mp4.c

video linux conference

        break;

    case VLC_FOURCC('m','p','g','a'):
        if( p_sys->b_mov )
            memcpy( fcc, ".mp3", 4 );
        else
        {
            memcpy( fcc, "mp4a", 4 );
            b_descr = VLC_TRUE;
        }
        break;

    default:
        memcpy( fcc, (char*)&p_stream->fmt.i_codec, 4 );
        break;
    }

    soun = box_new( fcc );
    for( i = 0; i < 6; i++ )
    {
        bo_add_8( soun, 0 );        // reserved;
    }
    bo_add_16be( soun, 1 );         // data-reference-index

    /* SoundDescription */
    if( p_sys->b_mov &&
        p_stream->fmt.i_codec == VLC_FOURCC('m','p','4','a') )
    {
        bo_add_16be( soun, 1 );     // version 1;
    }
    else
    {
        bo_add_16be( soun, 0 );     // version 0;
    }
    bo_add_16be( soun, 0 );         // revision level (0)
    bo_add_32be( soun, 0 );         // vendor
    // channel-count
    bo_add_16be( soun, p_stream->fmt.audio.i_channels );
    // sample size
    bo_add_16be( soun, p_stream->fmt.audio.i_bitspersample ?
                 p_stream->fmt.audio.i_bitspersample : 16 );
    bo_add_16be( soun, -2 );        // compression id
    bo_add_16be( soun, 0 );         // packet size (0)
    bo_add_16be( soun, p_stream->fmt.audio.i_rate ); // sampleratehi
    bo_add_16be( soun, 0 );                             // sampleratelo

    /* Extended data for SoundDescription V1 */
    if( p_sys->b_mov &&
        p_stream->fmt.i_codec == VLC_FOURCC('m','p','4','a') )
    {
        /* samples per packet */
        bo_add_32be( soun, p_stream->fmt.audio.i_frame_length );
        bo_add_32be( soun, 1536 ); /* bytes per packet */
        bo_add_32be( soun, 2 );    /* bytes per frame */
        /* bytes per sample */
        bo_add_32be( soun, 2 /*p_stream->fmt.audio.i_bitspersample/8 */);
    }

    /* Add an ES Descriptor */
    if( b_descr )
    {
        bo_t *box;

        if( p_sys->b_mov &&
            p_stream->fmt.i_codec == VLC_FOURCC('m','p','4','a') )
        {
            box = GetWaveTag( p_stream );
        }
        else
        {
            box = GetESDS( p_stream );
        }
        box_fix( box );
        box_gather( soun, box );
    }

    box_fix( soun );

    return soun;
}

static bo_t *GetVideBox( sout_mux_t *p_mux, mp4_stream_t *p_stream )
{

    bo_t *vide;
    char fcc[4] = "    ";
    int  i;

    switch( p_stream->fmt.i_codec )
    {
    case VLC_FOURCC('m','p','4','v'):
    case VLC_FOURCC('m','p','g','v'):
        memcpy( fcc, "mp4v", 4 );
        break;

    case VLC_FOURCC('M','J','P','G'):
        memcpy( fcc, "mjpa", 4 );
        break;

    case VLC_FOURCC('S','V','Q','1'):
        memcpy( fcc, "SVQ1", 4 );
        break;

    case VLC_FOURCC('S','V','Q','3'):
        memcpy( fcc, "SVQ3", 4 );
        break;

    case VLC_FOURCC('h','2','6','4'):
        memcpy( fcc, "avc1", 4 );
        break;

    default:
        memcpy( fcc, (char*)&p_stream->fmt.i_codec, 4 );
        break;
    }

    vide = box_new( fcc );
    for( i = 0; i < 6; i++ )
    {
        bo_add_8( vide, 0 );        // reserved;
    }
    bo_add_16be( vide, 1 );         // data-reference-index

    bo_add_16be( vide, 0 );         // predefined;
    bo_add_16be( vide, 0 );         // reserved;
    for( i = 0; i < 3; i++ )
    {
        bo_add_32be( vide, 0 );     // predefined;
    }

    bo_add_16be( vide, p_stream->fmt.video.i_width );  // i_width
    bo_add_16be( vide, p_stream->fmt.video.i_height ); // i_height

    bo_add_32be( vide, 0x00480000 );                // h 72dpi
    bo_add_32be( vide, 0x00480000 );                // v 72dpi

    bo_add_32be( vide, 0 );         // data size, always 0
    bo_add_16be( vide, 1 );         // frames count per sample

    // compressor name;
    for( i = 0; i < 32; i++ )
    {
        bo_add_8( vide, 0 );
    }

    bo_add_16be( vide, 0x18 );      // depth
    bo_add_16be( vide, 0xffff );    // predefined

    /* add an ES Descriptor */
    switch( p_stream->fmt.i_codec )
    {
    case VLC_FOURCC('m','p','4','v'):
    case VLC_FOURCC('m','p','g','v'):
        {
            bo_t *esds = GetESDS( p_stream );

            box_fix( esds );
            box_gather( vide, esds );
        }
        break;

    case VLC_FOURCC('S','V','Q','3'):
        {
            bo_t *esds = GetSVQ3Tag( p_stream );

            box_fix( esds );
            box_gather( vide, esds );
        }
        break;

    case VLC_FOURCC('h','2','6','4'):
        box_gather( vide, GetAvcCTag( p_stream ) );
        break;

    default:
        break;
    }

    box_fix( vide );

    return vide;
}

static bo_t *GetTextBox( sout_mux_t *p_mux, mp4_stream_t *p_stream )
{

    bo_t *text = box_new( "text" );
    int  i;

    for( i = 0; i < 6; i++ )
    {
        bo_add_8( text, 0 );        // reserved;
    }
    bo_add_16be( text, 1 );         // data-reference-index

    bo_add_32be( text, 0 );         // display flags
    bo_add_32be( text, 0 );         // justification
    for( i = 0; i < 3; i++ )
    {
        bo_add_16be( text, 0 );     // back ground color
    }

    bo_add_16be( text, 0 );         // box text
    bo_add_16be( text, 0 );         // box text
    bo_add_16be( text, 0 );         // box text
    bo_add_16be( text, 0 );         // box text

    bo_add_64be( text, 0 );         // reserved
    for( i = 0; i < 3; i++ )
    {
        bo_add_16be( text, 0xff );  // foreground color
    }

    bo_add_8 ( text, 9 );
    bo_add_mem( text, 9, "Helvetica" );

    box_fix( text );

    return text;
}

static bo_t *GetStblBox( sout_mux_t *p_mux, mp4_stream_t *p_stream )
{
    sout_mux_sys_t *p_sys = p_mux->p_sys;
    unsigned int i_chunk, i_stsc_last_val, i_stsc_entries, i, i_index;
    bo_t *stbl, *stsd, *stts, *stco, *stsc, *stsz, *stss;
    uint32_t i_timescale;
    int64_t i_dts, i_dts_q;

    stbl = box_new( "stbl" );

    /* sample description */
    stsd = box_full_new( "stsd", 0, 0 );
    bo_add_32be( stsd, 1 );
    if( p_stream->fmt.i_cat == AUDIO_ES )
    {
        bo_t *soun = GetSounBox( p_mux, p_stream );
        box_gather( stsd, soun );
    }
    else if( p_stream->fmt.i_cat == VIDEO_ES )
    {
        bo_t *vide = GetVideBox( p_mux, p_stream );
        box_gather( stsd, vide );
    }
    else if( p_stream->fmt.i_cat == SPU_ES )
    {
        box_gather( stsd, GetTextBox( p_mux, p_stream ) );
    }
    box_fix( stsd );

    /* chunk offset table */
    if( p_sys->i_pos >= (((uint64_t)0x1) << 32) )
    {
        /* 64 bits version */
        p_stream->b_stco64 = VLC_TRUE;
        stco = box_full_new( "co64", 0, 0 );
    }
    else
    {
        /* 32 bits version */
        p_stream->b_stco64 = VLC_FALSE;
        stco = box_full_new( "stco", 0, 0 );
    }
    bo_add_32be( stco, 0 );     // entry-count (fixed latter)

    /* sample to chunk table */
    stsc = box_full_new( "stsc", 0, 0 );
    bo_add_32be( stsc, 0 );     // entry-count (fixed latter)

    for( i_chunk = 0, i_stsc_last_val = 0, i_stsc_entries = 0, i = 0;
         i < p_stream->i_entry_count; i_chunk++ )
    {
        int i_first = i;

        if( p_stream->b_stco64 )
            bo_add_64be( stco, p_stream->entry[i].i_pos );
        else
            bo_add_32be( stco, p_stream->entry[i].i_pos );

        while( i < p_stream->i_entry_count )
        {
            if( i + 1 < p_stream->i_entry_count &&
                p_stream->entry[i].i_pos + p_stream->entry[i].i_size
                != p_stream->entry[i + 1].i_pos )
            {
                i++;
                break;
            }

            i++;
        }

        /* Add entry to the stsc table */
        if( i_stsc_last_val != i - i_first )
        {
            bo_add_32be( stsc, 1 + i_chunk );   // first-chunk
            bo_add_32be( stsc, i - i_first ) ;  // samples-per-chunk
            bo_add_32be( stsc, 1 );             // sample-descr-index
            i_stsc_last_val = i - i_first;
            i_stsc_entries++;
        }
    }

    /* Fix stco entry count */
    bo_fix_32be( stco, 12, i_chunk );
    msg_Dbg( p_mux, "created %d chunks (stco)", i_chunk );
    box_fix( stco );

    /* Fix stsc entry count */
    bo_fix_32be( stsc, 12, i_stsc_entries  );
    box_fix( stsc );

    /* add stts */
    stts = box_full_new( "stts", 0, 0 );
    bo_add_32be( stts, 0 );     // entry-count (fixed latter)

    if( p_stream->fmt.i_cat == AUDIO_ES )
        i_timescale = p_stream->fmt.audio.i_rate;
    else
        i_timescale = 1001;

    /* first, create quantified length */
    for( i = 0, i_dts = 0, i_dts_q = 0; i < p_stream->i_entry_count; i++ )
    {
        int64_t i_dts_deq = i_dts_q * I64C(1000000) / (int64_t)i_timescale;
        int64_t i_delta = p_stream->entry[i].i_length + i_dts - i_dts_deq;

        i_dts += p_stream->entry[i].i_length;

        p_stream->entry[i].i_length =
            i_delta * (int64_t)i_timescale / I64C(1000000);

        i_dts_q += p_stream->entry[i].i_length;
    }
    /* then write encoded table */
    for( i = 0, i_index = 0; i < p_stream->i_entry_count; i_index++)
    {
        int     i_first = i;
        int64_t i_delta = p_stream->entry[i].i_length;

        while( i < p_stream->i_entry_count )
        {
            i++;
            if( i >= p_stream->i_entry_count ||
                p_stream->entry[i].i_length != i_delta )
            {
                break;
            }
        }

        bo_add_32be( stts, i - i_first ); // sample-count
        bo_add_32be( stts, i_delta );     // sample-delta
    }
    bo_fix_32be( stts, 12, i_index );
    box_fix( stts );

    /* FIXME add ctts ?? FIXME */

    stsz = box_full_new( "stsz", 0, 0 );
    bo_add_32be( stsz, 0 );                             // sample-size
    bo_add_32be( stsz, p_stream->i_entry_count );       // sample-count
    for( i = 0; i < p_stream->i_entry_count; i++ )
    {
        bo_add_32be( stsz, p_stream->entry[i].i_size ); // sample-size
    }
    box_fix( stsz );

    /* create stss table */
    stss = NULL;
    for( i = 0, i_index = 0; i < p_stream->i_entry_count; i++ )
    {
        if( p_stream->entry[i].i_flags & BLOCK_FLAG_TYPE_I )
        {
            if( stss == NULL )
            {
                stss = box_full_new( "stss", 0, 0 );
                bo_add_32be( stss, 0 ); /* fixed later */
            }
            bo_add_32be( stss, 1 + i );
            i_index++;
        }
    }
    if( stss )
    {
        bo_fix_32be( stss, 12, i_index );
        box_fix( stss );
    }

    /* Now gather all boxes into stbl */
    box_gather( stbl, stsd );
    box_gather( stbl, stts );
    if( stss )
    {
        box_gather( stbl, stss );
    }
    box_gather( stbl, stsc );
    box_gather( stbl, stsz );
    p_stream->i_stco_pos = stbl->i_buffer + 16;
    box_gather( stbl, stco );

    /* finish stbl */
    box_fix( stbl );

    return stbl;
}

static int64_t get_timestamp();

static uint32_t mvhd_matrix[9] =
    { 0x10000, 0, 0, 0, 0x10000, 0, 0, 0, 0x40000000 };

static bo_t *GetMoovBox( sout_mux_t *p_mux )
{
    sout_mux_sys_t *p_sys = p_mux->p_sys;

    bo_t            *moov, *mvhd;
    int             i_trak, i;

    uint32_t        i_movie_timescale = 90000;
    int64_t         i_movie_duration  = 0;

    moov = box_new( "moov" );

    /* Create general info */
    for( i_trak = 0; i_trak < p_sys->i_nb_streams; i_trak++ )
    {
        mp4_stream_t *p_stream = p_sys->pp_streams[i_trak];
        i_movie_duration = __MAX( i_movie_duration, p_stream->i_duration );
    }
    msg_Dbg( p_mux, "movie duration %ds",
             (uint32_t)( i_movie_duration / (mtime_t)1000000 ) );

    i_movie_duration = i_movie_duration * i_movie_timescale / 1000000;

    /* *** add /moov/mvhd *** */
    if( !p_sys->b_64_ext )
    {
        mvhd = box_full_new( "mvhd", 0, 0 );
        bo_add_32be( mvhd, get_timestamp() );   // creation time
        bo_add_32be( mvhd, get_timestamp() );   // modification time
        bo_add_32be( mvhd, i_movie_timescale);  // timescale
        bo_add_32be( mvhd, i_movie_duration );  // duration
    }
    else
    {
        mvhd = box_full_new( "mvhd", 1, 0 );
        bo_add_64be( mvhd, get_timestamp() );   // creation time
        bo_add_64be( mvhd, get_timestamp() );   // modification time
        bo_add_32be( mvhd, i_movie_timescale);  // timescale
        bo_add_64be( mvhd, i_movie_duration );  // duration
    }
    bo_add_32be( mvhd, 0x10000 );           // rate
    bo_add_16be( mvhd, 0x100 );             // volume
    bo_add_16be( mvhd, 0 );                 // reserved
    for( i = 0; i < 2; i++ )
    {
        bo_add_32be( mvhd, 0 );             // reserved
    }
    for( i = 0; i < 9; i++ )
    {
        bo_add_32be( mvhd, mvhd_matrix[i] );// matrix
    }
    for( i = 0; i < 6; i++ )
    {
        bo_add_32be( mvhd, 0 );             // pre-defined
    }

    /* Next available track id */
    bo_add_32be( mvhd, p_sys->i_nb_streams + 1 ); // next-track-id

    box_fix( mvhd );
    box_gather( moov, mvhd );

    for( i_trak = 0; i_trak < p_sys->i_nb_streams; i_trak++ )
    {
        mp4_stream_t *p_stream;
        uint32_t     i_timescale;

        bo_t *trak, *tkhd, *edts, *elst, *mdia, *mdhd, *hdlr;
        bo_t *minf, *dinf, *dref, *url, *stbl;

        p_stream = p_sys->pp_streams[i_trak];

        if( p_stream->fmt.i_cat == AUDIO_ES )
            i_timescale = p_stream->fmt.audio.i_rate;
        else
            i_timescale = 1001;

        /* *** add /moov/trak *** */
        trak = box_new( "trak" );

        /* *** add /moov/trak/tkhd *** */
        if( !p_sys->b_64_ext )
        {
            if( p_sys->b_mov )
                tkhd = box_full_new( "tkhd", 0, 0x0f );
            else
                tkhd = box_full_new( "tkhd", 0, 1 );

            bo_add_32be( tkhd, get_timestamp() );       // creation time
            bo_add_32be( tkhd, get_timestamp() );       // modification time
            bo_add_32be( tkhd, p_stream->i_track_id );
            bo_add_32be( tkhd, 0 );                     // reserved 0
            bo_add_32be( tkhd, p_stream->i_duration *
                         (int64_t)i_movie_timescale /
                         (mtime_t)1000000 );            // duration
        }
        else
        {
            if( p_sys->b_mov )
                tkhd = box_full_new( "tkhd", 1, 0x0f );
            else
                tkhd = box_full_new( "tkhd", 1, 1 );

            bo_add_64be( tkhd, get_timestamp() );       // creation time
            bo_add_64be( tkhd, get_timestamp() );       // modification time
            bo_add_32be( tkhd, p_stream->i_track_id );
            bo_add_32be( tkhd, 0 );                     // reserved 0
            bo_add_64be( tkhd, p_stream->i_duration *
                         (int64_t)i_movie_timescale /
                         (mtime_t)1000000 );            // duration
        }