Joakim Plate elu...@ecce.se added the comment:
You are so right. Sorry for the noise.
FFmpeg issue tracker iss...@roundup.ffmpeg.org
https://roundup.ffmpeg.org/issue2563
Joakim Plate elu...@ecce.se added the comment:
You don't think something along the lines of the patch would be
appropriate too?
At the start of mpegts_push_data it can emit a packet if there is data
left over on is_start. If the given buffer then contains a full pes
header + packet
New submission from Joakim Plate elu...@ecce.se:
valgrind --tool=memcheck --leak-check=full ./ffmpeg_g -v 9 -loglevel 99
-i ~/brokenCut_issue899.ts 2 ~/test4.log
==3164== Memcheck, a memory error detector
==3164== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.
==3164== Using
Joakim Plate elu...@ecce.se added the comment:
I've been able to repro something similar when using libavformat in xbmc. I can
reproduce it on my sample file in
ffplay only if I change it so that non selected stream are not
AVDISCARD_ALL'ed. Ie
Joakim Plate elu...@ecce.se added the comment:
I think what he meant was it returns frame_size even if the supplied
amount of data was less than that. Atleast going by the check he added in
our codebase.
That can't be intended behaviour can
Joakim Plate elu...@ecce.se added the comment:
I suspect our implementation stems from way way back. Haven't looked at
if there was a better approach for it since then.
I think our implementation even came from mplayer's back in the day, so
if mplayer is now using something better, we aught
Joakim Plate elu...@ecce.se added the comment:
Done, switched to mplayer's current approach. So this ticket can probably
be closed.
FFmpeg issue tracker iss...@roundup.ffmpeg.org
https://roundup.ffmpeg.org/issue1849