search

dwbuiten / FFmpeg

mirror of git://git.videolan.org/ffmpeg.git
http://ffmpeg.org
Other
0 stars 1 forks source link

concat generates corrupted h264 #20

Closed michaelni closed 8 years ago

michaelni commented 8 years ago

./ffmpeg -f concat -i tickets/3108/concatfile.txt -codec copy out.avi

playing out.avi shows errors: [h264 @ 0x7f33480034a0] AVC: nal size 100663272KB sq= 0B f=0/0
[h264 @ 0x7f33480034a0] missing picture in access unit with size 14576 ... [NULL @ 0x7f33480034a0] AVC: nal size 1081312514B sq= 0B f=0/0
[NULL @ 0x7f33480034a0] missing picture in access unit with size 11730

file should be in https://trac.ffmpeg.org/raw-attachment/ticket/3108/examplefiles.zip

dwbuiten commented 8 years ago

As far as I can tell, this isn't a concatdec problem?

If I do: ffmpeg -i tickets/3108/concatfile.txt -codec copy out.h264, both old and new output is correct. But if I mux to mov, one sample has a 1-byte-different length. Further more, the decoded output of both is the same (md5). I'm not sure what this is.

dwbuiten commented 8 years ago

Gotta admit im pretty stumped on this one.

dwbuiten commented 8 years ago 
[16:33] <@Daemon404> oh... hmmm
[16:33] <@Daemon404> it is perhaps the way teh bsd inside concatdec.c is acting with the muxing
[16:33] <@Daemon404> because when you output to a format that uses annexb (.h264, .ts) it is fine
[16:33] <@Daemon404> but when it is a mp4-like output, it fucks up
[16:34] <@Daemon404> when it does mp4style->annexb->backtomp4
[16:34] <@Daemon404> likely because concatdec is doing bistream filtering in the demuxer itselg
[16:36] <@Daemon404> though im not sure where/why it messes up
[16:38] <@Daemon404> actually where the hell is concatdec creating streams
[16:41]  * Daemon404 donest understand this crap at all
dwbuiten commented 8 years ago

Fixed.