bug in rational arithmetic ?



Richard Fateman wrote:
> 
> bfloat(2.1d0) should be
> 
> 2.100000000000000088817841970012523233890533447265625b0
> 

Should we make this change?  I think we should.  As mentioned, it does 
break one test case, but that wasn't really right either, due to 
roundoff issues.

Ray