Merge pull request #1200 from schreibfaul1/MP3_MAIN_DATA_UNDERFLOW

ESPUIno non-progressive m4a files

Author: schreibfaul1

src/Audio.cpp

@@ -234,12 +234,13 @@ class Audio {
     bool         readMetadata(uint16_t b, uint16_t* readedBytes, bool first = false);
     int32_t      getChunkSize(uint16_t* readedBytes, bool first = false);
     bool         readID3V1Tag();
-    int32_t      newInBuffStart(int32_t m_resumeFilePos);
+    int32_t      newInBuffStart(int32_t resumeFilePos);
     boolean      streamDetection(uint32_t bytesAvail);
     uint32_t     m4a_correctResumeFilePos();
     uint32_t     ogg_correctResumeFilePos();
     int32_t      flac_correctResumeFilePos();
     int32_t      mp3_correctResumeFilePos();
+    int32_t      wav_correctResumeFilePos();
     uint8_t      determineOggCodec();
     void         strlower(char* str);
     void         trim(char* str);
@@ -322,6 +323,7 @@ class Audio {
 
     SemaphoreHandle_t mutex_playAudioData;
     SemaphoreHandle_t mutex_audioTask;
+    SemaphoreHandle_t mutex_audioTaskIsDecoding;
     TaskHandle_t      m_audioTaskHandle = nullptr;
 
 #pragma GCC diagnostic push
@@ -362,6 +364,7 @@ class Audio {
     uint32_t       m_avr_bitrate = 0;       // average bitrate, median calculated by VBR
     uint32_t       m_nominal_bitrate = 0;   // given br from header
     uint32_t       m_audioFilePosition = 0; // current position, counts every readed byte
+    uint32_t       m_audioDataReadPtr = 0;  // used in playAudioData
     uint32_t       m_audioFileSize = 0;     // local and web files
     int            m_readbytes = 0;         // bytes read
     uint32_t       m_metacount = 0;         // counts down bytes between metadata

src/Audio.h

@@ -98,6 +98,7 @@ struct m4aHdr_t { // used in read_M4A_Header
     size_t   audioDataPos;
     size_t   cnt;
     size_t   offset;
+    uint32_t mdat_startPos;
     uint32_t picPos;
     uint32_t picLen;
     uint32_t ilst_pos;
@@ -116,6 +117,7 @@ struct m4aHdr_t { // used in read_M4A_Header
     uint32_t stsz_table_pos;
     bool     progressive; // Progressive (moov before mdat)
     bool     version_flags;
+    bool     mdat_seen;
 };
 
 struct plCh_t { // used in playChunk

src/audiolib_structs.hpp

@@ -552,17 +552,17 @@ int32_t MP3Decoder::findSyncWord(uint8_t* buf, int32_t nBytes) {
     // };
 
     typedef struct {
-        uint8_t  mpeg_version; // 0=MPEG2.5, 1=reserved, 2=MPEG2, 3=MPEG1
-        uint8_t  layer;        // 0=reserved, 1=Layer III, 2=Layer II, 3=Layer I
-        bool     crc_protected;
-        uint8_t  bitrate_idx;
-        uint8_t  sample_rate_idx;
-        bool     padding;
-        uint8_t  channel_mode;
-        uint32_t frame_length = 0; // cytes
-        uint16_t sample_rate_hz;   // the actual sampling rate in Hz
-        uint16_t bitrate_kbps;     // the actual bit rate in Kbps
-        uint16_t samples_per_frame;
+        uint8_t  mpeg_version = 0; // 0=MPEG2.5, 1=reserved, 2=MPEG2, 3=MPEG1
+        uint8_t  layer = 0;        // 0=reserved, 1=Layer III, 2=Layer II, 3=Layer I
+        bool     crc_protected = 0;
+        uint8_t  bitrate_idx = 0;
+        uint8_t  sample_rate_idx = 0;
+        bool     padding = 0;
+        uint8_t  channel_mode = 0;
+        uint32_t frame_length = 0;   // cytes
+        uint16_t sample_rate_hz = 0; // the actual sampling rate in Hz
+        uint16_t bitrate_kbps = 0;   // the actual bit rate in Kbps
+        uint16_t samples_per_frame = 0;
     } Mp3FrameHeader_sync_t;
 
     // SamplingFrequenz-Lookup tables(Beispiel für MPEG1, MPEG2, MPEG2.5)
@@ -644,7 +644,7 @@ int32_t MP3Decoder::findSyncWord(uint8_t* buf, int32_t nBytes) {
         uint16_t sample_rate_hz = 0;
 
         // Mapping from MPEG version to sampling rate table
-        uint8_t sr_table_idx;
+        uint8_t sr_table_idx = 0;
         if (header_info->mpeg_version == 3)
             sr_table_idx = 0; // MPEG 1 (0b11)
         else if (header_info->mpeg_version == 2)
@@ -1093,11 +1093,31 @@ int32_t MP3Decoder::decode(uint8_t* inbuf, int32_t* bytesLeft, int16_t* outbuf)
     // Skip fake frames
     int frameLen = IsLikelyRealFrame(inbuf, *bytesLeft);
 
+    if(m_invalid_frame.start == true && m_invalid_frame.timer + 3000 > millis()) m_invalid_frame.start = false;
+
     if (frameLen <= 0) {
         int skip = -frameLen;
         if (skip > 0 && skip <= *bytesLeft) {
             *bytesLeft -= skip;
             MP3_LOG_DEBUG("Fakeframe, size %i", frameLen);
+
+            if (m_invalid_frame.start == false) { // fake frames control
+                m_invalid_frame.start = true;
+                m_invalid_frame.timer = millis();
+                m_invalid_frame.count1 = 0;
+                m_invalid_frame.count2 = 0;
+            } else {
+                m_invalid_frame.count1++;
+                if (m_invalid_frame.start && m_invalid_frame.timer + 1000 < millis()) { m_invalid_frame.count2++; }
+                if (m_invalid_frame.start && m_invalid_frame.timer + 2000 < millis()) {
+                    if (m_invalid_frame.count1 > 5 && m_invalid_frame.count2 > 5) {
+                        // network error
+                        m_invalid_frame.start = false;
+                        return MP3_NEED_RESTART;
+                    }
+                }
+            }
+
             return MP3_NONE; // fakeframe
         }
         // inbuf empty or unusable
@@ -1146,7 +1166,7 @@ int32_t MP3Decoder::decode(uint8_t* inbuf, int32_t* bytesLeft, int16_t* outbuf)
     if (m_MP3DecInfo->nSlots > *bytesLeft) {
         MP3ClearBadFrame(outbuf);
         MP3_LOG_DEBUG("MP3, indata underflow");
-        return MP3_NONE;
+        return MP3_MAIN_DATA_UNDERFLOW;
     }
 
     /* fill main data buffer with enough new data for this frame */

src/mp3_decoder/mp3_decoder.cpp

@@ -35,6 +35,8 @@ class MP3Decoder : public Decoder {
     enum {
         MP3_NONE = 0,
         MP3_ERR = -1,
+        MP3_MAIN_DATA_UNDERFLOW = - 2,
+        MP3_NEED_RESTART = -3,
         MP3_STOP = -100,
     };
 
@@ -59,6 +61,8 @@ class MP3Decoder : public Decoder {
     ps_ptr<MP3FrameInfo_t>  m_MP3FrameInfo;
     ps_ptr<char>            m_mpeg_version_str;
 
+    invalid_frame           m_invalid_frame;
+
     // internally used
     int32_t  IsLikelyRealFrame(const uint8_t* p, int32_t bytesLeft);
     void     MP3GetLastFrameInfo();

src/mp3_decoder/mp3_decoder.h

@@ -222,4 +222,12 @@ typedef struct {
     bool     padding;
     uint8_t  channel_mode;
     uint32_t frame_length; // In Bytes
-} Mp3FrameHeader;
+} Mp3FrameHeader;
+
+struct invalid_frame {
+    uint32_t timer = 0;
+    bool start = true;
+    uint32_t count1 = 0;
+    uint32_t count2 = 0;
+
+};