[Ardour-Dev] big commit in 2.0-ongoing SVN needs testing

[Robin Gareus]

Paul Davis wrote:
> I have just committed a huge change to the deep internals of 2.X which
> is intended to fix at least one known bug. The change is deep,
> pervasive and potentially crash-inducing. Along the way, I may have
> altered parts of how latency compensation operates.
> 
> I'd really appreciate as many people as possible doing a build and
> testing stuff out in whatever style they normally work. Latency
> compensation in particular needs careful checking, and my current
> cabling set up here in the studio makes this tricky for me to do in
> all the required ways.
> 
> Please don't report crashes without backtraces, which means that scons
> DEBUG=1 is a must.
> 
> Thanks for your help. This change should fix at least the "click at
> loop point" error, but there is potential for other subtle yet
> incorrect/dangerous behaviour to also be corrected by the change.
> 
> rev 4979 or later contains the new magic.

The good news is: I have not got it to crash!

I've made a few tests concerning latency compensation. There's still
some bugs in there:

new ardour session.
 - add two tracks with the same audio
 - add a LADSPA Inverter Plugin at Track 1
 - add eg. the DJ-EQ (with all controls at zero) on Track 2

Playing this session should result in silence. but it does not.

 - stop playback
 - bypass the DJ-EQ
 - start playback

Now the tracks cancel each other out.

 - keep playing and enable the DJ-EQ.

It remains in sync and quiet!

Another issue is compensating the port latency of the master-output port
relative to jack-transport (tested with jplay2 sync to jack-transport)
The output of ardour is late.
However I don't see a clean solution: Compensating this would mean that
ardour will never play the first N samples.
fi. at jack-transport position 00:00:00.0 ardour would need to output
audio at 00:00:00.1 - so that frame 00:00:00.1 hits the speakers at the
same time as jack-transport reaches this timestamp.

robin