http://bugs.winehq.org/show_bug.cgi?id=14717
--- Comment #252 from Alexander E. Patrakov patrakov@gmail.com 2011-12-24 04:01:48 CST --- I have looked at the patch, but not as thoroughly as I want. Some of the concerns about maintaining the buffer positions twice are addressed, but I don't yet understand why it is impossible that converted > count (i.e. that there is no buffer overflow) and why the calculation of the temporary buffer size is consistent with what speex thinks.
And I have to test the integer <-> float conversion manually for every possible case. In particular, I want to ensure that samples are passed through, unchanged, if there is only one stream with no volume applied and no resampling.