cocos2d, box2d and retina mode: give up, or try to get it to work?

I've been trying to get my box2d game to work in retina mode, and I'm running into a lot of irritating issues when running in higher-resolution mode.

cocos2d renders my graphics correctly in retina mode, but I'm finding the need for hack after hack to get box2d to work the same as in lower-resolution mode. For example, I'm finding I need to do something like this to get the debug draw shape size correct in retina mode:

b2Vec2 vertices[4];
vertices[0].Set(-0.5f / CC_CONTENT_SCALE_FACTOR(), 1.0f / CC_CONTENT_SCALE_FACTOR());
vertices[1].Set(-0.5f / CC_CONTENT_SCALE_FACTOR(), -1.0f / CC_CONTENT_SCALE_FACTOR());
vertices[2].Set(0.5f / CC_CONTENT_SCALE_FACTOR(), -1.0f / CC_CONTENT_SCALE_FACTOR());
vertices[3].Set(0.5f / CC_CONTENT_SCALE_FACTOR(), 1.0f / CC_CONTENT_SCALE_FACTOR());  
int32 count = 4;    
b2PolygonShape polygon;    
polygon.Set(vertices, count);

That hack (adjusting all vertex points by CC_CONTENT_SCALE_FACTOR()), of course inevitably leads to a density hack, to keep movement similar to lower-resolution mode:

b2FixtureDef fixtureDef;
fixtureDef.shape = &polygon;开发者_StackOverflow社区
fixtureDef.density = 1.0f * CC_CONTENT_SCALE_FACTOR();
fixtureDef.friction = 0.9f;
playerBody->CreateFixture(&fixtureDef);

And that leads to another adjustment hack when applying forces:

b2Vec2 force = b2Vec2(10.0f / CC_CONTENT_SCALE_FACTOR(), 15.0f / CC_CONTENT_SCALE_FACTOR());
playerBody->ApplyLinearImpulse(force, playerBody->GetPosition());

Keep in mind, I'm drawing in debug mode by scaling my debug draw calls, like so:

glPushMatrix();  
glScalef(CC_CONTENT_SCALE_FACTOR(), CC_CONTENT_SCALE_FACTOR(), 1.0f);
world->DrawDebugData();
glPopMatrix();

What is it that I'm fundamentally misunderstanding about getting box2d to work with retina mode? I'm using Steffen Itterheim's suggested Box2DHelper class in place of PTM_RATIO, but that just doesn't seem to be enough. Any ideas? I'm about to give up entirely on retina mode for my game at this point.

I'm using the GLESDebugDraw class that comes with Cocos2D+Box2D Xcode template and the way I make the debug draw to scale up when in retina mode is just as simple as:

m_debugDraw = new GLESDebugDraw( PTM_RATIO * CC_CONTENT_SCALE_FACTOR() );
world->SetDebugDraw(m_debugDraw);

I don't need to use CC_CONTENT_SCALE_FACTOR() anywhere else since, except in some explicitly named class methods/properties like contentSizeInPixels etc, Cocos2D uses point-based coordinate space in which coordinate values remain the same whether in retina mode or low-res mode, which means the same PTM_RATIO can be used to create Box2D fixtures or shapes in both modes.