4.5 2D physics - ruined my game >:(

I've just released a game on the Play store, "Clamber", which makes use of Unity 2D physics. The player drags the 'hands' of the character around to climb up through the level. This was achieved by setting the velocity of the hand to the frame-to-frame motion of the player's finger, divided by Time.deltatime.

In the previous version of Unity, all was silky smooth and perfect. In 4.5, it's garbage. The hand judders and falls behind the player's finger; it doesn't move smoothly at all despite zero code changes from the release build. It's not a frame-rate issue; the screen is still scrolling smoothly. I don't know what the heck has happened but it's utterly ruined the feel of the game and I'm obliged to roll back.

Conceivably it's the mouse/touch input rather than the physics that's the problem (which manifests both in the editor and on device), since all other physics based stuff (including the body of the character) seems to move smoothly. Anyone have any ideas?

EDIT: All the physics and time settings are the same in both. I've now rolled back to 4.3.4 and the problem has gone away.

 

Huh. Android, right? I don't see any changes to the touch controls in Android in the release notes. Have you tried using mouse input to distinguish whether it's coming from the input or the physics?

My off-hand guess would be that it is a physics thing, since that update broke my game's physics in several seemingly random places (ironically mostly bug fixes).

 

Add me to the list. I have a fairly simple 2D physics game and 4.5 has apparently changed the laws of physics completely. Firstly, everything is running about 1/3 the speed it was (not framerate, just the physics). All settings are the same as far as I can tell. Objects don't bounce or roll or fall the same way. Being a puzzle game I need to rely on the physics, and I was hoping that 4.5 would fix the issues with 4.3 (a show-stopping bug in Windows Store x86 builds and occasional inconsistent physics behaviour), which seems to be the case...but at the cost of altering the physics so much that the game is massively broken in 4.5.

 

Hey pumpkinszwan, have you tried messing with Rigidbody2D.inertia? They completely changed how rotational inertia is calculated, and exposed it there.

 

There is something wrong with the physics in the new version of Unity.
Look at the 2d demo project.
I recorded a video, that shows errors (vsync ON), the main character jumps quite weird.

https://www.youtube.com/watch?v=AHV8y08GEE8

Interesting is a fact, that interpolation works differently for vsync-off and vsync-on.
However if you disable an animator, then it works.
Now Rigidbody.Interpolate: interpolate causes that main character stutters, but background doesn't.

In unity 4.3.3 everything works correctly (free version).
You can not work now physics in the original demo 2d?
Do you also have a problems with physics in 2d demo project?

 

Yep, inertia is the key! I think I got the fix from one of your other posts. The only problem is that now I have to figure out the right values for all my items. At least my game is now acting more or less like in 4.3.

 

Same problem here, had to manually set inertia for all the ragdolls, cables and procedural controlled enemies.

MelvMay,
The inertia is calculated incorrectly, try making 2 circles of same mass but different radius, they'll have same inertia.
Inertia for circle is calculated like this:
I = m * r * r / 2
So bigger circle should have bigger inertia.
//another observation: inertia always seems to equal (mass + 0.1), regardless of object shape

For now i solved it with extra component that sets inertia. I'm porting the box2d inertia calculating algorithm now, so it can calculate the correct inertia automatically.

//done: http://forum.unity3d.com/threads/24...ia-fix-InertiaCalculator-component-and-editor

 

From the release notes:

•2D rigid-body rotational inertia no longer scales with collider size but rather with the rigid-body mass.

So... They're classifying that as a fix.

 

Thank you Melv

 

This is the first thing I noticed when I updated to 4.5. In my platformer, whenever I set a game object to active for example, my player stutters. It's even more noticeable if you have the main camera set as the child of the player as the whole screen stutters.

 

Here's the solution for inertia calculation based on explicit mass:
http://forum.unity3d.com/threads/24...ia-fix-InertiaCalculator-component-and-editor

Inertia is always proportional to mass, so you just need to calculate hypothetical mass and inertia based on hypothetical density, and:
explicit_inertia = hypothetic_inertia * explicit_mass / hypothetic mass;

My game relies heavily on physics, occasionally with things like this:


It relies on complex rigidbody control, exposing inertia property in 4.5 was a real gift, but new inertia calculation algorithm is completely wrong. It might work for simple single-bodied objects, but none of the above is possible with new inertia calculation.

My suggestion would be to use Box2D inertia calculation (like in thread above), and add inertia_multiplier property to rigidbodies, so it would be possible to make inertia proportionally higher or lower.

 

Thanks, MelvMay!
Sorry, i might've overreacted, 4.5 update is awesome.

 

Wow these things are amazing! :mrgreen: