[FFmpeg-devel] RTMP Audio and video packet sequence

Mirko Puliafito mirko.puliafito at gmail.com
Mon Sep 2 17:17:36 CEST 2013

Hi all,

I experienced some problem with RTMP decoder using some particolar FMS
servers like crtmpserver or wowza. RTMP packets are segmented into fixed
size chuncks. When decoding Audio+Video stream sometimes it happens that
video and audio chunks are mixed together and video segments are not
I.E. this case could happen: V - A - V - V. If the first Video packet of
the sequence reports a dimension bigger than the chunck the decoder expects
the second packet to be video but it's not. This is generating a read error
and a EIO closure. I have a patch for this (libavformat
rtmppkt.c/rtmppkt.h) tested successfully on the reported servers I need to
merge with current master (it's now on 0.10).
Can anybody give me a feedback on this problem and tell me if a patch for
this is useful?


More information about the ffmpeg-devel mailing list