On Fri Jun 14 19:03:55 2024 +0000, Davide Beatrici wrote:
On that note, I was thinking: should we perhaps force 10 ms even if the UNIX driver reports a higher period?
We should, yeah, it should depend strictly on what Windows has as buffer length and not what backend hints. Back then it was fixed this way in xaudio / faudio. I am not sure if mmdevapi ever queried that from backends but back then it had voluntary buffer size and that was also causing problems (1d66a108b901ebfb11b134e9a6895b1d87333638).