log

age author description
Fri, 25 Nov 2016 13:37:02 +0000 Chris Cannam Switch the multiple-audio tests to using amplitudefollower -- percussiononsets is too sensitive to the very low-level noise that comes from variations in mp3 decoder fixed-point arithmetic function choice piper-nopiper
Fri, 25 Nov 2016 11:41:53 +0000 Chris Cannam Update svcore audio reader tests so as to check for truncated decode piper-nopiper
Thu, 24 Nov 2016 17:06:31 +0000 Chris Cannam Add MAD_BUFFER_GUARD padding at end of mp3 buffer, in order to ensure last frame is decoded successfully (otherwise the decoded audio is truncated). Another thing learned from madplay. piper-nopiper
Thu, 24 Nov 2016 13:38:45 +0000 Chris Cannam Do that only for the first frame piper-nopiper
Thu, 24 Nov 2016 13:32:04 +0000 Chris Cannam Filter out Xing/LAME info frames, rather than letting them go to the mp3 decoder as if they were audio frames. Fixes the 1152-sample zero pad at start of some decoded mp3 files (distinct from decoder delay). The logic here is based on the madplay code. piper-nopiper
Wed, 23 Nov 2016 14:51:03 +0000 Chris Cannam Tag load diagnostic piper-nopiper
Wed, 23 Nov 2016 14:34:27 +0000 Chris Cannam A bit more diagnostic info piper-nopiper