From: Vaughan Johnson <vjohnson@co...> - 2004-08-07 03:30:22
> Revision : 220.127.116.11.4.4
> Date : 2004/8/6 5:59:42
> Author : 'mbrubeck'
> State : 'Exp'
> Lines : +1 -6
> Description :
> Revert last change.(Incorrect backport of James Crook's fix from HEAD,
> which was already applied to this branch in the previous revision.)
Sorry, Matt. I was looking for "// JKC", and the code was different
enough that I didn't think the setTempo(1.0f) was new -- if that was the
fix. Anyway, I should have checked with you.
I started this because of the report from Ross McMillan, "RE:
[Audacity-users] Change Pitch DOES change the tempo". I first sent him a
build from AUDACITY_1_2, recalling that you'd mentioned porting James's
fix. I thought that the fix, although made for a different problem he
was having, was what had fixed this problem in my AudacityChangePitchVST
plugin, & in HEAD. But Ross wrote back that it didn't work in my
AUDACITY_1_2 build (& I later learned, nor did my incorrect backport).
Now I confirm that the problem occurs with AUDACITY_1_2, and with HEAD
-- but not with AudacityChangePitchVST.
So, I'll try to debug it (tomorrow), but if anybody has any ideas,
please let me know. I've seen it both over-return and under-return. In
my testing, unlike what Ross reported, it seems dependent on the
selection size. It's returning on the order of thousands of samples
different than the input, per processing call, and does not converge to
zero-sum. I hope the way I did it in AudacityChangePitchVST is
repeatable and portable. :-)
From: Matt Brubeck <mbrubeck@cs...> - 2004-08-07 16:16:31
Vaughan Johnson wrote:
> > Revision : 18.104.22.168.4.4
> > Date : 2004/8/6 5:59:42
> > Author : 'mbrubeck'
> > Revert last change.(Incorrect backport of James Crook's fix from HEAD,
> > which was already applied to this branch in the previous revision.)
> Sorry, Matt. I was looking for "// JKC", and the code was different
> enough that I didn't think the setTempo(1.0f) was new -- if that was
> the fix. Anyway, I should have checked with you.
No problem -- the diffs were confusing, partly because I didn't copy
JKC's comment, and also because the SoundTouch code on HEAD had
line-ending issues that made the CVS deltas hard to read. [For the
record, the setTempo(tempo) from James' fix was actually in a different
function than the setTempo(1.0f) line.]