In the last few years i coded my own physics engine - but i got stuck and decided to use jbox2d instead.

Now i added it to my java-based entity component game engine and its not working correctly...

There are gaps on every body - like the shape extends a lot more than the original shape definition and its like it floating around the bodies and the gravity affection is extremely fast. Using a gravity vector of (0, -10) is like instantly let the body crushes to the ground.

Is jbox2d so much different than normal box2d - or is my implementation wrong?

This is the physics systems which registers and unregisters new entities and creates or destroys box2d bodies for all the shapes/fixtures and of course steps the world:

http://pastebin.com/rERkFTwV

And the rendering does not operate on the body directly, but use the entities transform component which are two-way syncronized with the bodies position/rotation in the physics/system, also use the shapes for rendering which was the base for the box2d shape definition - so if box2d has not change this, the values should be equal to my shapes.

http://pastebin.com/5qsE2TGM

Where is my implementation wrong?? (See attached image)

For clarity sakes i uploaded a small video showing the problem:

After analyzing the created bodies, everything seems to be created correctly.

The polygon-shapes created just contain the same "vertices" i use for rendering. I set the half size of the polygon shape definitions which is equal to my box geometry component which are also defined in half sizes....

And another weird thing happened, if i syncronize the body position / rotation before i world step, the two rotated boxes on the bottom rotates around O_o - which should never happen, because they are static bodies! Maybe the rotation conversion is shit? I do for the step deg2rad which is simply (PI / 180.0f * transformComponent.rotation) and vice versa after the world step rad2deg (body.getTransform().getAngle() * 180.0f / PI), because my rotation property in the transform component is in degree.

What the hell is wrong???

Here a video just with the two static bodies which rotates:

Btw. i also removed the degree to radian conversions, because it was not really needed. It is better to store the transformation angle always in radians which are the same as in box2d - but the behavior is the same... static rotated box, rotates O_o

I created a one java class lwjgl (opengl) application which integrates jbox2d and it has the same issues:

- Gravity is too heavy

- Gaps everywhere

Why??? I dont know...

Here are the source - you need eclipse, a java project, the LWJGL library + util jar and of course jbox2d jar + sl4j jar.

http://pastebin.com/GLeGmmmC

Please help me!!!

Try scaling up the simulation by some factor so everything is bigger, currently your boxes are like 10 cm big which seems quite small. That might explain the gaps (since they would not be noticeable at a bigger scale) and gravity (which will seem stronger at smaller scales).

Try scaling up the simulation by some factor so everything is bigger, currently your boxes are like 10 cm big which seems quite small. That might explain the gaps (since they would not be noticeable at a bigger scale) and gravity (which will seem stronger at smaller scales).

I tried to scale it by the modelview matrix to a factor of 30, but got the same result. And it should not matter which scale i use, for the rendering after all - because box2d does not alter the coordinate system, either use the units passed. From the box2d doc i read that the vector units are optimized to operate in meters and if i use 0.1 meters for a box and draw it with 0.1, its just a matter what scale i use for rendering - but its still 0.1 meters.

Ya you was right there, scaling the bodies helps... i scaled the body sizes by a factor of 10 and now everything looks almost right.