Closed Andreya-Autumn closed 2 months ago
Adding an all sounds off call before unstreaming changed nothing.
Overlapping voices seem to trigger it somehow somehow, the noises always seem to happen at the start of the release stage.
I'm also getting a real weird version of it now: First hit plays right, second hit one or both channels cuts out halfway through the kick sound and stays dead. I think it's outputting some crazy DC offset on the gone channel, but metering plugins downstream are struggling to even show what's happening.
The standalone showed this nicely. With a long release and hammering with 120bpm 1/8 notes when you re-load an identical session, the prior session notes get 'stuck'
In debug mode those stuck notes just eat note slots but in release the memory they have gets all gnarly and they output occasional splatting noise, I think when one of them tries to improperly re-use it
There was more than one place where you needed to terminate notes than the one I shared on discord (both an audio-on and audio-off version and a cautionary one in a zone) so I did all of those and then a case which blew up on the third reload with my midi stressor and standalone didn't blow up past 50. So I think that's it.
I think there may be more to it. I'm still getting the weirdness in Reaper here. Will do a clean build and a clean project to be sure.
If you can share the project and the sample it would help too I guess
Will do
Seems to be fixed in #985 indeed! Tested with a few samples at different bit depths > 16.
Thanks
A 32 bit windows build could have shown it in 16 bit samples also is my guess. Good its fixed
Make a reaper project with a shortcircuit playing a four on the floor or something Put a scope and a limiter after it if you wanna see what's going on/not trip the automute While playing, re-load the project (recent projects menu is useful) You may then get stuff like this:
If there are voices active when the re-load happens, the repro rate goes up by a lot. But I have also seen it a couple times without voices active.
This all seems to happen independent of proc routing. And changing oversampling doesn't stop it from happening, though it might change the character of the noise bursts.