Re: [Denemo-devel] Showstopper bug?

[Richard Shann]

Dominic,

I missed seeing this question from you...

<das> rshann: the comment mentions a "pause mechanism", how does that
work?

The broad outline is that each time the portaudio backend makes its
callback for another buffer full of samples there is a check, and (at
least roughly) if the music is paused it is given a buffer filled with
0.0 values.

I think other backends may need to use a different mechanism (because
they don't initiate the request for samples?), and if that were
developed then, for uniformity, portaudio backend could possibly follow
suit and use the same. But I didn't really understand what these
backends would be doing, they being not so easy to get working, e.g.
setting up Alsa connections, or starting up jack etc.

Richard

On Tue, 2012-05-15 at 09:16 +0100, Richard Shann wrote:
> Dominic,
> 
> Further to this I just tried the git blame command and got this:
> 
> git blame -C -L 68,68 src/portmidibackend.c
> a03ecf7d (rshann 2012-01-09 18:16:18 +0000 68) return;
> 
> Looking up that commit I find this comment:
> 
> Implement the pause feature for playback. Note playback via midi backend
> is disabled as it continues to update the playback time regardless of
> pause.
> 
> So it seems that if you re-enable the portmidi backend it will have the
> bad effect that during a pause the playback time will continue to get
> updated. The playhead will continue to move along the notes, although
> the portaudio output is paused.
>  So that will be why I disconnected the code. To re-connect, it will
> need to use the same sort of code as the portaudio backend to handle
> time while pause is in effect.
> 
> Richard
> 
> 
> On Tue, 2012-05-15 at 09:08 +0100, Richard Shann wrote:
> > Dominic,
> > when you said on irc #denemo
> > > what problem is the return statement in process_midi() (in 
> > > portmidibackend.c) supposed to solve? what would happen if i remove it?
> > 
> > I answered thinking you were referring to the first return statement in 
> > process_midi().
> > 
> > But, now I look further I see the your point - I have put in another
> > return to ignore the midi output backend. As the comment says, this is
> > because I introduced the pause feature involving handling the time
> > differently and I expected the portmidi output code would need altering.
> > I haven't got any MIDI output device to test such code, so I don't know
> > what it will do without that return statement - perhaps I should simply
> > have let it happen and wait for feedback?
> > If you can test it I would be most interested to know what happens, it
> > can't possibly work complete with Pause, but it could be that it would
> > still play. (In a similar vein, you will have noticed that I
> > re-connected the non-threaded audio input code for inputting notes from
> > a microphone, and it just works; it is safest to disconnect stuff that
> > *may* cause problems while developing, but it is always worth trying to
> > reconnect it afterwards to see if it does just work anyway).
> > 
> > Richard
> > 
>