100% consistent crash with pcontrol and FrameLib

[Former user]

In most scenarios I would use poly~ to mute anything I didn't want to eat up CPU, but in the attached patch I can use pcontrol to crash a patch with a framelib network doing a particular series of actions.

I apologise if the patch is confusing. Basically, I have put the process that crashes inside 'crash' and another very simple process inside 'no_crash'. The 'no_crash' patch just generates a random number between -1, 1 every sample while 'crash' is a bit more complex.

To force the crash toggle the toggle to 1 for 'crash' then turn it off. Doing the same with 'no_crash' does not cause max to close.

I would probably never need pcontrol but perhaps this points to a bug with other repercussions that I'm not capable of finding myself.

https://pastebin.com/5qYEh0XS <-- Logs

James

[Alex Harker]

I don't need to read the code to know that switching parts of the network off with pcotronl is bad news. The workaround for now (and possibly forever) is not to do it.

Having said that:

A - I like the sounds this is making B - this doesn't crash my computer so maybe the specifics of what causes this crash are fixed in the more up to date version C there are still definite scenarios that I could probably easily recreate in which frame lib would crash

I'll investigate more later at low priority as this is an edge case, and there are some other more pressing issues to do with poly~....

[Former user]

Thanks for taking a look. I don't remember why I used pcontrol over poly~ for this specific example but thought the behaviour was interesting (at least enough to put it up here for you and others to see).

[Alex Harker]

OK. I've looked into this and it got fixed more or less by accident. However, I've managed to recreate the crash by allowing enable messages to be passed through to audio objects and then it crashes again. When I remove the code that passes them through that goes away.

So - I'm not 100% clear on why the version at that time crashed, but I suspect that I was going through an iteration of the wrapper that allowed the enable messages through some how. It's not worth chasing as I've verified that pcontrol and related messages have no effect on fl. objects now. Turning audio on and off in a poly~ is fine, because that is like freezing all time, whereas here the issue is parts of a network getting out of step with other parts.