How to parse Server's "frame meta"?

[nnnpa31]

Hi, I'm trying to write a go implementation of the Scrcpy client. When I insert the obtained h264 stream data into the libavcodec's decoder, I sometimes get this error

[h264 @ 0x2e56a40] No start code is found.
[h264 @ 0x2e56a40] Error splitting the input into NAL units.
4 Invalid data found when processing input

I think this is because I am not splitting the NALU data correctly and wrapping it into the AVPacket.
To simplify the code, as you did in #646, I want to parse the frame meta data to read the complete NALU packet one by one.
I didn't understand the Server java code and I'm using version 1.25.
Can you give me some clues?

[rom1v]

Here are relevant pointers:

scrcpy/app/src/demuxer.c

Lines 77 to 96 in cb20bcb

	// The video stream contains raw packets, without time information. When we
	// record, we retrieve the timestamps separately, from a "meta" header
	// added by the server before each raw packet.
	//
	// The "meta" header length is 12 bytes:
	// [. . . . . . . .|. . . .]. . . . . . . . . . . . . . . ...
	// <-------------> <-----> <-----------------------------...
	// PTS packet raw packet
	// size
	//
	// It is followed by <packet_size> bytes containing the packet/frame.
	//
	// The most significant bits of the PTS are used for packet flags:
	//
	// byte 7 byte 6 byte 5 byte 4 byte 3 byte 2 byte 1 byte 0
	// CK...... ........ ........ ........ ........ ........ ........ ........
	// ^^<------------------------------------------------------------------->
	// || PTS
	// | `- key frame
	// `-- config packet

scrcpy/server/src/main/java/com/genymobile/scrcpy/Streamer.java

Lines 83 to 100 in cb20bcb

	private void writeFrameMeta(FileDescriptor fd, int packetSize, long pts, boolean config, boolean keyFrame) throws IOException {
	headerBuffer.clear();
	long ptsAndFlags;
	if (config) {
	ptsAndFlags = PACKET_FLAG_CONFIG; // non-media data packet
	} else {
	ptsAndFlags = pts;
	if (keyFrame) {
	ptsAndFlags |= PACKET_FLAG_KEY_FRAME;
	}
	}
	headerBuffer.putLong(ptsAndFlags);
	headerBuffer.putInt(packetSize);
	headerBuffer.flip();
	IO.writeFully(fd, headerBuffer);
	}

[nnnpa31]

Thank you very much! I think this is enough to solve my needs:

 // The video stream contains raw packets, without time information. When we 
 // record, we retrieve the timestamps separately, from a "meta" header 
 // added by the server before each raw packet. 
 // 
 // The "meta" header length is 12 bytes: 
 // [. . . . . . . .|. . . .]. . . . . . . . . . . . . . . ... 
 //  <-------------> <-----> <-----------------------------... 
 //        PTS        packet        raw packet 
 //                    size 
 // 
 // It is followed by <packet_size> bytes containing the packet/frame. 
 // 
 // The most significant bits of the PTS are used for packet flags: 
 // 
 //  byte 7   byte 6   byte 5   byte 4   byte 3   byte 2   byte 1   byte 0 
 // CK...... ........ ........ ........ ........ ........ ........ ........ 
 // ^^<-------------------------------------------------------------------> 
 // ||                                PTS 
 // | `- key frame 
 //  `-- config packet 

I will close this issue.