Reverb effect still audible after setting effect slot gain to 0

[Vercidium]

When changing the AL_EFFECTSLOT_GAIN of an AL_EFFECT_EAXREVERB effect from 1.0f to 0.0f, sometimes the reverb is still audible for a few seconds.

It happens intermittently, and I've only noticed it when:

• moving the listener position
• setting AL_EAXREVERB_DECAY_TIME via alEffectf each frame. Even if it's setting to the same value (e.g. 12.0f) each frame.

This video shows 4 spatialised sources sending their input unfiltered into the same EAX reverb effect, and then toggling the AL_EFFECTSLOT_GAIN of the effect between 1.0f and 0.0f. Normally it cuts off instantly, but at the end of the video it is still audible for a few seconds, then cuts off.

The reverb effect's gain and decay time is visible at the bottom of the screen:

https://www.youtube.com/watch?v=Y75Y1RdBmXk

I've reproduced this on v1.25.0 and v1.25.2.

[kcat]

That fading at the end like that should only happen when certain reverb properties are modified. It starts processing a second reverb pipeline as the previous one fades out with the sound it had, to avoid an abrupt and disorienting change to the audible environment. For example, entering a carpeted hallway with little reverberation when leaving a large echoing concert hall; it's very disorienting to have the concert hall echoes stop on a dime, the echoes from the concert hall should fade out smoothly as they had been when the hallway's softer and shorter echoes start (this is the kind of thing that happens with hardware, from what I've been able to tell).

Although changing the effect slot gain by itself shouldn't cause that to happen. The effect slot gain isn't considered part of the environment for that purpose, and the caller can change it as fast or slow as it wants. The properties that would cause that decaying fade are Density, Diffusion, Decay Time, Decay HF and LF Ratio, Modulation Time and Depth, and HF and LF Reference Frequency. The values actually need to be different, simply setting the same value again won't cause it (though it does an equality comparison, so it could still happen even if the difference is imperceptible; I have been meaning to look into how much of a similarity threshold there can be to change those properties in-place without any audible noise or disorientation).

[Vercidium]

The properties that would cause that decaying fade are Density, Diffusion, Decay Time, Decay HF and LF Ratio, Modulation Time and Depth, and HF and LF Reference Frequency

That makes sense - if I leave these properties as is and toggle the effect slot gain from 1.0f to 0.0f, the reverb cuts off correctly.

Is it possible that the previous reverb pipeline's output is not affected by the new effect slot gain?

[kcat]

Is it possible that the previous reverb pipeline's output is not affected by the new effect slot gain?

That's right. When the effect slot gain and the effect properties are changed at the same time (that is, they're changed close enough in time that they both get scheduled to apply together with the internal update), the previous pipeline is left as-is to fade out with what it had, as the new properties are applied to the newly activated pipeline. Though it may be possible for me to change it so the effect slot gain applies to the old and new pipeline together instead of just the new pipeline, since it's more about how the effect is mixed out and not really part of the environment (the reverb itself has a gain property for that, separate from the effect slot gain).