Fix build with bullet double precision library

[rapsys]

This pull request should fix issue #191

[logzero]

I think it should be enough if you just cast the constants btScalar(0), btScalar(1) etc, would be less noisy that way.

[logzero]

Although I'd be also fine with explicit instantiation Min<btScalar>(...) etc.

[rapsys]

Although I'd be also fine with explicit instantiation Min<btScalar>(...) etc.

Sorry, I am not a c++ expert, I don't see what you mean.

An example of code would be welcome.

[logzero]

Min/Max/Clamp are so called function templates expecting the same argument type. The compiler will complain if they get mixed type arguments (like float and double for example), thus the casting business.

Alternatively you can explicitly tell the compiler which type of function should be used (instantiated) by writing Min<float> or Min<double> etc and the compiler will cast the arguments accordingly if possible.

So instead of

Min(btScalar(a), btScalar(b));

you can write

Min<btScalar>(a, b);

[logzero]

I've pushed a slightly cleaner variant of the fix.

Thanks for looking into it.

[rapsys]

Thanks for the upstream integration.