(Version of) patch from Richard Watts: allow forcing of MPEG2 or MPEG4 ADTS in

esmerge.

--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4019

adts.c

@@ -45,12 +45,15 @@
  *
  * - `file` is the file descriptor of the ADTS file to read from
  * - `frame` is the ADTS frame that is read
+ * - `flags` indicates if we are forcing the recognition of "emphasis"
+ *   fields, etc.
  *
  * Returns 0 if all goes well, EOF if end-of-file is read, and 1 if something
  * goes wrong.
  */
 extern int read_next_adts_frame(int            file,
-                                audio_frame_p *frame)
+                                audio_frame_p *frame,
+                                unsigned int   flags)
 {
 #define JUST_ENOUGH 6 // just enough to hold the bits of the headers we want
 
@@ -59,6 +62,7 @@ extern int read_next_adts_frame(int            file,
   byte   header[JUST_ENOUGH];
   byte  *data = NULL;
   int    frame_length;
+  int    has_emphasis = 0;
 
   offset_t  posn = tell_file(file);
 #if DEBUG
@@ -102,16 +106,21 @@ extern int read_next_adts_frame(int            file,
     printf("   layer is %d, not 0 (in frame at " OFFSET_T_FORMAT ")\n",
            layer,posn);
 
-  if (id == 1)
+  // Experience appears to show that emphasis doesn't exist in MPEG-2 AVC.
+  // But it does exist in (ID=1) MPEG-4 streams.
+  // 
+  // .. or if forced.
+
+  has_emphasis = (flags & ADTS_FLAG_NO_EMPHASIS) ? 0 :
+    ((flags & ADTS_FLAG_FORCE_EMPHASIS) || !id);
+
+  if (!has_emphasis)
   {
-    // We assume the Emphasis field is not present
-    // (experience appears to show that it is not used for MPEG-2 AVC)
     frame_length = ((header[3] & 0x03) << 11) | (header[4] << 3) |
       ((unsigned)(header[5] & 0xE0) >> 5);
   }
   else
   {
-    // We assume the Emphasis field *is* present
     frame_length = (header[4] << 5) | ((unsigned)(header[5] & 0xF8) >> 3);
   }
 #if DEBUG

adts_defns.h

@@ -32,6 +32,17 @@
 #include "audio_defns.h"
 // AAC ADTS provides audio in frames of constant time
 
+// Flags for ``read_next_adts_frame``
+//
+// Specify this flag to indicate that there is no emphasis field in the ADTS
+// header. Generally, MPEG-2 ADTS audio (ID=0) has no emphasis field and
+// MPEG-4 (ID=1) has emphasis, but some H.264/AAC streams have MPEG-4 ADTS
+// with no emphasis and in those cases, you'll need this flag.
+#define ADTS_FLAG_NO_EMPHASIS    (1<<0)
+// Specify this flag to indicate that there is always an emphasis field, even
+// if the ID says there isn't one - included for symmetry with NO_EMPHASIS.
+#define ADTS_FLAG_FORCE_EMPHASIS (1<<1)
+
 #endif // _adts_defns
 
 // Local Variables:

adts_fns.h

@@ -39,12 +39,15 @@
  *
  * - `file` is the file descriptor of the ADTS file to read from
  * - `frame` is the ADTS frame that is read
+ * - `flags` indicates if we are forcing the recognition of "emphasis"
+ *   fields, etc.
  *
  * Returns 0 if all goes well, EOF if end-of-file is read, and 1 if something
  * goes wrong.
  */
 extern int read_next_adts_frame(int            file,
-                                audio_frame_p *frame);
+                                audio_frame_p *frame,
+                                unsigned int   flags);
 
 
 #endif // _adts_fns

audio.c

@@ -98,8 +98,12 @@ extern int read_next_audio_frame(int            file,
 {
   switch (audio_type)
   {
+  case AUDIO_ADTS_MPEG2:
+    return read_next_adts_frame(file,frame,ADTS_FLAG_NO_EMPHASIS);
+  case AUDIO_ADTS_MPEG4:
+    return read_next_adts_frame(file,frame,ADTS_FLAG_FORCE_EMPHASIS);
   case AUDIO_ADTS:
-    return read_next_adts_frame(file,frame);
+    return read_next_adts_frame(file,frame,0);
   case AUDIO_L2:
     return read_next_l2audio_frame(file,frame);
   default:

audio_defns.h

@@ -46,6 +46,9 @@ typedef struct audio_frame *audio_frame_p;
 #define AUDIO_ADTS      ADTS_AUDIO_STREAM_TYPE
 #define AUDIO_L2        MPEG2_AUDIO_STREAM_TYPE
 
+#define AUDIO_ADTS_MPEG2 0x100
+#define AUDIO_ADTS_MPEG4 0x101
+
 #define AUDIO_STR(x)    ((x)==AUDIO_UNKNOWN?"unknown":  \
                          (x)==AUDIO_ADTS   ?"ADTS":     \
                          (x)==AUDIO_L2     ?"MPEG2":    \

esmerge.c

@@ -178,6 +178,8 @@ static int merge_with_avs(avs_context_p  video_context,
   switch (audio_type)
   {
   case AUDIO_ADTS:
+  case AUDIO_ADTS_MPEG2:
+  case AUDIO_ADTS_MPEG4:
     prog_type[1] = ADTS_AUDIO_STREAM_TYPE;
     break;
   case AUDIO_L2:
@@ -392,6 +394,8 @@ static int merge_with_h264(access_unit_context_p  video_context,
   switch (audio_type)
   {
   case AUDIO_ADTS:
+  case AUDIO_ADTS_MPEG2:
+  case AUDIO_ADTS_MPEG4:
     prog_type[1] = ADTS_AUDIO_STREAM_TYPE;
     break;
   case AUDIO_L2:
@@ -570,6 +574,8 @@ static void print_usage()
     "\n"
     "  -adts             The audio stream is ADTS (the default)\n"
     "  -l2               The audio stream is MPEG layer 2 audio\n"
+    "  -mp2adts          The audio stream is MPEG-2 style ADTS regardless of ID bit\n"
+    "  -mp4adts          The audio stream is MPEG-4 style ADTS regardless of ID bit\n"
     "\n"
     "Limitations\n"
     "===========\n"
@@ -671,6 +677,14 @@ int main(int argc, char **argv)
       {
         video_type = VIDEO_H264;
       }
+      else if (!strcmp("-mp2adts", argv[ii]))
+      {
+        audio_type = AUDIO_ADTS_MPEG2;
+      }
+      else if (!strcmp("-mp4adts", argv[ii]))
+      {
+        audio_type = AUDIO_ADTS_MPEG4;
+      }
       else if (!strcmp("-avs",argv[ii]))
       {
         video_type = VIDEO_AVS;