allow sh_audio struct to be initialized by demuxer, add parsing of mp4 esds header to mov demuxer, init faad from info from mov header

git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@5207 b3059339-0415-0410-9bf9-f77b7e298cf2

dec_audio.c

@@ -328,15 +328,17 @@ int mplayer_audio_read(char *buf,int size){
 int init_audio(sh_audio_t *sh_audio){
 int driver=sh_audio->codec->driver;
 
-sh_audio->samplesize=2;
+if(!sh_audio->samplesize)
+  sh_audio->samplesize=2;
+if(!sh_audio->sample_format)
 #ifdef WORDS_BIGENDIAN
-sh_audio->sample_format=AFMT_S16_BE;
+  sh_audio->sample_format=AFMT_S16_BE;
 #else
-sh_audio->sample_format=AFMT_S16_LE;
+  sh_audio->sample_format=AFMT_S16_LE;
 #endif
-sh_audio->samplerate=0;
+//sh_audio->samplerate=0;
 //sh_audio->pcm_bswap=0;
-sh_audio->o_bps=0;
+//sh_audio->o_bps=0;
 
 sh_audio->a_buffer_size=0;
 sh_audio->a_buffer=NULL;
@@ -840,15 +842,14 @@ case AFM_AAC: {
   faacDecConfigurationPtr faac_conf;
   faac_hdec = faacDecOpen();
 
-#if 0
+#if 1
   /* Set the default object type and samplerate */
   /* This is useful for RAW AAC files */
   faac_conf = faacDecGetCurrentConfiguration(faac_hdec);
   if(sh_audio->samplerate)
     faac_conf->defSampleRate = sh_audio->samplerate;
-  /* XXX: is outputFormat samplesize of compressed data or samplesize of
+   /* XXX: FAAD support FLOAT output, how do we handle
-   * decoded data, maybe upsampled? Also, FAAD support FLOAT output,
+    * that (FAAD_FMT_FLOAT)? ::atmos
-   * how do we handle that (FAAD_FMT_FLOAT)? ::atmos
     */
   if(sh_audio->samplesize)
     switch(sh_audio->samplesize){
@@ -865,7 +866,7 @@ case AFM_AAC: {
 	faac_conf->outputFormat = FAAD_FMT_32BIT;
 	break;
     }
-  faac_conf->defObjectType = LTP; // => MAIN, LC, SSR, LTP available.
+  //faac_conf->defObjectType = LTP; // => MAIN, LC, SSR, LTP available.
 
   faacDecSetConfiguration(faac_hdec, faac_conf);
 #endif
@@ -887,7 +888,11 @@ case AFM_AAC: {
     mp_msg(MSGT_DECAUDIO,MSGL_V,"FAAD: Negotiated samplerate: %dHz  channels: %d\n", faac_samplerate, faac_channels);
     sh_audio->channels = faac_channels;
     sh_audio->samplerate = faac_samplerate;
+    if(!sh_audio->i_bps) {
+      mp_msg(MSGT_DECAUDIO,MSGL_WARN,"FAAD: compressed input bitrate missing, assuming 128kbit/s!\n");
       sh_audio->i_bps = 128*1000/8; // XXX: HACK!!! There's currently no way to get bitrate from libfaad2! ::atmos
+    } else 
+      mp_msg(MSGT_DECAUDIO,MSGL_V,"FAAD: got %dkbit/s rate from MP4 header!\n",sh_audio->i_bps*8/1000);
   }  
 	    
 } break;		

libmpdemux/demux_mov.c

@@ -562,9 +562,37 @@ static void lschunks(demuxer_t* demuxer,int level,off_t endpos,mov_track_t* trak
 	    mov_build_index(trak);
 	    switch(trak->type){
 	    case MOV_TRAK_AUDIO: {
+#if 0				   
+		struct {
+		   int16_t version;		// 0 or 1 (version 1 is qt3.0+)
+		   int16_t revision;		// 0
+		   int32_t vendor_id;		// 0
+		   int16_t channels;		// 1 or 2  (Mono/Stereo)
+		   int16_t samplesize;		// 8 or 16 (8Bit/16Bit)	
+		   int16_t compression_id;	// if version 0 then 0
+		  				// if version 1 and vbr then -2 else 0
+		   int16_t packet_size;		// 0
+		   uint32_t sample_rate;	// samplerate (Hz)
+		   // qt3.0+ (version == 1)
+		   uint32_t samples_per_packet;	// 0 or num uncompressed samples in a packet
+		   				// if 0 below three values are also 0
+		   uint32_t bytes_per_packet;	// 0 or num compressed bytes for one channel
+		   uint32_t bytes_per_frame;	// 0 or num compressed bytes for all channels
+		   				// (channels * bytes_per_packet)
+		   uint32_t bytes_per_sample;	// 0 or size of uncompressed sample
+		   // if samples_per_packet and bytes_per_packet are constant (CBR)
+		   // then bytes_per_frame and bytes_per_sample must be 0 (else is VBR)
+		   // ---
+		   // optional additional atom-based fields
+		   // ([int32_t size,int32_t type,some data ],repeat)
+		} my_stdata;		  
+#endif		
 		sh_audio_t* sh=new_sh_audio(demuxer,priv->track_db);
 		sh->format=trak->fourcc;
 
+// assumptions for below table: short is 16bit, int is 32bit		
+// short values are usually one byte leftpadded by zero		
+//   int values are usually two byte leftpadded by zero		
 //  stdata[]:
 //	8   short	version
 //	10  short	revision
@@ -579,8 +607,48 @@ static void lschunks(demuxer_t* demuxer,int level,off_t endpos,mov_track_t* trak
 //	32  int		bytes_per_packet
 //	36  int		bytes_per_frame
 //	40  int		bytes_per_sample
+// my recoveries from .mp4 files with correct index (counting from 0):
+// it's always start with padding/real start		
+//    8/ 9  short	version	
+//   10/11  short	revision	
+//   12/14  int		vendor_id	
+//   16/17  short	channels
+//   18/19  short	samplesize
+//   20/21  short	compression_id		
+//   22/23  short	packet_size (XXX: this overlaps with sample_rate ?)		
+//   22/24  int		sample_rate		
+// esds atom:
+//   28/31  int		atom size (bytes of int size, int type and data)
+//   32/32  int		atom type (fourc charater code -> esds)		
+//   62/63  int  	compressed datarate (Bits)
 
-		mp_msg(MSGT_DEMUX, MSGL_INFO, "Audio bits: %d  chans: %d\n",trak->stdata[19],trak->stdata[17]);
+#define char2short(x,y) ((x[y]<<8)|x[y+1])
+#define char2int(x,y) ((x[y]<<24)|(x[y+1]<<16)|(x[y+2]<<8)|x[y+3])
+
+		sh->samplesize=char2short(trak->stdata,18)/8;
+		sh->channels=char2short(trak->stdata,16);
+		sh->samplerate=char2short(trak->stdata,24);
+
+		mp_msg(MSGT_DEMUX, MSGL_INFO, "Audio bits: %d  chans: %d\n",
+		    trak->stdata[19],trak->stdata[17]);
+		mp_msg(MSGT_DEMUX, MSGL_INFO, "Audio sample rate: %d\n",
+		    char2short(trak->stdata,24));
+		if((trak->stdata[9]==0) && trak->stdata_len >= 36) { // version 0 with extra atoms
+		    int atom_len = char2int(trak->stdata,28);
+		    switch(char2int(trak->stdata,32)) { // atom type
+		      case MOV_FOURCC('e','s','d','s'):
+			mp_msg(MSGT_DEMUX, MSGL_INFO, "MOV: MP4 esds audio atom found (%d)!\n", atom_len);
+			if(atom_len >= 28)
+			  mp_msg(MSGT_DEMUX, MSGL_INFO, "Audio compressed datarate: %dkbit/s\n",
+			      char2int(trak->stdata,62)/1000);
+			  sh->i_bps=char2int(trak->stdata,62)/8;
+			break;
+		      default:
+			mp_msg(MSGT_DEMUX, MSGL_INFO, "MOV: unknown audio atom %c%c%c%c found (%d)!\n",
+			    trak->stdata[32],trak->stdata[33],trak->stdata[34],trak->stdata[35],
+			    atom_len);
+		    }
+		}  
 		mp_msg(MSGT_DEMUX, MSGL_INFO, "Fourcc: %.4s\n",&trak->fourcc);
 #if 0
 		{ FILE* f=fopen("stdata.dat","wb");