I committed a fix for the dc-offset caused by samples that don't end at 0 to redacted.
This only affects samples with a non-modulated playback-speed, in the case of a modulated playback speed, playback might be reversed and the fix might cause extra clicks (and a envelope is more likely to be used). I now remember I made it like that on purpose, but I should not have made that reasoning affect non-modulated samples; in that case playback will clearly never reverse after having reached the end. Yours, Kas.
