Perfect physics equivalence between Corona and Gideros

Drix

Dear friends,

I'm converting a game from Corona to Gideros and everything has been working fine, except the physics. For this game, it would be very important I could set the right parameters in order to produce the same physics behavior it was in Corona. A minimal difference on the trajectory of objects can make the level not be solvable.

OK, in Corona this game uses:

- content = {width = 320, height = 480, scale = "letterbox", fps = 30}
- physics.start() -- default value (body.isSleepingAllowed = true)
- physics.setScale(100)
- physics.setGravity(0, 9.8) -- default value
- physics.setPositionIterations(8) -- default value
- physics.setVelocityIterations(3) -- default value
- physics.setTimeStep(-1) -- default value (must be 1/30 for this case, right?)

In Gideros this game uses:

- Scalemode = Letterbox, Logical dimensions = 320 x 480 px
- application:setFps(30)
- b2.setScale(100)
- b2.World.new(0, 9.8, true)
- b2.World:step(1/30, 8, 3)

This produces very approximate results, but still not enough to generate a satisfactory equivalence. Is there anything I'm missing?

@atilim, would it be possible to achieve a perfect physics equivalence between Corona and Gideros?

Sorry for my bad English and thanks in advance

plamen

@Drix Is it possible to post also the fixtures and body definition for an object who doesn't act equally.

Drix

@plamen, I was coding two test samples (one in Corona and other in Gideros) to demonstrate the results and I realized the last Corona's build is performing the physics simulation a bit different from the old builds.
Probably they improved the physics in their SDK and unfortunately my game, that was created using an old build (845), is now facing physics issues with their last build.

The good news is that the results using the last Corona's build looks very similar to the results from Gideros.

So... I think I'll have a tough job ahead by recalibrating everything on all levels, unless I could identify what has changed in Corona regarding their physics.

I am attaching the test samples.

plamen

@Drix: I downloaded and started both projects. They run exactly the same way in Gideros and Corona. Here is what i think you probably miss. When you set your frame rate in Gideros at 30 fps it will work fine on actual device but it doesn't work on Gideros player so you have to go to the player menus and set the frame rate manually for 30fps. /default frame rate of Gideros player is 60 fps/ My player was already set because i am working on game targeted at 30 fps. My setup is: Win7/32bit 4G ram Intel Core2quad. Nothing special. I hope this will help you. I

Drix

Hi @plamen, thanks for replying.

I had already tried those fps settings in Gideros Player. Thanks anyway.

I think that is a Corona/Box2D issue. They upgraded Box2D library to version 2.2.1 after the build 2012.840.

Did you try running the project in a older Corona build? The ball is performing a different trajectory in older builds.

Is there a way to use Box2D 2.2.0 in Gideros?

gorkem

@drix theoretically yes, but will take lots of time to rebuild the system only for a version change for a specific test. I hope you understand.

Drix

@gorkem: No problem. Sooner ou later I would have to update the game.

Thanks

plamen

@Drix. I tested it in latest Corona build . Cant do it in older.

john26

One difference I noticed is in setting velocities. In Corona I used:

ego:setLinearVelocity(-100,vy)

while in Gideros I used

ego.body:setLinearVelocity(-3.5,vy)

(this assumes default scale factor of 30 which I did not set)

to get the same effect! I believe the scalefactor is applied only to positions, not velocities in Gideros, while it seems to be applied to both in Corona.

so if you are setting velocities and forces you will need to scale between Gideros and Corona.

ar2rsawseen

I believe the scalefactor is applied only to positions, not velocities in Gideros, while it seems to be applied to both in Corona.

Yes you are right and that's a good point

john26

At the time I just used trial and error to get to 3.5 instead of 100. Now I see it's due to the factor of 30. (thanks to your reply to my other post)

atilim

@john26 @ar2rsawseen - here is the list where scale factor is used:

1. getPosition, setPosition, getWorldCenter, getLocalCenter functions
2. queryAABB coordinates
3. rayCast coordinates
4. the applying point of applyForce and applyLinearImpulse
5. coordinates/dimensions of PolygonShape, CircleShape, EdgeShape, ChainShape
6. coordinates of anchor points
7. length/maxLength parameters of joints

And scale factor is not used on velocities, forces, accelerations, gravity etc... Still you need to think in meters while passing these parameters (e.g. 9.8 m/s^2)

atilim

@Drix Box2D 2.2.1 is used in Gideros.

john26

Thanks very much, @atilim, that's very clear! One minor question: why is rotation in degrees for sprites but radians for bodies?

OZApps

Thanks very much, @atilim, that's very clear! One minor question: why is rotation in degrees for sprites but radians for bodies?

There was a discussion on this a while ago, you can find it here

http://www.giderosmobile.com/forum/discussion/comment/13987