Fix Parallax2D physics interpolation

[rburing]

The camera_screen_center value was stale in case of interpolation, so in that case we compute it using the affine inverse of the interpolated camera transform.

• Fixes Physics 2D Interpolation does not work on Parallax2D. #91625

[akien-mga]

CC @markdibarry

[markdibarry]

I did some quick tests, and checked the screen center values while using camera drag with and without physics interpolation. LGTM.

[lawnjelly]

I'm not clear on what adj_screen_pos should be in the case of camera rotation. Should it be the top left of the rotated screen, or should it be the centre minus half the screen size?

I suspect they can both just derive the adj_screen_pos from the xform, this would make the code a bit simpler and less bug prone in future. The cost of a single xform for each camera doesn't seem worth worrying about if it makes a single path for all cases.

[markdibarry]

@lawnjelly Oh good catch. I didn't test with rotation and totally should have. The original behavior is still correct, but the interpolated version now has different behavior. I used a camera with position and rotation smoothing to show the difference.

Without interpolation:

untitled.mp4

With interpolation:

2024-05-09.08-21-26.mp4

Even worse when you try to do a full 360:

untitled.mp4

[rburing]

@markdibarry Could you share your test scene as a ZIP file?

[markdibarry]

Here you go!
parallaxmrp.zip

[rburing]

Thanks for the reviews and the test project! This should work fine now.

[markdibarry]

We're getting close! The only problems I found was when I zoomed a camera in or out.

When zoomed out I noticed it was scrolling differently with interpolation on, and eventually broke when it got to the the loop point:

Just to preempt the question of why the first image has the looping looking weird too, that's outside the normal bounds of the viewport if the zoom was set to 1. If you were to increase the repeat_times it works correctly, but still doesn't with interpolation:

[rburing]

@markdibarry I don't understand how you got to the situation in the screenshot, please upload the new MRP.

[markdibarry]

@rburing I just set the zoom on the camera to 0.5.

[rburing]

@markdibarry I made a silly mistake, the PR should be good to go now.

It works because the final lines of Camera2D::get_camera_transform() are:

godot/scene/2d/camera_2d.cpp

Lines 244 to 246 in be56cab

	camera_screen_center = xform.xform(0.5 * screen_size);
	return xform.affine_inverse();

[markdibarry]

Ran through tests for most things I think would be affected. LGTM. Great job!

[lawnjelly]

I think this looks as correct as possible, and I guess testing will reveal any problems here.

(It's been a while since I looked at Camera2D, and the whole logic of doing processing in get_camera_transform() has long needed refactoring.)

[akien-mga]

Thanks!