[FFmpeg-devel] [PATCH v2 3/7] avformat/matroskadec: export Dynamic HDR10+ packet side data

[James Almer]

On 3/29/2023 9:43 PM, Andreas Rheinhardt wrote:
>> @@ -3615,12 +3635,58 @@ static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
>>   }
>>   
>>   static int matroska_parse_block_additional(MatroskaDemuxContext *matroska,
>> -                                           AVPacket *pkt,
>> +                                           MatroskaTrack *track, AVPacket *pkt,
>>                                              const uint8_t *data, int size, uint64_t id)
>>   {
>> -    uint8_t *side_data = av_packet_new_side_data(pkt,
>> -                                                 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
>> -                                                 size + 8);
>> +    uint8_t *side_data;
>> +    int res;
>> +
>> +    switch (id) {
>> +    case 4: {
>> +        GetByteContext bc;
>> +        int country_code, provider_code;
>> +        int provider_oriented_code, application_identifier;
>> +        size_t hdrplus_size;
>> +        AVDynamicHDRPlus *hdrplus;
>> +
>> +        if (!track->blockaddid_itu_t_t35)
>> +            break; //ignore
>> +
>> +        bytestream2_init(&bc, data, size);
>> +
>> +        /* ITU-T T.35 metadata */
>> +        country_code  = bytestream2_get_byte(&bc);
>> +        provider_code = bytestream2_get_be16(&bc);
> This will add implicit overread checks; but you should check size
> explicitly instead.
> (The current behaviour of the bytestream2 API is to return 0 upon
> overreads; yet one should not rely on this and it could e.g. be changed
> so that bytestream2_get_byte returns 0xB5 and bytestream2_get_be16 0x3C
> upon overread.)

The only reason I'm using bytestream2 here was to not do a size check, 
as Anton suggested.
Added one in any case, as well as using the u variants of 
bytestream2_get_* since it will not overread now.