Re: [Denemo-devel] make dist failing

[Richard Shann]

On Wed, 2014-01-29 at 15:59 -0400, Paul Brossier wrote:
> On 01/29/2014 01:12 PM, Richard Shann wrote:
> > On Wed, 2014-01-29 at 12:41 -0400, Paul Brossier wrote:
> >> On 01/29/2014 11:17 AM, Richard Shann wrote:
> >>> On Wed, 2014-01-29 at 10:33 -0400, Paul Brossier wrote:
> >>>> Hi all,
> >> [...]
[...]
> >>
> >> the performance of the onset detection should change only little.
> >> one important change is that the onset timestamps are not quantized
> >> to hop_size anymore, but given in samples.
> >
> > The onset detection code in src/sourceaudio.c is just cloned from
> > your onset example (and, likewise, but several years earlier the
> > pitch detection in src/pitchrecog.c was cloned from your example
> > code).
> 
> i updated src/sourceaudio.c using the new aubio_onset_t. it simplifies
> the code quite a bit. i will do the same for pitchrecog.c now.
I look forward to seeing this - we can make a new release of Denemo once
that is sorted.

> 
> i deleted the #ifndef G_OS_WIN32/#endif. if deleting any aubio object
> causes a crash on windows, aubio should be fixed.

I suspected the problem lay elsewhere than in aubio (e.g. in how I was
calling the aubio routines), but didn't have time to delve deeper. It
will be good to see if the problem has gone in the new version.

> 
> another thing: is multichannel processing needed, or should the input be
> downmixed instead (I would vote for the second option)?

(I'm not exactly sure of the terminology here), the input is assumed to
be stereo just because that is what is most common, but no use is made
of that fact, and I guess the onset detection does not try to exploit
this either, so accepting mono input and adding together stereo would be
better. It's really little more than a first draft to find out if it is
useful to do this sort of thing (I have never heard of another program
allowing you to synchronize a notated score with a live performance for
transcription onto the notated score, so I don't know if it is really
useful or not).

> 
> also, I see that sndfile is used for reading the file. we could use
> aubio_source_t instead, so that compressed files (mp3

I understand that mp3 could invite the lawyers around to claim royalty
payments

>  et all) could also
> be open by denemo (using libav on unix/windows systems, and natively on
> mac osx).
that sounds good - I don't recall the reason for the choice of
libsndfile, it may already have been in use or it may just have been
ignorance :)

Thanks for the help

Richard