mpf_urandomb and precision

[Marc Glisse]

Hello,

when I read the documentation of mpf_urandomb:

"Generate a uniformly distributed random float in rop, such that 0 <= rop 
< 1, with nbits significant bits in the mantissa."

I expected that if nbits was large, the function would increase the 
precision of rop so the nbits would fit. Instead, the code reduces nbits 
so that it fits prec. Should I append "or less if the precision of rop is 
smaller" to the documentation?

-- 
Marc Glisse