| From Vaughan Johnson <vaughan@...>
| Sun, 04 Nov 2012 11:19:46 -0800
| Subject: [Audacity-quality] [Audacity-devel] Fw: SoX resampling
> On 11/4/2012 8:01 AM, Rob Sykes wrote:
> > From: Vaughan Johnson <vaughan@...>
> >> To: audacity-devel@...
> >> Okay, with commit r 12031, I think the new refactoring for libsoxr is
> >> completed:
> >> * libsoxr for all const-rate resampling,
> > Thanks Vaughan. Some of the m4 stuff seems to have fallen through the cracks somewhere along the line; this patch http://dl.dropbox.com/u/8835547/3resamplers.patch hopefully covers this, the extra words re variable/constant rate, and (for the future) being able to find libsoxr as a system lib.
> Applied and committed. Thanks!
Not to pick nits in this sterling effort, but I noticed a couple of typos
in configure in:
"AC_MSG_NOTICE([Libsamplerate not explicity requested, using libresample])
dnl if we get here, then libsample rate has been explicity specified."
i.e. "explicity" and "libsample rate".
> > With this patch on top of r12032, I can confirm that on linux, const-rate is working and the two var-rate resamplers can be selected and are working (time-track bug notwithstanding). One problem though, the quality for const-rate resampling (Hiqh quality conversion) doesn't seem to be changing according to the selected preference e.g. the execution times and the spectrograms with Low Q. and Best Q., appear to be the same.
> What time-track bug is that? The crash when using libsamplerate?
> Regarding pref for const-rate, good catch. That code was duplicated in
> the var-rate constructors, so I first did that, but just revised it to
> do the common stuff in the base class constructor, so the descendants
> don't duplicate it. In short, should be fixed in commit r 12035.