[FFmpeg-devel] [RFC] avcodec: Add native DCA decoder based on libdcadec.

[Andreas Cadhalpun]

On 06.01.2016 23:17, Andreas Cadhalpun wrote:
> On 06.01.2016 18:32, foo86 wrote:
>> Otherwise testing coverage will be decreased somewhat. The easiest way to do
>> this is to modify ff_dca2_check_crc() to always return 0.
> 
> I tried this (comment out everything in ff_dca2_check_crc except 'return 0')
> and there doesn't seem to be much difference.

The reason why this didn't make any difference is that this function wasn't
executed at all for my fuzzed samples. :-/
I fixed that and coverage is now much better. Hence I also have a few more
(minor) comments:

On 03.01.2016 18:49, foo86 wrote:
> +            for (i = 0; i < spkr_remap_nsets; i++) {
> +                // Number of channels to be decoded for speaker remapping
> +                int nch_for_remaps = get_bits(&s->gb, 5) + 1;
> +
> +                for (j = 0; j < nspeakers[i]; j++) {
> +                    // Decoded channels to output speaker mapping mask
> +                    int remap_ch_mask = get_bits(&s->gb, nch_for_remaps);

Here nch_for_remaps can be up to 32, so this has to use get_bits_long, as
get_bits only supports reading 1-25 bits.

> +        for (i = 0; i < s->nmixoutconfigs; i++) {
> +            for (j = 0; j < nchannels_dmix; j++) {
> +                // Mix output mask
> +                int mix_map_mask = get_bits(&s->gb, s->nmixoutchs[i]);

Here s->nmixoutchs[i] can be zero. If that should not happen, there needs
to be an error check and otherwise it should use get_bitsz, because
get_bits doesn't support reading 0 bits.

Anyway, I still think the code is pretty robust. :-)

I'd be glad to increase fuzz-testing coverage further, but I'm lacking
input examples. It would be great if you could share some (tiny) samples
triggering the HEADER_XCH/HEADER_XXCH cases and/or *_down_mix functions.

Best regards,
Andreas