mkv.cpp: I was wrong, the dirent export works
[vlc.git] / modules / packetizer / h264.c
index 5002d8c..e0855b4 100644 (file)
@@ -6,7 +6,7 @@
  *
  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
  *          Eric Petit <titer@videolan.org>
- *          Gildas Bazin <gbazin@netcourrier.com>
+ *          Gildas Bazin <gbazin@videolan.org>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -42,6 +42,8 @@ static int  Open ( vlc_object_t * );
 static void Close( vlc_object_t * );
 
 vlc_module_begin();
+    set_category( CAT_SOUT );
+    set_subcategory( SUBCAT_SOUT_PACKETIZER );
     set_description( _("H264 video packetizer") );
     set_capability( "packetizer", 50 );
     set_callbacks( Open, Close );
@@ -52,6 +54,7 @@ vlc_module_end();
  * Local prototypes
  ****************************************************************************/
 static block_t *Packetize( decoder_t *, block_t ** );
+static block_t *PacketizeAVC1( decoder_t *, block_t ** );
 
 struct decoder_sys_t
 {
@@ -69,6 +72,19 @@ struct decoder_sys_t
     unsigned int i_flags;
 
     vlc_bool_t   b_sps;
+
+    /* avcC data */
+    int i_avcC_length_size;
+
+    /* Useful values of the Sequence Parameter Set */
+    int i_log2_max_frame_num;
+    int b_frame_mbs_only;
+
+    /* Useful values of the Slice Header */
+    int i_nal_type;
+    int i_nal_ref_idc;
+    int i_idr_pic_id;
+    int i_frame_num;
 };
 
 enum
@@ -101,6 +117,8 @@ enum nal_priority_e
 
 static block_t *ParseNALBlock( decoder_t *, block_t * );
 
+static block_t *nal_get_annexeb( decoder_t *, uint8_t *p, int );
+
 /*****************************************************************************
  * Open: probe the packetizer and return score
  *****************************************************************************/
@@ -110,7 +128,11 @@ static int Open( vlc_object_t *p_this )
     decoder_sys_t *p_sys;
 
     if( p_dec->fmt_in.i_codec != VLC_FOURCC( 'h', '2', '6', '4') &&
-        p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') )
+        p_dec->fmt_in.i_codec != VLC_FOURCC( 'H', '2', '6', '4') &&
+        p_dec->fmt_in.i_codec != VLC_FOURCC( 'V', 'S', 'S', 'H') &&
+        p_dec->fmt_in.i_codec != VLC_FOURCC( 'v', 's', 's', 'h') &&
+        ( p_dec->fmt_in.i_codec != VLC_FOURCC( 'a', 'v', 'c', '1') ||
+          p_dec->fmt_in.i_extra < 7 ) )
     {
         return VLC_EGENERIC;
     }
@@ -135,33 +157,59 @@ static int Open( vlc_object_t *p_this )
     p_sys->i_flags = 0;
     p_sys->b_sps   = VLC_FALSE;
 
+    p_sys->i_nal_type = -1;
+    p_sys->i_nal_ref_idc = -1;
+    p_sys->i_idr_pic_id = -1;
+    p_sys->i_frame_num = -1;
+
     /* Setup properties */
     es_format_Copy( &p_dec->fmt_out, &p_dec->fmt_in );
     p_dec->fmt_out.i_codec = VLC_FOURCC( 'h', '2', '6', '4' );
+    /* FIXME: FFMPEG isn't happy at all if you leave this */
+    if( p_dec->fmt_out.i_extra ) free( p_dec->fmt_out.p_extra );
+    p_dec->fmt_out.i_extra = 0; p_dec->fmt_out.p_extra = 0;
 
-#if 0
-    if( p_dec->fmt_in.i_extra )
+    if( p_dec->fmt_in.i_codec == VLC_FOURCC( 'a', 'v', 'c', '1' ) )
     {
-        /* We have a vol */
-        p_dec->fmt_out.i_extra = p_dec->fmt_in.i_extra;
-        p_dec->fmt_out.p_extra = malloc( p_dec->fmt_in.i_extra );
-        memcpy( p_dec->fmt_out.p_extra, p_dec->fmt_in.p_extra,
-                p_dec->fmt_in.i_extra );
-
-        msg_Dbg( p_dec, "opening with vol size:%d", p_dec->fmt_in.i_extra );
-        m4v_VOLParse( &p_dec->fmt_out,
-                      p_dec->fmt_out.p_extra, p_dec->fmt_out.i_extra );
+        uint8_t *p = &((uint8_t*)p_dec->fmt_in.p_extra)[4];
+        int i_sps, i_pps;
+        int i;
+
+        /* Parse avcC */
+        p_sys->i_avcC_length_size = 1 + ((*p++)&0x03);
+
+        /* Read SPS */
+        i_sps = (*p++)&0x1f;
+
+        for( i = 0; i < i_sps; i++ )
+        {
+            int i_length = GetWBE( p );
+            block_t *p_sps = nal_get_annexeb( p_dec, p + 2, i_length );
+
+            ParseNALBlock( p_dec, p_sps );
+            p += 2 + i_length;
+        }
+        /* Read PPS */
+        i_pps = *p++;
+        for( i = 0; i < i_pps; i++ )
+        {
+            int i_length = GetWBE( p );
+            block_t *p_pps = nal_get_annexeb( p_dec, p + 2, i_length );
+
+            ParseNALBlock( p_dec, p_pps );
+            p += 2 + i_length;
+        }
+        msg_Dbg( p_dec, "avcC length size=%d sps=%d pps=%d",
+                 p_sys->i_avcC_length_size, i_sps, i_pps );
+
+        /* Set callback */
+        p_dec->pf_packetize = PacketizeAVC1;
     }
     else
     {
-        /* No vol, we'll have to look for one later on */
-        p_dec->fmt_out.i_extra = 0;
-        p_dec->fmt_out.p_extra = 0;
+        /* Set callback */
+        p_dec->pf_packetize = Packetize;
     }
-#endif
-
-    /* Set callback */
-    p_dec->pf_packetize = Packetize;
 
     return VLC_SUCCESS;
 }
@@ -196,7 +244,7 @@ static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
         {
             case STATE_NOSYNC:
                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
-                        &p_sys->i_offset, p_sys->startcode, 4 ) == VLC_SUCCESS )
+                      &p_sys->i_offset, p_sys->startcode+1, 3 ) == VLC_SUCCESS)
                 {
                     p_sys->i_state = STATE_NEXT_SYNC;
                 }
@@ -219,10 +267,15 @@ static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
             case STATE_NEXT_SYNC:
                 /* Find the next startcode */
                 if( block_FindStartcodeFromOffset( &p_sys->bytestream,
-                        &p_sys->i_offset, p_sys->startcode, 4 ) != VLC_SUCCESS )
+                      &p_sys->i_offset, p_sys->startcode, 3 ) != VLC_SUCCESS)
                 {
-                    /* Need more data */
-                    return NULL;
+                    if( block_FindStartcodeFromOffset( &p_sys->bytestream,
+                          &p_sys->i_offset, p_sys->startcode+1, 3 ) !=
+                        VLC_SUCCESS )
+                    {
+                        /* Need more data */
+                        return NULL;
+                    }
                 }
 
                 /* Get the new fragment and set the pts/dts */
@@ -252,7 +305,70 @@ static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
     }
 }
 
-static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret, uint8_t *src, int i_src )
+/****************************************************************************
+ * PacketizeAVC1: the whole thing
+ ****************************************************************************/
+static block_t *PacketizeAVC1( decoder_t *p_dec, block_t **pp_block )
+{
+    decoder_sys_t *p_sys = p_dec->p_sys;
+    block_t       *p_block;
+    block_t       *p_ret = NULL;
+    uint8_t       *p;
+
+    if( !pp_block || !*pp_block ) return NULL;
+
+    p_block = *pp_block;
+    *pp_block = NULL;
+
+    for( p = p_block->p_buffer; p < &p_block->p_buffer[p_block->i_buffer]; )
+    {
+        block_t *p_pic;
+        int i_size = 0;
+        int i;
+
+        for( i = 0; i < p_sys->i_avcC_length_size; i++ )
+        {
+            i_size = (i_size << 8) | (*p++);
+        }
+
+        if( i_size > 0 )
+        {
+            block_t *p_part = nal_get_annexeb( p_dec, p, i_size );
+
+            p_part->i_dts = p_block->i_dts;
+            p_part->i_pts = p_block->i_pts;
+
+            /* Parse the NAL */
+            if( ( p_pic = ParseNALBlock( p_dec, p_part ) ) )
+            {
+                block_ChainAppend( &p_ret, p_pic );
+            }
+        }
+        p += i_size;
+    }
+
+    return p_ret;
+}
+
+static block_t *nal_get_annexeb( decoder_t *p_dec, uint8_t *p, int i_size )
+{
+    block_t *p_nal;
+
+    p_nal = block_New( p_dec, 3 + i_size );
+
+    /* Add start code */
+    p_nal->p_buffer[0] = 0x00;
+    p_nal->p_buffer[1] = 0x00;
+    p_nal->p_buffer[2] = 0x01;
+
+    /* Copy nalu */
+    memcpy( &p_nal->p_buffer[3], p, i_size );
+
+    return p_nal;
+}
+
+static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret,
+                             uint8_t *src, int i_src )
 {
     uint8_t *end = &src[i_src];
     uint8_t *dst = malloc( i_src );
@@ -261,7 +377,8 @@ static void nal_get_decoded( uint8_t **pp_ret, int *pi_ret, uint8_t *src, int i_
 
     while( src < end )
     {
-        if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00  && src[2] == 0x03 )
+        if( src < end - 3 && src[0] == 0x00 && src[1] == 0x00 &&
+            src[2] == 0x03 )
         {
             *dst++ = 0x00;
             *dst++ = 0x00;
@@ -285,7 +402,8 @@ static inline int bs_read_ue( bs_t *s )
     }
     return( ( 1 << i) - 1 + bs_read( s, i ) );
 }
-static inline int bs_read_se( bs_t *s ) 
+
+static inline int bs_read_se( bs_t *s )
 {
     int val = bs_read_ue( s );
 
@@ -298,68 +416,114 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
     decoder_sys_t *p_sys = p_dec->p_sys;
     block_t *p_pic = NULL;
 
-    const int i_ref_idc = (p_frag->p_buffer[4] >> 5)&0x03;
-    const int i_nal_type= p_frag->p_buffer[4]&0x1f;
+    const int i_nal_ref_idc = (p_frag->p_buffer[3] >> 5)&0x03;
+    const int i_nal_type = p_frag->p_buffer[3]&0x1f;
 
-    if( p_sys->b_slice &&
-        ( i_nal_type == NAL_SLICE || i_nal_type == NAL_SLICE_IDR ||
-          i_nal_type == NAL_SLICE_DPC || i_nal_type == NAL_SPS || i_nal_type == NAL_PPS ) )
+    if( p_sys->b_slice && !p_sys->b_sps )
     {
-        if( p_sys->b_sps )
-        {
-            p_pic = block_ChainGather( p_sys->p_frame );
-            p_pic->i_dts = p_sys->i_dts;
-            p_pic->i_pts = p_sys->i_pts;
-            p_pic->i_length = 0;    /* FIXME */
-            p_pic->i_flags = p_sys->i_flags;
-        }
-        else
-        {
-            block_ChainRelease( p_sys->p_frame );
-        }
+        block_ChainRelease( p_sys->p_frame );
+        msg_Warn( p_dec, "waiting for SPS" );
 
-        /* reset context */
+        /* Reset context */
         p_sys->p_frame = NULL;
         p_sys->b_slice = VLC_FALSE;
-        //p_sys->i_dts += 40000;
     }
 
-    if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
+    if( !p_sys->b_sps &&
+        i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
+    {
+        p_sys->b_slice = VLC_TRUE;
+        /* Fragment will be discarded later on */
+    }
+    else if( i_nal_type >= NAL_SLICE && i_nal_type <= NAL_SLICE_IDR )
     {
         uint8_t *dec;
-        int     i_dec;
+        int i_dec, i_first_mb, i_slice_type, i_frame_num, i_pic_flags = 0;
+        vlc_bool_t b_pic = VLC_FALSE;
         bs_t s;
 
-        p_sys->b_slice = VLC_TRUE;
-        p_sys->i_dts   = p_frag->i_dts;
-        p_sys->i_pts   = p_frag->i_pts;
-
         /* do not convert the whole frame */
-        nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], __MIN( p_frag->i_buffer - 5, 60 ) );
+        nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
+                         __MIN( p_frag->i_buffer - 4, 60 ) );
         bs_init( &s, dec, i_dec );
 
-        /* i_first_mb */
-        bs_read_ue( &s );
-        /* picture type */
-        switch( bs_read_ue( &s ) )
+        /* first_mb_in_slice */
+        i_first_mb = bs_read_ue( &s );
+
+        /* slice_type */
+        switch( (i_slice_type = bs_read_ue( &s )) )
         {
             case 0: case 5:
-                p_sys->i_flags = BLOCK_FLAG_TYPE_P;
+                i_pic_flags = BLOCK_FLAG_TYPE_P;
                 break;
             case 1: case 6:
-                p_sys->i_flags =BLOCK_FLAG_TYPE_B;
+                i_pic_flags = BLOCK_FLAG_TYPE_B;
                 break;
             case 2: case 7:
-                p_sys->i_flags = BLOCK_FLAG_TYPE_I;
+                i_pic_flags = BLOCK_FLAG_TYPE_I;
                 break;
             case 3: case 8: /* SP */
-                p_sys->i_flags = BLOCK_FLAG_TYPE_P;
+                i_pic_flags = BLOCK_FLAG_TYPE_P;
                 break;
             case 4: case 9:
-                p_sys->i_flags = BLOCK_FLAG_TYPE_I;
+                i_pic_flags = BLOCK_FLAG_TYPE_I;
                 break;
         }
 
+        /* pic_parameter_set_id */
+        bs_read_ue( &s );
+        /* frame_num */
+        i_frame_num = bs_read( &s, p_sys->i_log2_max_frame_num + 4 );
+
+        /* Detection of the first VCL NAL unit of a primary coded picture
+         * (cf. 7.4.1.2.4) */
+        if( i_frame_num != p_sys->i_frame_num ||
+            ( (i_nal_ref_idc != p_sys->i_nal_ref_idc) &&
+              (!i_nal_ref_idc || !p_sys->i_nal_ref_idc) ) )
+        {
+            b_pic = VLC_TRUE;
+        }
+        p_sys->i_frame_num = i_frame_num;
+        p_sys->i_nal_ref_idc = i_nal_ref_idc;
+
+        if( !p_sys->b_frame_mbs_only )
+        {
+            /* field_pic_flag */
+            if( bs_read( &s, 1 ) )
+            {
+                /* bottom_field_flag */
+                bs_read( &s, 1 );
+            }
+        }
+
+        if( i_nal_type == NAL_SLICE_IDR )
+        {
+            /* id_pic_id */
+            int i_idr_pic_id = bs_read_ue( &s );
+            if( p_sys->i_nal_type != i_nal_type ) b_pic = VLC_TRUE;
+            if( p_sys->i_idr_pic_id != i_idr_pic_id ) b_pic = VLC_TRUE;
+            p_sys->i_idr_pic_id = i_idr_pic_id;
+        }
+        p_sys->i_nal_type = i_nal_type;
+
+        if( b_pic && p_sys->b_slice )
+        {
+            p_pic = block_ChainGather( p_sys->p_frame );
+            p_pic->i_dts = p_sys->i_dts;
+            p_pic->i_pts = p_sys->i_pts;
+            p_pic->i_length = 0;    /* FIXME */
+            p_pic->i_flags = p_sys->i_flags;
+
+            /* Reset context */
+            p_sys->p_frame = NULL;
+            p_sys->b_slice = VLC_FALSE;
+        }
+
+        p_sys->b_slice = VLC_TRUE;
+        p_sys->i_flags = i_pic_flags;
+        p_sys->i_dts   = p_frag->i_dts;
+        p_sys->i_pts   = p_frag->i_pts;
+
         free( dec );
     }
     else if( i_nal_type == NAL_SPS )
@@ -369,9 +533,12 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
         bs_t s;
         int i_tmp;
 
+        msg_Dbg( p_dec, "found NAL_SPS" );
+
         p_sys->b_sps = VLC_TRUE;
 
-        nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
+        nal_get_decoded( &dec, &i_dec, &p_frag->p_buffer[4],
+                         p_frag->i_buffer - 4 );
 
         bs_init( &s, dec, i_dec );
         /* Skip profile(8), constraint_set012, reserver(5), level(8) */
@@ -379,7 +546,7 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
         /* sps id */
         bs_read_ue( &s );
         /* Skip i_log2_max_frame_num */
-        bs_read_ue( &s );
+        p_sys->i_log2_max_frame_num = bs_read_ue( &s );
         /* Read poc_type */
         i_tmp = bs_read_ue( &s );
         if( i_tmp == 0 )
@@ -415,26 +582,26 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
         p_dec->fmt_out.video.i_height = 16 * ( bs_read_ue( &s ) + 1 );
 
         /* b_frame_mbs_only */
-        i_tmp = bs_read( &s, 1 );
-        if( i_tmp == 0 )
+        p_sys->b_frame_mbs_only = bs_read( &s, 1 );
+        if( p_sys->b_frame_mbs_only == 0 )
         {
             bs_skip( &s, 1 );
         }
         /* b_direct8x8_inference */
         bs_skip( &s, 1 );
 
-        /* crop */
+        /* crop */
         i_tmp = bs_read( &s, 1 );
         if( i_tmp )
         {
             /* left */
-            p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
+            bs_read_ue( &s );
             /* right */
-            p_dec->fmt_out.video.i_width -= 2 * bs_read_ue( &s );
+            bs_read_ue( &s );
             /* top */
-            p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
+            bs_read_ue( &s );
             /* bottom */
-            p_dec->fmt_out.video.i_height -= 2 * bs_read_ue( &s );
+            bs_read_ue( &s );
         }
 
         /* vui */
@@ -466,9 +633,8 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
                     h = bs_read( &s, 16 );
                 }
                 p_dec->fmt_out.video.i_aspect =
-                    VOUT_ASPECT_FACTOR *
-                    w / h *
-                    p_dec->fmt_out.video.i_width / p_dec->fmt_out.video.i_height;
+                    VOUT_ASPECT_FACTOR * w / h * p_dec->fmt_out.video.i_width /
+                    p_dec->fmt_out.video.i_height;
             }
         }
 
@@ -477,15 +643,14 @@ static block_t *ParseNALBlock( decoder_t *p_dec, block_t *p_frag )
     else if( i_nal_type == NAL_PPS )
     {
         bs_t s;
-        bs_init( &s, &p_frag->p_buffer[5], p_frag->i_buffer - 5 );
+        bs_init( &s, &p_frag->p_buffer[4], p_frag->i_buffer - 4 );
 
         /* TODO */
+        msg_Dbg( p_dec, "found NAL_PPS" );
     }
 
-
     /* Append the block */
     block_ChainAppend( &p_sys->p_frame, p_frag );
 
     return p_pic;
 }
-