#165 Error in documentation of MersenneTwisterUniformRng

[Mikail Sheikh]

The documentation of the random number generators states that the values fall in the open interval (0.0, 1.0), i.e. without the end points.

If MersenneTwisterUniformRng compiled with 32 bit Real numbers
When QuantLib is compiled with 32 bit floating point precision QL_REAL=float the line

return (Real(nextInt32()) + 0.5)/4294967296.0;

can evalueate to

return (float(4294967295) + 0.5)/4294967296.0;

This evaluates to 1.0f exactly

[Luigi Ballabio]

As compiling with QL_REAL=float is discouraged (the test suite fails) I think mentioning this might be more confusing than useful.