Closed lars18th closed 4 years ago
Interesting obseravtion. I have never tested the Windows version, but I suspect it may have to do with threading. The frame types are logged from the main application, while the debug values are printed by the ffmpeg libraries.
@plebreton Do you have an idea why this would happen?
Hi,
Yes, I will have a look. Not sure if I can do it tonight though (it is the end of the day in my time zone).
I will see what I can do. Tomorrow at worst.
Just a small update on this. I have been able to reproduce the problem, but I haven't solved it yet
Hi, any news about this trouble?
Hi @slhck ,
Please check my solution #19 .
Hi,
After testing the CI build of the Windows version, the problem continues in Windows:
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 101010101010101010101010101010101010101010101010101010<< frame_type: P; pkt_size: 10780 >>
10101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
[mpeg2video @ 0000000000588080] 10101010101010101010101010101010101010101010101010101010101010101010
But, with the new format the parsing can be overcomed in the python parser.
I'll try to provide a PR that "replaces" the << frame_type: ... >>\n
with the next line.
Using this way the parsing will provide the same results.
Regards.
Please wait a day before providing a PR, as I am about to restructure the tool into a proper Python package. I will be done with this tomorrow! Then your changes will be based on the latest version.
Updated PR #24 to fix it.
Hi @slhck ,
I confirm that in Windows using the last ci pre-compiled binary this problem has gone. :tada: So I close the issue.
However, please @slhck check to increase the number of threads: https://github.com/slhck/ffmpeg-debug-qp/blob/master/ffmpeg_debug_qp.c#L225 The tool is now thread-safe and the number can be increased without disturbing results. I recommend to use '3' threads.
Regards.
Hi,
I'm using the Windows precompiled version. Almost in that version when a new message "frame_type:" appears it is printed in the middle of the quantization values.
Example:
It appears with MPEG-2 and H.264. I hope you can fix it. Regards.