RFA: Support infinity, NaN, and denormalized numbers in floatformat.c

[Andrew Cagney]

> That sounds reasonable to me.  floatformat_to_doublest could contain a
> check on HAVE_LONG_DOUBLE and call one or the other.

something like

floatformat_from_doublest()
{
#if HAVE_LONG_DOUBLE
   return floatformat_from_long_double (...)
#else
   return floatformat_from_double (...)
}

Yes, that would keep GDB insulated from liberty's configuration internals.

(Any long term signs of a portable target floating point library (so 
that the code can completly avoid host fpu assumptions?).  glibc's 
floating-point code has been suggested in the past.)

Andrew