Bullet 2.66 released (minor fixes for 2.65)

[Erwin Coumans]

http://www.bulletphysics.com/Bullet/php ... =18&t=1783

Not all features made it, they will be added for next release. See the Bullet todo here

Thanks again for all your feedback and contributions, and happy holidays!
Erwin

[Enrico]

There is a typo in Extras/BulletMultiThreaded/SpuParallelSolver.cpp:31:

Code: Select all

#include "LinearMath/btQuickProf.h"

when it should be

Code: Select all

#include "LinearMath/btQuickprof.h"

Note the small 'p'-character in the second version. Or rename the mentioned header file

[reltham]

The convexTest hits a triangle mesh, the m_hitNormalLocal in the convexResult is already in world space, but the code in AddSingleResult() is transforming it again thinking it's not in worldspace (the bool param normalInWorldSpace is false).

[snowcat]

The convexTest hits a triangle mesh, the m_hitNormalLocal in the convexResult is already in world space, but the code in AddSingleResult() is transforming it again thinking it's not in worldspace (the bool param normalInWorldSpace is false).

I found this bug too. Cost me 2 days debug my program, and finally found that's a bug in Bullet. ( 2.67Alpha has this bug too.)

Code: Select all

bool	btContinuousConvexCollision::calcTimeOfImpact(
				const btTransform& fromA,
				const btTransform& toA,
				const btTransform& fromB,
				const btTransform& toB,
				CastResult& result)
{
....
				n = pointCollector.m_normalOnBInWorld;
....
		result.m_normal = n;
}

the normal is in world, but btCollisionWorld::objectQuerySingle call addsingleresult with wrong parameter when collisionShape->isConcave:

Code: Select all

							bool	normalInWorldSpace = false;
							return m_resultCallback->AddSingleResult(convexResult,normalInWorldSpace);

btw: I think btCollisionWorld::objectQuerySingle's code is copied from btCollisionWorld::rayTestSingle. But they have little different make this bug.

[John McCutchan]

Thanks for the bug report. A fix has been committed to SVN.