[Question]: Question about m_ReadData.SetSize(0); causing premature AMF_EOF in BitStreamParserH264.cpp

[octalflow]

Hi,

I have a question regarding BitStreamParserH264.cpp.

At the very bottom of the FindSPSandPPS() function, the code is implemented as follows:

m_pStream->Seek(amf::AMF_SEEK_BEGIN, 0, NULL);
// It will fail if SPS or PPS are absent
extraDataBuilder.GetExtradata(m_Extradata);
m_ReadData.SetSize(0);

My question is: Is calling m_ReadData.SetSize(0); at this point absolutely necessary?

Consider a case where the file contains only one sequence and immediately ends with the following NALUs:
00 00 00 01 09 + 00 00 00 01 67 + 00 00 00 01 68 + 00 00 00 01 65 (EOF)

If a file ends like this, it causes an error condition. At this moment, m_bEof is already evaluated as true. Because m_ReadData.SetSize(0); is forced right after, the check inside QueryOutput() gets triggered immediately:

if ((m_bEof && m_ReadData.GetSize() == 0) || (m_maxFramesNumber && m_PacketCount >= m_maxFramesNumber))
{
    return AMF_EOF;
}

As a result, it hits the condition and unexpectedly returns AMF_EOF, preventing the actual frame data from being dispatched.

Should m_ReadData.SetSize(0); really be enforced here, or is this a potential edge-case bug in the sample parser? I would appreciate your insights.

Thank you!

[MikhailAMD]

Hi, first of all, this is purely sample code. It may don't handle corner cases. Secondly, the idea behind FindSPSandPPS() is to scan elementary stream, generate extradata (similar to what one finds in file containers) and reset the parser to the initial state. From that perspective, SetSize(0) is needed. Extradata helps to fully initialize decoder. If not provided, decoder will initialize on the first frame or later, when it finds SPS/PPS. It may cause glitches in timing. To address you issue I would set m_bEof = false at the end of FindSPSandPPS() function.

[octalflow]

thank you.