contrib/src/faad2/faad2-fix-cpe-reconstruction.patch

   1 hDecoder->element_alloced denotes whether or not we have allocated memory for
   2 usage in terms of the specified channel element. Given that it previously only
   3 had two states (1 meaning allocated, and 0 meaning not allocated), it would not
   4 allocate enough memory for parsing a CPE it if is preceeded by a SCE (and
   5 therefor crash).
   6
   7 These changes fixes the issue by making sure that we allocate additional memory
   8 if so is necessary, and the set of values for hDecoder->element_alloced[n] is
   9 now:
  10
  11  0 = nothing allocated
  12  1 = allocated enough for SCE
  13  2 = allocated enough for CPE
  14
  15 All branches that depend on hDecoder->element_alloced[n] prior to this patch
  16 only checks if the value is, or is not, zero. The added state, 2, is therefor
  17 correctly handled automatically.
  18
  19 --- faad2/libfaad/specrec.c     2017-03-12 20:16:58.305887783 +0100
  20 +++ faad2/libfaad/specrec.c     2017-03-25 03:48:19.875321643 +0100
  21 @@ -1110,13 +1110,13 @@
  22  #ifdef PROFILE
  23      int64_t count = faad_get_ts();
  24  #endif
  25 -    if (hDecoder->element_alloced[hDecoder->fr_ch_ele] == 0)
  26 +    if (hDecoder->element_alloced[hDecoder->fr_ch_ele] != 2)
  27      {
  28          retval = allocate_channel_pair(hDecoder, cpe->channel, (uint8_t)cpe->paired_channel);
  29          if (retval > 0)
  30              return retval;
  31
  32 -        hDecoder->element_alloced[hDecoder->fr_ch_ele] = 1;
  33 +        hDecoder->element_alloced[hDecoder->fr_ch_ele] = 2;
  34      }
  35
  36      /* dequantisation and scaling */
  37