In our production avalon instance we've had failed encoding jobs due to ffmpeg reporting a progress of 99% when it finishes encoding. The ffmpeg adapter currently uses the ceiling of percentage thus allowing a 1% drift but the files that failed are very short (<2 sec) so a very small variance in output time in comparison to input time can lead to a greater than 1% difference. This PR adds a class attribute on the ffmpeg adapter that allows implementers to lower the threshold if they choose.
In our production avalon instance we've had failed encoding jobs due to ffmpeg reporting a progress of 99% when it finishes encoding. The ffmpeg adapter currently uses the ceiling of percentage thus allowing a 1% drift but the files that failed are very short (<2 sec) so a very small variance in output time in comparison to input time can lead to a greater than 1% difference. This PR adds a class attribute on the ffmpeg adapter that allows implementers to lower the threshold if they choose.