Hi Sven and Felipe, Thanks for the report on those architectures and the intermediate upload. I think I found a better way to handle that test that I proposed in https://gitlab.freedesktop.org/pulseaudio/pulseaudio/merge_requests/7 The idea is rather than allowing a random number of rounding issues that grows over time, potentially hiding other errors, we allow the values returned to have a difference of + or - 1, which allows the rounding issues without allowing bigger deviation. Tested on amd64, it works as expected, so I updated the patch proposed earlier via the salsa MR: https://salsa.debian.org/pulseaudio-team/pulseaudio/merge_requests/2
Let me know if you'd like to change something in it. Thanks Joseph
