[Acoustic Simulation] Under certain conditions, audio breaks everywhere forever

[NOTEKAMI (NOT)]

While creating a dynamic rain system I came across a strange bug, I can’t explain why it happens, but I assume it has to do with rotated parts

When camera get closer to “tent”, this bug occurs

---

Expected behavior

Audio will still play sounds effects

---

Repo file in private message

A private message is associated with this bug report

[ReallyLongArms]

Thanks for the report; will check out the placefile

[ReallyLongArms]

Hey @NOTEKAMI – it looks like the MaterialAudio script can end up positioning some parts that have AudioEmitters at suuuuuuuuuper far away positions; I added this line as a diagnostic

		if AudioEmitterPart.Position.Y < -1000000000000 then
			print("Woah")
		end

and it fires whenever I place my camera near the tent, just like you said.

When this happens, some calculations inside the audio engine end up generating infinity or NaN, which corrupts a lot of other math as a result, and all audio goes silent

We’re going to patch our code to make it more robust against this, but you may also be able to work around it by clamping the positions of those AudioEmitterParts

[NOTEKAMI (NOT)]

I decided to check what I get when using workspace:Blockcast() and I sometimes get NaN, but all the data I pass looks normal, does this means that this is also an engine bug?

Edit:

Filled a bug report for this.

Blockcast returning math.huge or NaN position despite normal args

While testing acoustic audio, I found a bug that completely broke it, I thought that the problem was in the audio (kinda yes and no), BUT the main problem was that :Blockcast() somehow gets a NaN or math.huge RaycastResult.Position, despire all arguments are normal. This issue only (AFAIK) occurs when :Blockcast() hits rotated parts (I haven’t tested it on Sherecast and Shapecast) Expected behavior :Blockcast() will return a usual position, not math.huge or NaN Repo file in private …

[ReallyLongArms]

Thanks for filing that – yeah it might be a separate/additional issue

[ReallyLongArms]

Hey @NOTEKAMI – we added extra checks to prevent NaNs from getting into acoustic simulation; does this resolve the problem on your end?

[NOTEKAMI (NOT)]

Hey, now the audio doesn’t break when receiving NaN!
But unfortunately the main problem with Blockcast (Shapecast and Spherecast i belive too) still returns NaN under certain conditions