Fix z-billboard + y to velocity transform alignment to correctly respect non-uniform scale

[conorlawton]

Adds an alignment option for GPUParticles3D which does not alter the scale of the particle transform. This allows for manipulating the scale of the particle in the particle process material (e.g. being able to stretch it based on the particle's velocity).

[QbieShay]

Hey 👋 this should already be possible with what Godot provides. Can you supply an example project that uses it, highlighting which parts are dependant on this PR?

[conorlawton]

Hi there, so the current z-billboard + y to velocity alignment (which should only control particle rotation) is averaging the scale of the particle. This causes some strange behaviour when trying to alter the particle scale from within the particle process shader.

With the change:

Without:

The custom part of the particle shader in the attached project.

Project:
Particles.zip

[QbieShay]

Ah. This looks like the flag is turning the non-uniform scale into uniform scale. I think this is a bug.

[conorlawton]

I had considered removing the scaling from the existing alignment case but was afraid that it'd break existing effects for no real gain.

If you'd like, I could commit again to remove the case I added and remove the scaling on the existing case.

[QbieShay]

No, the existing one should keep scaling, but in fact it should keep scaling separately per-axis, while it seems that now it's using the same value for all axes.

[conorlawton]

Sorry, by "remove the scaling" I meant removing the averaging and retain the scaling per axis.

[QbieShay]

Ah, then yes. I believe we should do that. I only have concerns regarding particle trails, because this option was meant initially for particle trails. It could break with nonuniform+trails, but we'll see. I pinged reduz on rocket chat about it.

But otherwise, then we agree. It should respect the nonuniform scale.

[QbieShay]

Could you test the interaction of nonuniform scale and particle trails?

[conorlawton]

Particle trails scale as expected.
With non-uniform scaling:

Without (old):

[QbieShay]

Didn't test myself but the changes are good and we discussed both here and on RC.

[akien-mga]

I tried to rename the PR to be clearer about what it does, assuming I understood correctly. Could you also amend the commit similarly? It currently starts with "Removed additional case", which doesn't make much sense without context (I think it refers to previous work in this PR which has since been squashed?).

[QbieShay]

Yep you're correct. I'd say this is a fix actually. "Fixed z-billboard-y-velocity to correctly respect non-uniform scale instead of averaging scale"

[akien-mga]

Thanks! And congrats for your first merged Godot contribution 🎉