nonuniform treatment of EXP() and SIN() in FLOAT()



Please report this bug to sourceforge (see bug_report() ).

I suspect that float is overenthusiastically avoiding the case
float(x^2) => x^2.0.

In the meantime, a workaround is float(...),numer.

> When applied to compositions like EXP(EXP(2)), or EXP(SIN(2)),
> the float() function seems unable to get a complete result.
> If we consider SIN(SIN(2)), or SIN(EXP(2)) instead, the result
> is complete. The bfloat() works as expected. The output is below.