Fix RBSP issue, where 0x03 should be removed. v5.0.178 v6.0.75

[feixintianxia]

ISO_IEC_14496-10-AVC-2012.pdf, page 65
7.4.1.1 Encapsulation of an SODB within an RBSP (informative)

... 00 00 03 xx, the 03 byte should be drop where xx represents any 2 bit pattern: 00, 01, 10, or 11.

---

Co-authored-by: john hondaxiao@tencent.com
Co-authored-by: chundonglinlin chundonglinlin@163.com

[xiaozhihong]

Thank you for the PR. This PR is an enhancement PR.
I'm not sure if we need to handle preventing competition codes in this way, because only 0x00 0x00 0x00, 0x00 0x00 0x01, 0x00, 0x00, 0x02, 0x00, 0x00, 0x03 will add competition codes. So your condition if (tmp > 3) { may always be true.

TRANS_BY_GPT3

[feixintianxia]

Thank you for your reply. I referred to the following code.

https://github.com/mirror/x264/blob/master/common/bitstream.c

static uint8_t *nal_escape_c( uint8_t *dst, uint8_t *src, uint8_t *end )
{
    if( src < end ) *dst++ = *src++;
    if( src < end ) *dst++ = *src++;
    while( src < end )
    {
        if( src[0] <= 0x03 && !dst[-2] && !dst[-1] )      // this 
            *dst++ = 0x03;
        *dst++ = *src++;
    }
    return dst;
}

[winlinvip]

There indeed is this logic, that is, if the original data in the video is 00 00 01, this would conflict with the prefix of annexb, so it should be specially encoded.

TRANS_BY_GPT4

[winlinvip]

Please add the corresponding utest cases. For the various scenarios mentioned in the specifications, test cases need to be added for verification and testing.

TRANS_BY_GPT4

[xiaozhihong]

Thank you for the PR. This PR is an enhancement PR. I'm not sure if we need to handle preventing competition codes in this way, because only 0x00 0x00 0x00, 0x00 0x00 0x01, 0x00, 0x00, 0x02, 0x00, 0x00, 0x03 will add competition codes. So your condition if (tmp > 3) { may always be true.

TRANS_BY_GPT3

@feixintianxia I revisited the code logic, it seems I misunderstood it before, this modification should be fine. Could you please help to add some unit tests? Thanks.

TRANS_BY_GPT4

[winlinvip]

Nice work! Thank you all @feixintianxia @xiaozhihong @chundonglinlin ! 👍

[winlinvip]

I added some utests and found that the boundary conditions were not handled correctly, the following case will fail:

VOID TEST(KernelCodecTest, VideoFormatRbspEdge)
{
    if (true) {
        vector<uint8_t> nalu = {0x00, 0x00, 0x03};
        vector<uint8_t> expect = {0x00, 0x00, 0x03};

        vector<uint8_t> rbsp(nalu.size());
        SrsBuffer b((char*)nalu.data(), nalu.size());
        int nb_rbsp = srs_rbsp_remove_emulation_bytes(&b, rbsp);

        ASSERT_EQ(nb_rbsp, (int)expect.size());
        EXPECT_TRUE(srs_bytes_equals(rbsp.data(), expect.data(), nb_rbsp));
    }

Fixed by 5e6784c

TRANS_BY_GPT4