CODEWITHSUNDEEP
c++qt4udph.264libav
Colin Ray
Colin Ray

Reputation: 175

Decoding UDP Streams with LibAV: Insufficient Buffer Size

I'm attempting to using the libav libraries with Qt to decode h.264 UDP video streams, but when I use avformat_open_file, I get the following error message: 

[udp @ 0x102b5bee0] Part of datagram lost due to insufficient buffer size

The message appears about 10 times and then the attempt fails. I'm trying to decode 4 different UDP streams simultaneously, each stream with a frame rate of 25fps. Wireshark indicates that the packets are between 3000 and 10000 bytes.

Here is my call to avformat_open_input:

avformat_open_input(&formatContext, udpUrl.toStdString().c_str(), NULL, NULL)

formatContext is NULL at the time that this call is made, and udpUrl is in the format "udp://ipaddress:port".

If someone could shed some light on this issue for me it would be greatly appreciated!

Upvotes: 2

Views: 4407

Answers (3)

lu_zero
lu_zero

Reputation: 998

Make sure you set the buffer_size and pkt_size options to be big enough.

udp, as any other protocols (e.g. tcp, sctp) can deliver generic formats (so mpegts, nut, mkv) or you can use rtp and rtsp and the respectively coupled formats.

Upvotes: 0

Colin Ray
Colin Ray

Reputation: 175

After digging around a bit more it looks like FFMPEG expects you to first encode the stream to MPEGTS when streaming raw H.264 over UDP. Sure enough, encoding the output stream to MPEGTS did the trick.

Upvotes: 1

Chad
Chad

Reputation: 19052

That error happens when you try to read a UDP datagram into a buffer that is too small. Since UDP does not guarantee delivery, this packet simply gets truncated (or dropped) depending on the implementation.

From a quick look at the documentation, it looks like you can specify the datagram size as an option on your URL (See section 6.16 UDP).

Try changing your URL to something like: udp://ipaddress:port?buffer_size=10240

Upvotes: 1

Related Questions