sin etc of huge numbers

[davjam]

sin 1e100 is clearly nonsense. Should it return NaN? Some experiments with smaller numbers:

sin 1e14
-0.2094084333834997
sin 1e14 :: Float
0.9671488

In https://www.python.org/:

import math
math.sin(1e14)
-0.2094083074964523

In https://www.tutorialspoint.com/execute_lisp_online.php:

(format t "~d" (sin 1e14))
0.96697646

(iPhone & Window calculators give ~-0.209, as does sinLarge with piRat to 1000 fracs)

[davjam]

ISO 10967-2 requires big angle r(F) - the max angle that sin, etc, is required to be valid for, and absolute_precision_underflow error (giving a qNaN) for angles outside of this.