seeing "skip" messages, Tidal stops playback

[kindohm]

Describe the bug
Sometimes Tidal produces "Skip" messages in the ghci output. Correlated: when this happens, Tidal can't really keep up and seems to have trouble correctly sequencing events until it is restarted.

To Reproduce
I was doing something like this:

let times = "<3 5 7 10>"
let denom = ("4" * (discretise 1 $ range 0.5 1 $ rand) * (range 0.5 1 $ slow 3 sine))

d1 $ plyWith times (# gain 0.8) $ slow (times/denom) $ fast 4 $ s "bd"

Which produced in the output:

tidal> skip: 9
skip: 8
skip: 36

Tidal kind of "stopped" playing for a bit, then kind of caught up, but then even after modifying the pattern to be much simpler (e.g. s "bd") the timing/sequencing was very erratic.

Environment

• OS: MacOS
• Tidal version: 1.4.4

[yaxu]

Strange that the timing goes permanently awry after a skip, I haven't seen that behaviour. Do you get late messages in supercollider as well, when that happens?

[kindohm]

I don't recall if I see late messages in SC or not. When I get a chance I will try to cause the issue and provide more info here.

[kindohm]

Just caused the issue now:

Tidal code:

  d1
    $ fast (smooth "<1 1.2 0.8 2 1.5 3>") 
    $ s "bass*16" 
    # n 1 # gain 1.2 # cut 1 # cps 1

GHCI output:

tidal> tidal> tidal> tidal> Warning: GHCi | skip: 15
Warning: GHCi | skip: 22
Warning: GHCi | skip: 88
tidal> Warning: GHCi | skip: 209
tidal> 

Lots of late messages in SuperDirt:

late 0.183419958
late 0.651652582
late 1.724338114
late 4.974297644
late 5.009923078
late 5.033939331
late 5.069567054
late 5.093585585
late 5.129235541
late 5.153226388

Tidal code no longer evaluates in my editor. GHCI output has stopped. Sometimes sound comes out at unpredictable times.

Rebooting SuperCollider alone does not resolve the issue. I need to restart my editor in order to resume coding.

If I restart my editor and eval the same code, I immediately get skip messages.

I had to slow cps down to 0.1 for the skip messages to stop appearing.

Granted, my pattern is producing a lot of events, and I would expect the late messages. But I'm surprised the skip messages happen for low values of cps. The fastest that that above pattern would go is fast 3 $ s "bass*16", and that sounds just fine (at cps 1) when not interpolating fast with smooth.

I can see how smooth might cause Tidal to grind hard when doing the interpolation calculation. I'm probably pushing it too hard, but this is a perfect example of seeing skip.

[yaxu]

Ok I can reproduce this strange behaviour.

I run the pattern, then stop it.

Then I run d1 $ sound "cp", and the claps come at uneven intervals, with ghc pegged at 100% cpu.

I stop it and the cpu drops to normal. Very strange !

[yaxu]

This is fine:

d1 $ fast (slow 8 $ saw + 1) 
   $ s "bass*16"

This isn't:

d1 $ fast (slow 8 $ saw + 1) 
   $ s "bass*16"
   # cps 1

This is fine:

d1 $ fast (segment 8 $ slow 8 $ saw + 1) 
   $ s "bass*16"
   # cps 1

So it seems to be a weird problem around continuous changes to fast and the handling of cps

[yaxu]

These examples work fine for me now..