#1130
- Shapes are now directly attached to Bodies when they're created (similar to love 0.7 and older).
- Fixtures are removed.
- All methods that were in Fixtures now exist in Shapes.
- All APIs that used or returned a Fixture now do the same with a Shape.
- Add new love.physics.new*Shape variants that take a Body as the first parameter.
- Deprecate the new*Shape APIs that don't take a Body.
- Deprecate love.physics.newFixture (the deprecated function now returns a Shape).
- Replace Body:getFixture and Body:getFixtures with Body:getShape and Body:getShapes (Body:getFixtures is deprecated).
- Replace World:queryFixturesInArea and World:getFixturesInArea with World:queryShapesInArea and World:getShapesInArea (queryFixturesInArea is deprecated).
- Replace Contact:getFixtures with Contact:getShapes (Contact:getFixtures is deprecated).
- Replace all love.physics callback Fixture parameters with Shape parameters.
- Deprecate ChainShape:getChildEdge.
All the Memoizer stuff was pointless: when you create a Fixture, it clones the b2Shape, so there was never going to be a memoized Shape to find. And cloning the Shape is incorrect as well, since you should able to modify the Shape retrieved through getShape and have it affect the Fixture. Everything's much simpler now.
while technically density is (according to the Box2D docs) in kg/m^2, it seems like a number that can be safely left un-scaled, for simplicity's sake; the concept of a shape/fixture's density seems sufficiently divorced from the MKS system as to render this a non-issue. if this proves to be a problem in use, scaling can be re-added.
Constructing invalid shapes, e.g. a rectangle with width 0, could result
in a b2Assert to fail. In turn, Shape::~Shape tried to destroy a the shape
with an uninitialized pointer, yielding a segfault. Add NULL-initializers
and checks for NULL to prevent this.
Sasha Szpakowski
Bart van Strien
rude
Aidan Gauland
Bill Meltsner
bart@bartbes.ath.cx
bart@bart