Unwanted midi changes. about bschaffl HOT 9 CLOSED

Thanks for reporting. After a while of playing around I observed a similar (or the same?) bug and I analyzed the dump. I found "old" (from a stopped previous run) midi messages sticking in front of the midi event queue and these messages block all new messages. I will look how to fix it.

from bschaffl.

I can reproduce this problem with Ardour 5.12. It looks like the same problem as described by myself in issue #1 . If the NOTE_OFF for the first note is scheduled shortly after the NOTE_ON for the second note and both notes have the same pitch you will hear (almost) nothing (= "dropped notes"). This shouldn't happen if the notes have different pitches.

I will do a dump to confirm it. If this overlap comes from the input, a "smart quantization" algorithm (see README.md#TODO's) should fix it. If it only appears on the output side (I'm afraid) then it's a bug.

from bschaffl.

It was a bug. Fixed in f5cf026 . Also added MIDI scheduler dump to stderr.

from bschaffl.

That's better; with the default settings, it bypasses.
When I change the swing slider I get a swing pattern, but also periods of silence, and sometimes I need to start and stop Ardour to get sound back.

from bschaffl.

That's better; with the default settings, it bypasses.

What do you mean with "bypasses"? You should hear the origial 1/16 note pattern. If you don't hear anything or if you hear something else then it's a bug.

When I change the swing slider I get a swing pattern, but also periods of silence...

I'll try to reproduce it.

from bschaffl.

What do you mean with "bypasses"?

Sorry, that was unclear. Yes, it passes trough the midi unchanged, as it should.

from bschaffl.

Quick fix in 94ad289 . This fix will kill all "old" scheduled midi messages from the queue. I hope this will fix the problem you described.

For the future I have look for a smarter solution. Otherwise other ideas, like MIDI filtering and smart quantization (fitting NOTE_OFF to the end of a step and NOTE_ON to the start of a step) will not work.

from bschaffl.

Yes, it works!
This thing is even more fun to play with then I imagined when I asked for it.
Thank you so much!

It indeed look like you need to look for a smarter solution, as there are some timing bugs remaining.
For example the first bar of a song plays unchanged for me.

from bschaffl.

You reported the "first bar bug" to be solved in #3 . Thanks.

Smart quantization is now introduced in c2eec87 . This means that NOTE_ON signals can now be mapped to (the amp of) the just starting step whereas NOTE_OFF signal can be mapped to the ending step by definition of a range. The edge problem is solved.

Still TODO: classic quantization (re-positioning of not exactly placed MIDI signals.

from bschaffl.