Closed sdroege closed 3 months ago
@sdroege Thanks and sorry for late reply. Don't think i have any strong opinions how to fix it, in the software where this was triggered we're looking for NaN and Inf samples before using ebur128 as a workaround atm.
This avoids returning non-sensical results, or in case of history-based loudness range calculations a panic.
All filter processing already handles NaN gracefully.
Fixes https://github.com/sdroege/ebur128/issues/43