fix DecoderConfigDescriptor parsing

Author: yilun-zhangs

code/src/iamf_dec/IAMF_decoder.c

@@ -729,9 +729,15 @@ static int iamf_codec_conf_get_sampling_rate(IAMF_CodecConf *c) {
     static int sf[] = {96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050,
                        16000, 12000, 11025, 8000,  7350,  0,     0,     0};
 
-    /* DecoderConfigDescriptor (14 bytes) + DecSpecificInfoTag (1 byte) */
-    if (c->decoder_conf_size < 16) return IAMF_ERR_BAD_ARG;
-    bs(&b, c->decoder_conf + 15, c->decoder_conf_size - 15);
+    /* DecoderConfigDescriptor */
+    bs(&b, c->decoder_conf, c->decoder_conf_size);
+    bs_get32b(&b, 8);
+    bs_getExpandableSize(&b);
+    bs_skipABytes(&b, 13);
+
+    /* DecSpecificInfoTag */
+    bs_get32b(&b, 8);
+    bs_getExpandableSize(&b);
 
     type = bs_get32b(&b, 5);
     if (type == 31) bs_get32b(&b, 6);

code/src/iamf_dec/aac/IAMF_aac_decoder.c

@@ -23,6 +23,7 @@
 #include "IAMF_debug.h"
 #include "IAMF_types.h"
 #include "aac_multistream_decoder.h"
+#include "bitstream.h"
 
 #ifdef IA_TAG
 #undef IA_TAG
@@ -58,35 +59,28 @@ typedef struct IAMF_AAC_Context {
 static int iamf_aac_init(IAMF_CodecContext *ths) {
   IAMF_AAC_Context *ctx = (IAMF_AAC_Context *)ths->priv;
   uint8_t *config = ths->cspec;
-  int len = ths->clen;
+  BitStream b;
   int ret = 0;
 
-  if (!ths->cspec || ths->clen <= 0) {
-    return IAMF_ERR_BAD_ARG;
-  }
+  bs(&b, ths->cspec, ths->clen);
+  if (bs_get32b(&b, 8) != 0x04) return IAMF_ERR_BAD_ARG;
+  bs_getExpandableSize(&b);
 
-  int idx = 1;
-  if (config[idx] != 0x40                     /* Audio ISO/IEC 14496-3 */
-      || (config[idx + 1] >> 2 & 0x3f) != 5   /* AudioStream */
-      || (config[idx + 1] >> 1 & 0x1) != 0) { /* upstream */
+  if (bs_get32b(&b, 8) != 0x40 || bs_get32b(&b, 6) != 5 || bs_get32b(&b, 1))
     return IAMF_ERR_BAD_ARG;
-  }
-  idx += 13;
-  if (config[idx] != 0x05) {
-    return IAMF_ERR_BAD_ARG;  // MP4DecSpecificDescrTag
-  }
-  ++idx;
 
-  ths->cspec = &config[idx];
-  ths->clen = len - idx;
+  bs_skipABytes(&b, 11);
+
+  if (bs_get32b(&b, 8) != 0x05) return IAMF_ERR_BAD_ARG;
+
+  ths->clen = bs_getExpandableSize(&b);
+  ths->cspec = config + bs_tell(&b);
   ia_logd("aac codec spec info size %d", ths->clen);
 
   ctx->dec = aac_multistream_decoder_open(ths->cspec, ths->clen, ths->streams,
                                           ths->coupled_streams,
                                           AUDIO_FRAME_PLANE, &ret);
-  if (!ctx->dec) {
-    return IAMF_ERR_INVALID_STATE;
-  }
+  if (!ctx->dec) return IAMF_ERR_INVALID_STATE;
 
   ctx->out = (short *)malloc(sizeof(short) * MAX_AAC_FRAME_SIZE *
                              (ths->streams + ths->coupled_streams));

code/src/iamf_dec/bitstream.c

@@ -141,6 +141,23 @@ uint64_t bs_getAleb128(BitStream *b) {
   return ret;
 }
 
+/// @brief Read descriptor size of ISO/IEC 14496-1.
+uint32_t bs_getExpandableSize(BitStream *b) {
+  uint32_t ret = 0;
+  uint8_t byte;
+
+  for (uint32_t i = 0; i < 4; i++) {
+    byte = b->data[b->b8sp + i];
+    ret = (ret << 7) | (byte & 0x7f);
+    if (!(byte & 0x80)) {
+      b->b8sp += (i + 1);
+      break;
+    }
+  }
+
+  return ret;
+}
+
 int32_t bs_read(BitStream *b, uint8_t *data, int n) {
   bs_align(b);
   if (data) memcpy(data, &b->data[b->b8sp], n);

code/src/iamf_dec/bitstream.h

@@ -42,6 +42,7 @@ uint32_t bs_getA8b(BitStream *b);
 uint32_t bs_getA16b(BitStream *b);
 uint32_t bs_getA32b(BitStream *b);
 uint64_t bs_getAleb128(BitStream *b);
+uint32_t bs_getExpandableSize(BitStream *b);
 int32_t bs_read(BitStream *b, uint8_t *data, int n);
 int32_t bs_readString(BitStream *b, char *data, int n);
 uint32_t bs_tell(BitStream *b);

code/src/iamf_dec/vlogging_tool_sr.c

@@ -753,7 +753,7 @@ static void write_mix_presentation_log(uint64_t idx, void* obu, char* log) {
 
       // loudness
       log += write_yaml_form(log, 3, "loudness:");
-      log += write_yaml_form(log, 4, "info_type: %u",
+      log += write_yaml_form(log, 4, "info_type_bit_masks: [%u]",
                              submix->loudness[j].info_type);
       log += write_yaml_form(log, 4, "integrated_loudness: %d",
                              submix->loudness[j].integrated_loudness);