path: root/libmpdemux/demux_ogg.c

#include "config.h"

#ifdef HAVE_OGGVORBIS

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <math.h>

#include "../mp_msg.h"
#include "../help_mp.h"
#include "stream.h"
#include "demuxer.h"
#include "stheader.h"

#ifdef TREMOR
#include <tremor/ogg.h>
#include <tremor/ivorbiscodec.h>
#else
#include <ogg/ogg.h>
#include <vorbis/codec.h>
#endif

#ifdef HAVE_OGGTHEORA
#include <theora/theora.h>
#endif

#define BLOCK_SIZE 4096

/// Vorbis decoder context : we need the vorbis_info for vorbis timestamping
/// Shall we put this struct def in a common header ?
typedef struct ov_struct_st {
  vorbis_info      vi; /* struct that stores all the static vorbis bitstream
			  settings */
  vorbis_comment   vc; /* struct that stores all the bitstream user comments */
  vorbis_dsp_state vd; /* central working state for the packet->PCM decoder */
  vorbis_block     vb; /* local working space for packet->PCM decode */
  float            rg_scale; /* replaygain scale */
#ifdef TREMOR
  int              rg_scale_int;
#endif
} ov_struct_t;

/* Theora decoder context : we won't be able to interpret granule positions
 * without using theora_granule_time with the theora_state of the stream.
 * This is duplicated in `vd_theora.c'; put this in a common header?
 */
#ifdef HAVE_OGGTHEORA
typedef struct theora_struct_st {
    theora_state st;
    theora_comment cc;
    theora_info inf;
} theora_struct_t;
#endif

//// OggDS headers
// Header for the new header format
typedef struct stream_header_video
{
	ogg_int32_t	width;
	ogg_int32_t	height;
} stream_header_video;
	
typedef struct stream_header_audio
{
	ogg_int16_t	channels;
	ogg_int16_t	blockalign;
	ogg_int32_t	avgbytespersec;
} stream_header_audio;

typedef struct stream_header
{
	char	streamtype[8];
	char	subtype[4];

	ogg_int32_t	size;				// size of the structure

	ogg_int64_t	time_unit;			// in reference time
	ogg_int64_t	samples_per_unit;
	ogg_int32_t default_len;		// in media time

	ogg_int32_t buffersize;
	ogg_int16_t	bits_per_sample;

	union
	{
		// Video specific
		stream_header_video	video;
		// Audio specific
		stream_header_audio	audio;
	} sh;
} stream_header;

/// Our private datas

typedef struct ogg_syncpoint {
  int64_t granulepos;
  off_t page_pos;
} ogg_syncpoint_t;

/// A logical stream
typedef struct ogg_stream {
  /// Timestamping stuff
  float samplerate; /// granulpos 2 time
  int64_t lastpos;
  int32_t lastsize;

  // Logical stream state
  ogg_stream_state stream;
  int hdr_packets;
  int vorbis;
  int theora;
} ogg_stream_t;

typedef struct ogg_demuxer {
  /// Physical stream state
  ogg_sync_state sync;
  /// Current page
  ogg_page page;
  /// Logical streams
  ogg_stream_t *subs;
  int num_sub;
  ogg_syncpoint_t* syncpoints;
  int num_syncpoint;
  off_t pos, last_size;
} ogg_demuxer_t;

#define NUM_VORBIS_HDR_PACKETS 3

/// Some defines from OggDS
#define PACKET_TYPE_HEADER   0x01
#define PACKET_TYPE_BITS         0x07
#define PACKET_LEN_BITS01       0xc0
#define PACKET_LEN_BITS2         0x02
#define PACKET_IS_SYNCPOINT  0x08

extern int index_mode;

//-------- subtitle support - should be moved to decoder layer, and queue
//                          - subtitles up in demuxer buffer...

#include "../subreader.h"
#include "../libvo/sub.h"
#define OGG_SUB_MAX_LINE 128

static subtitle ogg_sub;
extern subtitle* vo_sub;
static float clear_sub;
//FILE* subout;

static
uint16_t get_uint16 (const void *buf)
{
  uint16_t      ret;
  unsigned char *tmp;

  tmp = (unsigned char *) buf;

  ret = tmp[1] & 0xff;
  ret = (ret << 8) + (tmp[0] & 0xff);

  return (ret);
}

static
uint32_t get_uint32 (const void *buf)
{
  uint32_t      ret;
  unsigned char *tmp;

  tmp = (unsigned char *) buf;

  ret = tmp[3] & 0xff;
  ret = (ret << 8) + (tmp[2] & 0xff);
  ret = (ret << 8) + (tmp[1] & 0xff);
  ret = (ret << 8) + (tmp[0] & 0xff);

  return (ret);
}

static
uint64_t get_uint64 (const void *buf)
{
  uint64_t      ret;
  unsigned char *tmp;

  tmp = (unsigned char *) buf;

  ret = tmp[7] & 0xff;
  ret = (ret << 8) + (tmp[6] & 0xff);
  ret = (ret << 8) + (tmp[5] & 0xff);
  ret = (ret << 8) + (tmp[4] & 0xff);
  ret = (ret << 8) + (tmp[3] & 0xff);
  ret = (ret << 8) + (tmp[2] & 0xff);
  ret = (ret << 8) + (tmp[1] & 0xff);
  ret = (ret << 8) + (tmp[0] & 0xff);

  return (ret);
}

void demux_ogg_init_sub () {
  int lcv;
  if(!ogg_sub.text[0]) // not yet allocated
  for (lcv = 0; lcv < SUB_MAX_TEXT; lcv++) {
    ogg_sub.text[lcv] = (char*)malloc(OGG_SUB_MAX_LINE);
  }
}

void demux_ogg_add_sub (ogg_stream_t* os,ogg_packet* pack) {
  int lcv;
  int line_pos = 0;
  int ignoring = 0;
  char *packet = pack->packet;

  mp_msg(MSGT_DEMUX,MSGL_DBG2,"\ndemux_ogg_add_sub %02X %02X %02X '%s'\n",
      (unsigned char)packet[0],
      (unsigned char)packet[1],
      (unsigned char)packet[2],
      &packet[3]);

  ogg_sub.lines = 0;
  if (((unsigned char)packet[0]) == 0x88) { // some subtitle text
    // Find data start
    int32_t duration = 0;
    int16_t hdrlen = (*packet & PACKET_LEN_BITS01)>>6, i;
    hdrlen |= (*packet & PACKET_LEN_BITS2) <<1;
    lcv = 1 + hdrlen;
    for (i = hdrlen; i > 0; i--) {
      duration <<= 8;
      duration |= (unsigned char)packet[i];
    }
    if ((hdrlen > 0) && (duration > 0)) {
      float pts;
      if(pack->granulepos == -1)
        pack->granulepos = os->lastpos + os->lastsize;
      pts = (float)pack->granulepos/(float)os->samplerate;
      clear_sub = pts + (float)duration/1000.0;
    }
    while (1) {
      int c = packet[lcv++];
      if(c=='\n' || c==0 || line_pos >= OGG_SUB_MAX_LINE-1){
	  ogg_sub.text[ogg_sub.lines][line_pos] = 0; // close sub
          if(line_pos) ogg_sub.lines++;
	  if(!c || ogg_sub.lines>=SUB_MAX_TEXT) break; // EOL or TooMany
          line_pos = 0;
      }
      switch (c) {
        case '\r':
        case '\n': // just ignore linefeeds for now
                   // their placement seems rather haphazard
          break;
        case '<': // some html markup, ignore for now
          ignoring = 1;
          break;
        case '>':
          ignoring = 0;
          break;
        default:
          if(!ignoring) 
	  ogg_sub.text[ogg_sub.lines][line_pos++] = c;
          break;
      }
    }
  }

  mp_msg(MSGT_DEMUX,MSGL_DBG2,"ogg sub lines: %d  first: '%s'\n",
      ogg_sub.lines, ogg_sub.text[0]);
#ifdef USE_ICONV
  subcp_recode1(&ogg_sub);
#endif
  vo_sub = &ogg_sub;
  vo_osd_changed(OSDTYPE_SUBTITLE);
}


// get the logical stream of the current page
// fill os if non NULL and return the stream id
static  int demux_ogg_get_page_stream(ogg_demuxer_t* ogg_d,ogg_stream_state** os) {
  int id,s_no;
  ogg_page* page = &ogg_d->page;

  s_no = ogg_page_serialno(page);

  for(id= 0; id < ogg_d->num_sub ; id++) {
    if(s_no == ogg_d->subs[id].stream.serialno)
      break;
  }

  if(id == ogg_d->num_sub) {
    // If we have only one vorbis stream allow the stream id to change
    // it's normal on radio stream (each song have an different id).
    // But we (or the codec?) should check that the samplerate, etc
    // doesn't change (for radio stream it's ok)
    if(ogg_d->num_sub == 1 && ogg_d->subs[0].vorbis) {
      ogg_stream_reset(&ogg_d->subs[0].stream);
      ogg_stream_init(&ogg_d->subs[0].stream,s_no);
      id = 0;
    } else
      return -1;
  }
  
  if(os)
    *os = &ogg_d->subs[id].stream;

  return id;

}

static unsigned char* demux_ogg_read_packet(ogg_stream_t* os,ogg_packet* pack,void *context,float* pts,int* flags) {
  unsigned char* data;

  *pts = 0;
  *flags = 0;

  if(os->vorbis) {
    data = pack->packet;
    if(*pack->packet & PACKET_TYPE_HEADER)
      os->hdr_packets++;
    else if (context )
    {
       vorbis_info *vi = &((ov_struct_t*)context)->vi;

       // When we dump the audio, there is no vi, but we dont care of timestamp in this case
       int32_t blocksize = vorbis_packet_blocksize(vi,pack) / vi->channels;
       // Calculate the timestamp if the packet don't have any
       if(pack->granulepos == -1) {
	  pack->granulepos = os->lastpos;
	  if(os->lastsize > 0)
	     pack->granulepos += os->lastsize;
       }
       *pts = pack->granulepos / (float)vi->rate;
       os->lastsize = blocksize;
       os->lastpos = pack->granulepos;
    }
# ifdef HAVE_OGGTHEORA
  } else if (os->theora) {
     /* we pass complete packets to theora, mustn't strip the header! */
     data = pack->packet;
     os->lastsize = 1;
     
     /* header packets beginn on 1-bit: thus check (*data&0x80).  We don't
	have theora_state st, until all header packets were passed to the
	decoder. */
     if (context != NULL && !(*data&0x80))
     {
	theora_state *st;
	int64_t usable_granulepos;

	st = &((theora_struct_t*)context)->st;
	*pts = theora_granule_time (st, pack->granulepos);
	if (*pts >= 0)
	{
	   usable_granulepos = (int64_t)ceil (*pts * os->samplerate - 0.5);
	   os->lastpos = usable_granulepos;
	}
	else
	{
	   os->lastpos++;
	   *pts = (double)os->lastpos / (double)os->samplerate;
	}
     }
#endif /* HAVE_OGGTHEORA */
  } else {
    // Find data start
    int16_t hdrlen = (*pack->packet & PACKET_LEN_BITS01)>>6;
    hdrlen |= (*pack->packet & PACKET_LEN_BITS2) <<1;
    data = pack->packet + 1 + hdrlen;
    // Calculate the timestamp
    if(pack->granulepos == -1)
      pack->granulepos = os->lastpos + os->lastsize;
    // If we alredy have a timestamp it can be a syncpoint
    if(*pack->packet & PACKET_IS_SYNCPOINT)
      *flags = 1;
    *pts =  pack->granulepos/os->samplerate;
    // Save the packet length and timestamp
    os->lastsize = 0;
    while(hdrlen) {
      os->lastsize <<= 8;
      os->lastsize |= pack->packet[hdrlen];
      hdrlen--;
    }
    os->lastpos = pack->granulepos;
  }
  return data;
}

/// Calculate the timestamp and add the packet to the demux stream
// return 1 if the packet was added, 0 otherwise
static int demux_ogg_add_packet(demux_stream_t* ds,ogg_stream_t* os,ogg_packet* pack) {
  demuxer_t* d = ds->demuxer;
  demux_packet_t* dp;
  unsigned char* data;
  float pts = 0;
  int flags = 0;
  void *context = NULL;

  if (ds == d->sub) { // don't want to add subtitles to the demuxer for now
    demux_ogg_add_sub(os,pack);
    return 0;
  }
  // If packet is an header we jump it except for vorbis and theora
  // (PACKET_TYPE_HEADER bit doesn't even exist for theora ?!)
  if((*pack->packet & PACKET_TYPE_HEADER) && 
     (ds != d->audio || ( ((sh_audio_t*)ds->sh)->format != 0xFFFE || os->hdr_packets >= NUM_VORBIS_HDR_PACKETS ) ) &&
     (ds != d->video || (((sh_video_t*)ds->sh)->format != 0xFFFC)))
    return 0;

  // For vorbis packet the packet is the data, for other codec we must jump
  // the header
  if(ds == d->audio && ((sh_audio_t*)ds->sh)->format == 0xFFFE)
     context = ((sh_audio_t *)ds->sh)->context;
  if (ds == d->video && ((sh_audio_t*)ds->sh)->format == 0xFFFC)
     context = ((sh_video_t *)ds->sh)->context;
  data = demux_ogg_read_packet(os,pack,context,&pts,&flags);

  /// Clear subtitles if necessary (for broken files)
  if ((clear_sub > 0) && (pts >= clear_sub)) {
    ogg_sub.lines = 0;
    vo_sub = &ogg_sub;
    vo_osd_changed(OSDTYPE_SUBTITLE);
    clear_sub = -1;
  }
  /// Send the packet
  dp = new_demux_packet(pack->bytes-(data-pack->packet));
  memcpy(dp->buffer,data,pack->bytes-(data-pack->packet));
  dp->pts = pts;
  dp->flags = flags;
  ds_add_packet(ds,dp);
  if(verbose>1)printf("New dp: %p  ds=%p  pts=%5.3f  len=%d  flag=%d  \n",
      dp, ds, pts, dp->len, flags);
  return 1;
}

/// Build a table of all syncpoints to make seeking easier
void demux_ogg_build_syncpoints_table(demuxer_t* demuxer) {
  ogg_demuxer_t* ogg_d = demuxer->priv;
  stream_t *s = demuxer->stream;
  ogg_sync_state* sync = &ogg_d->sync;
  ogg_page* page= &ogg_d->page;
  ogg_stream_state* oss;
  ogg_stream_t* os;
  ogg_packet op;
  int np,sid,p;
  void *context = NULL;
  off_t pos, last_pos;
  pos = last_pos = demuxer->movi_start;

  // Reset the stream
  stream_seek(s,demuxer->movi_start);
  ogg_sync_reset(sync);

  // Get the serial number of the stream we use
  if(demuxer->video->id >= 0) {
    sid = demuxer->video->id;
    /* demux_ogg_read_packet needs decoder context for Theora streams */
    if (((sh_video_t*)demuxer->video->sh)->format == 0xFFFC)
      context = ((sh_video_t*)demuxer->video->sh)->context;
  }
  else {
    sid = demuxer->audio->id;
    /* demux_ogg_read_packet needs decoder context for Vorbis streams */
    if(((sh_audio_t*)demuxer->audio->sh)->format == 0xFFFE)
      context = ((sh_audio_t*)demuxer->audio->sh)->context;
  }
  os = &ogg_d->subs[sid];
  oss = &os->stream;

  while(1) {
    np = ogg_sync_pageseek(sync,page);
    if(np < 0) { // We had to skip some bytes
      mp_msg(MSGT_DEMUX,MSGL_ERR,"Bad page sync while building syncpoints table (%ld)\n",-np);
      pos += -np;
      continue;
    }
    if(np <= 0) { // We need more data
      char* buf = ogg_sync_buffer(sync,BLOCK_SIZE);
      int len = stream_read(s,buf,BLOCK_SIZE);
      if(len == 0 && s->eof)
	break;
      ogg_sync_wrote(sync,len);
      continue;
    }
    // The page is ready
    //ogg_sync_pageout(sync,page);
    if(ogg_page_serialno(page) != os->stream.serialno) { // It isn't a page from the stream we want
      pos += np;
      continue;
    }
    if(ogg_stream_pagein(oss,page) != 0) {
      mp_msg(MSGT_DEMUX,MSGL_ERR,"Pagein error ????\n");
      pos += np;
      continue;
    }
    p = 0;
    while(ogg_stream_packetout(oss,&op) == 1) {
      float pts;
      int flags;
      demux_ogg_read_packet(os,&op,context,&pts,&flags);
      if(flags || (os->vorbis && op.granulepos >= 0)) {
	ogg_d->syncpoints = (ogg_syncpoint_t*)realloc(ogg_d->syncpoints,(ogg_d->num_syncpoint+1)*sizeof(ogg_syncpoint_t));
	ogg_d->syncpoints[ogg_d->num_syncpoint].granulepos = op.granulepos;
	ogg_d->syncpoints[ogg_d->num_syncpoint].page_pos = (ogg_page_continued(page) && p == 0) ? last_pos : pos;
	ogg_d->num_syncpoint++;
      }
      p++;
    }
    if(p > 1 || (p == 1 && ! ogg_page_continued(page)))
      last_pos = pos;
    pos += np;
    mp_msg(MSGT_DEMUX,MSGL_INFO,"Building syncpoint table %d%\r",pos*100/s->end_pos);
  }
  mp_msg(MSGT_DEMUX,MSGL_INFO,"\n");

  mp_msg(MSGT_DEMUX,MSGL_V,"Ogg syncpoints table builed: %d syncpoints\n",ogg_d->num_syncpoint);

  stream_reset(s);
  stream_seek(s,demuxer->movi_start);
  ogg_sync_reset(sync);
  for(np = 0 ; np < ogg_d->num_sub ; np++) {
    ogg_stream_reset(&ogg_d->subs[np].stream);
    ogg_d->subs[np].lastpos = ogg_d->subs[np].lastsize = ogg_d->subs[np].hdr_packets = 0;
  }

  
  // Get the first page
  while(1) {
    np = ogg_sync_pageout(sync,page);
    if(np <= 0) { // We need more data
      char* buf = ogg_sync_buffer(sync,BLOCK_SIZE);
      int len = stream_read(s,buf,BLOCK_SIZE);
      if(len == 0 && s->eof) {
	mp_msg(MSGT_DEMUX,MSGL_ERR,"EOF while trying to get the first page !!!!\n");
	break;
      }
      
      ogg_sync_wrote(sync,len);
      continue;
    }
    demux_ogg_get_page_stream(ogg_d,&oss);
    ogg_stream_pagein(oss,page);
    break;
  }
  
}

extern void print_wave_header(WAVEFORMATEX *h);
extern void print_video_header(BITMAPINFOHEADER *h);

/// Open an ogg physical stream
int demux_ogg_open(demuxer_t* demuxer) {
  ogg_demuxer_t* ogg_d;
  stream_t *s;
  char* buf;
  int np,s_no, n_audio = 0, n_video = 0, n_text = 0;
  ogg_sync_state* sync;
  ogg_page* page;
  ogg_packet pack;
  sh_audio_t* sh_a;
  sh_video_t* sh_v;

#ifdef USE_ICONV
  subcp_open();
#endif

  clear_sub = -1;
  s = demuxer->stream;

  ogg_d = (ogg_demuxer_t*)calloc(1,sizeof(ogg_demuxer_t));
  sync = &ogg_d->sync;
  page = &ogg_d->page;

  ogg_sync_init(sync);

  while(1) {
    /// Try to get a page
    ogg_d->pos += ogg_d->last_size;
    np = ogg_sync_pageseek(sync,page);
    /// Error
    if(np < 0) {
      mp_msg(MSGT_DEMUX,MSGL_DBG2,"OGG demuxer : Bad page sync\n");
      free(ogg_d);
      return 0;
    }
    /// Need some more data
    if(np == 0) {
      int len;
      buf = ogg_sync_buffer(sync,BLOCK_SIZE);
      len = stream_read(s,buf,BLOCK_SIZE);      
      if(len == 0 && s->eof) {
	free(ogg_d);
	return 0;
      }
      ogg_sync_wrote(sync,len);
      continue;
    }
    ogg_d->last_size = np;
    // We got one page now

    if( ! ogg_page_bos(page) ) { // It's not a begining page
      // Header parsing end here, we need to get the page otherwise it will be lost
      int id = demux_ogg_get_page_stream(ogg_d,NULL);
      if(id >= 0)
	ogg_stream_pagein(&ogg_d->subs[id].stream,page);
      else
	mp_msg(MSGT_DEMUX,MSGL_ERR,"OGG : Warning found none bos page from unknown stream %d\n",ogg_page_serialno(page));
      break;
    }

    /// Init  the data structure needed for a logical stream
    ogg_d->subs = (ogg_stream_t*)realloc(ogg_d->subs,(ogg_d->num_sub+1)*sizeof(ogg_stream_t));
    memset(&ogg_d->subs[ogg_d->num_sub],0,sizeof(ogg_stream_t));
    /// Get the stream serial number
    s_no = ogg_page_serialno(page);
    ogg_stream_init(&ogg_d->subs[ogg_d->num_sub].stream,s_no);
    mp_msg(MSGT_DEMUX,MSGL_DBG2,"OGG : Found a stream with serial=%d\n",s_no);
    // Take the first page
    ogg_stream_pagein(&ogg_d->subs[ogg_d->num_sub].stream,page);
    // Get first packet of the page
    ogg_stream_packetout(&ogg_d->subs[ogg_d->num_sub].stream,&pack);

    // Reset our vars
    sh_a = NULL;
    sh_v = NULL;

    // Check for Vorbis
    if(pack.bytes >= 7 && ! strncmp(&pack.packet[1],"vorbis", 6) ) {
      sh_a = new_sh_audio(demuxer,ogg_d->num_sub);
      sh_a->format = 0xFFFE;
      ogg_d->subs[ogg_d->num_sub].vorbis = 1;
      n_audio++;
      mp_msg(MSGT_DEMUX,MSGL_V,"OGG : stream %d is vorbis\n",ogg_d->num_sub);

      // check for Theora
#   ifdef HAVE_OGGTHEORA
    } else if (pack.bytes >= 7 && !strncmp (&pack.packet[1], "theora", 6)) {
	int errorCode = 0;
	theora_info inf;
	theora_comment cc;
	
	theora_info_init (&inf);
	theora_comment_init (&cc);
	
	errorCode = theora_decode_header (&inf, &cc, &pack);
	if (errorCode)
	    mp_msg(MSGT_DEMUX,MSGL_ERR,"Theora header parsing failed: %i \n",
		   errorCode);
	else
	{
	    sh_v = new_sh_video(demuxer,ogg_d->num_sub);

	    sh_v->context = NULL;
	    sh_v->bih = (BITMAPINFOHEADER*)calloc(1,sizeof(BITMAPINFOHEADER));
	    sh_v->bih->biSize=sizeof(BITMAPINFOHEADER);
	    sh_v->bih->biCompression= sh_v->format = 0xFFFC;
	    sh_v->fps = ((double)inf.fps_numerator)/
		(double)inf.fps_denominator;
	    sh_v->frametime = ((double)inf.fps_denominator)/
		(double)inf.fps_numerator;
	    sh_v->disp_w = sh_v->bih->biWidth = inf.width;
	    sh_v->disp_h = sh_v->bih->biHeight = inf.height;
	    sh_v->bih->biBitCount = 24;
	    sh_v->bih->biPlanes = 3;
	    sh_v->bih->biSizeImage =