[3.x] Bullet: Revert "Fixed gravity scale"

[Klowner]

This reverts commit d250ade which appears to have been made in error? (Looks suspicious, commented out line followed by a line which ignores the parameters in favor of a hard-coded zero vector).

Aside from ensuring the gravity vector is normalized (since it gets multiplied by gravityMagnitude), I don't see how reverting this would cause any issues.

Fixes #32196

[Klowner]

Also fixes #35378

[akien-mga]

This reverts commit d250ade which appears to have been made in error? (Looks suspicious, commented out line followed by a line which ignores the parameters in favor of a hard-coded zero vector).

See the PR which includes this commit: #13046.

It states:

• Fixed gravity scale - Fixes RigidBody falling slightly with Gravity Scale set to 0 RigidBody falling slightly with Gravity Scale set to 0 #12770

Can you confirm that this does not reintroduce #12770?

Also fixes #35378

Please list it in the OP so that the issue can properly be identified as linked / closed when the PR is merged (this actually doesn't work when merging PRs in non-master branch but it's for the principle/consistency :)).

[Klowner]

Bah, thank you, I was unaware of that one.

Looks like this does trigger that regression. Gravity must be affecting the body's velocity for a tick or two before the gravity scale setter is called.

I'll do a little more poking.

[akien-mga]

Did you have any luck looking into this further?

[Klowner]

Thanks for pinging me on this! I looked at this again today with (very) fresh eyes and I think I understand the problem but I don't have very strong feelings about any potential solutions.

The problem with the regression is due to the order of things being initialized, kinda:

1. Bullet Space is constructed
2. Gravity is set on Bullet Space (without this change, this gravity is always set to (0,0,0))
3. RigidBody is created
4. RigidBody gravity scale setter is called (and sets a "scratched" flag to indicate that changes have been made, fine)
5. RigidBody is added to the Bullet Space and the Bullet Space automatically calls body->setGravity() with the Space's gravity that was initially set in step 2 ⚠️ the RigidBody's gravity scale is not accounted for. ⚠️
6. Physics simulation tick occurs
7. RigidBody's dispatch_callbacks() is called which detects the "scratched" flag from step 4 and then calculates a new gravity which gets multiplied by the RigidBody's gravity scale before calling body->setGravity().
8. Second physics tick is performed, and our RigidBody with gravity now set to zero already has 1 physics's tick worth of gravitational velocity applied, so it takes a few more frames to settle.

One idea I had was to set the RigidBody's gravity after it's added to a space to overwrite the incorrect value. Seems kinda dumb but it seems like that might be the only way to include the body's gravity scale? Any input appreciated :)