Fix Bone2D scaling

[TokageItLab]

Co-authored-by: Thiago Lages de Alencar thiagola92@gmail.com @thiagola92

• Supersedes / Closes Fix scaling last bone length with negative scales #81048
• Fixes The last Bone2D in the chain will scale incorrectly #80643

To minimize the effect of the fix; Context: #81048 (comment) Revised overall behavior.

[TokageItLab]

I think the problem was that the gizmo scaling was scaling after the rotation. So, perhaps the scaling of Node is done before the rotation? I think that is where the direction mismatch was occurring.

Also, the abs() was needed in #81048 because there was a reverse rotation there to cancel first rotation (it was done rotation->scaling->rotation.inv). So now I make to simplify the calculations for the deformation so that they match the Node scaling.

@thiagola92 Does this change make sense to you?

[thiagola92]

I think the problem was that the gizmo scaling was scaling after the rotation. So, perhaps the scaling of Node is done before the rotation? I think that is where the direction mismatch was occurring.

Also, the abs() was needed in #81048 because there was a reverse rotation there to cancel first rotation (it was done rotation->scaling->rotation.inv). So now I make to simplify the calculations for the deformation so that they match the Node scaling.

@thiagola92 Does this change make sense to you?

I tested and it seems to be working 👍

I'm still confuse about how the rotation was affecting everything.
In my head get_rotation() was useless in this lines because would cancel itself:

real_t angle_to_use = get_rotation() + bone_angle;
//
rel = rel.rotated(-get_rotation());

But if I'm understanding correctly, there were some scale operation during this step... I'm understanding correctly?

[thiagola92]

Note: This would fix Bone2D length when scaling

To fix height too would probably need something like #81544 (which I used abs() probably without needing)

[TokageItLab]

I'm still confuse about how the rotation was affecting everything.
In my head get_rotation() was useless in this lines because would cancel itself:

In the old code, the vector after rotation (Bone Angle + rotation) was globally scaled.

This code globally scales the vector after rotation by Bone Angle only.

Evenually, Bone Angle is a Bone2D property, so it should be handled by Bone2D, but rotation is a Node2D property, so transformation/scaling for rotation should be handled by Node2D.