Fix GPUParticles2D not randomizing seed when set_one_shot is called

[TCROC]

gpu_particelse_2d needed fixing in this regard as well as pointed out by @clayjohn and @QbieShay: #101596 (comment)

The seed should be randomized by calls to set_one_shot. Although the behavior is redundant imo, @clayjohn made a solid point that we should still do it so that the behavior remains the same for gpu particles between 4.3 and 4.4.

[TCROC]

Why does it want me to update the docs? I presume this is a false flag? All I changed was the internals of the set_one_shot method.

[AThousandShips]

It's not a false positive, it changes the default value, as set_one_shot is called in the constructor

Did you intend this to change the default seed?

[TCROC]

Ah fascinating... This is the same issue we ran into in gpu_particles_3d although it didn't flag the default value change... but it should still apply. Previously this was randomized but in a process method due to the side effect of a boolean. Now it is set directly in the method. This definitely isn't something we took into consideration... what are your thoughts on this @clayjohn and @QbieShay?

@AThousandShips here's the convo where we've been discussing in detail: https://chat.godotengine.org/channel/rendering/thread/Mgd4xD8bH3TZKrQND

[TCROC]

@AThousandShips is Math::rand() deterministic? If so, I can update the docs to make this PR happy... otherwise its just gonna be luck of the draw on the PR

[AThousandShips]

It is if you set the seed, but don't think we do that in the documentation generation, but I think the default seed should probably not be set randomly if it wasn't set to that before, but that needs to be discussed, I think the way this is done should be solved locally

(No need to tag me immediately again after tagging me)

[TCROC]

My bad. Won't tag as much. Thanks for the heads up.

I think the way this is done should be solved locally

What do you mean by that?

[AThousandShips]

As in it should probably ensure it doesn't reset the seed when constructing, I.e. only when calling set_one_shot manually, or we can reset the seed manually after set_one_shot(false) in the constructor

[TCROC]

Ah I see. I like the 2nd approach. Is clean and simple. I will go do that

[TCROC]

Why do you think this PR didn't have the same issue? #101596

[AThousandShips]

I don't know I haven't studied the code closely, you'd have to go through the code and see if there are any differences in the initialization

[TCROC]

Sounds good. I don't think either of the 2 PRs need to be held up on those questions though. A third PR to make 2d and 3d particles behave consistently may be better suited as we can discuss all those changes in one place. These 2 PRs should make the behavior of 2d and 3d particles the same going from 4.3 to 4.4. At least as far as clay and qbie noticed in the rocket chat

[Calinou]

Looks good to me.

[QbieShay]

We spoke about this in the rendering meeting.

			if (!use_fixed_seed) {
				set_seed(Math::rand());
			}

should be removed both from gpuparticle2d and gpuparticle3d.

the function calls in the constructor (set seed and set use fixed seed) should be put both here and in gpuparticles3d

[TCROC]

Sounds good. I will do that for this PR and open a new pr for 3d particles

[QbieShay]

You can do all in the same PR, it's better for reverting just in case we need to.

[TCROC]

Makes sense! Will do! :)

[TCROC]

We spoke about this in the rendering meeting.

			if (!use_fixed_seed) {
				set_seed(Math::rand());
			}

should be removed both from gpuparticle2d and gpuparticle3d.

the function calls in the constructor (set seed and set use fixed seed) should be put both here and in gpuparticles3d

@QbieShay Done 😎

[Repiteo]

Thanks!