameSize)
         {
-            int currLen = m_saveBuffer.m_buffer.size();
-            if (currLen)
+            savedlen = m_saveBuffer.m_buffer.size();
+            if (savedlen)
             {
                 ab = new AudioBuffer(m_saveBuffer);
                 m_saveBuffer.clear();
+                firstlen = m_audioFrameSize - savedlen;
             }
             else
                 ab = new AudioBuffer;
 
             int overage;
 
-            if ((overage = (currLen + len) % m_audioFrameSize))
+            if ((overage = (len + savedlen) % m_audioFrameSize))
             {
                 long long newtime = timecode + ((len / m_bytes_per_frame) /
                                                 (m_eff_audiorate / 1000));
                 len -= overage;
-                m_saveBuffer.setData(&((unsigned char *)buffer)[len], overage,
-                                     newtime);
+                m_saveBuffer.setData(&buf[len], overage, newtime);
             }
         }
         else
         {
             ab = new AudioBuffer;
         }
 
-        m_last_audiotime = timecode + ((len / m_bytes_per_frame) /
-                                       (m_eff_audiorate / 1000));
-        ab->appendData((unsigned char *)buffer, len, m_last_audiotime);
+        int bufflen = (m_audioFrameSize ? m_audioFrameSize : len);
+        int index = 0;
 
-        m_bufferList.append(ab);
+        do
+        {
+            if (!len)
+                break;
+
+            int blocklen = (firstlen ? firstlen : bufflen);
+
+            m_last_audiotime = timecode + ((blocklen / m_bytes_per_frame) /
+                                           (m_eff_audiorate / 1000));
+            ab->appendData(&buf[index], blocklen, m_last_audiotime);
+            m_bufferList.append(ab);
+
+            timecode = m_last_audiotime;
+            index += blocklen;
+            firstlen = 0;
+
+            ab = new AudioBuffer;
+        } while (index < len);
+
+        // The last buffer is actually unused
+        delete ab;
 
         return true;
     }