There is a test for 'NaN ** 0' returning 0 in S02-types/nan.t skipped for the
reason 'unspecced and inconsistent'. Also there is a comment above the test:
# if we say that 0**0 and Inf**0 both give 1 (sse below), then for which
# number or limit whould $number ** 0 be different from 1? so maybe just say
# that NaN ** 0 == 1?
MoarVM and JVM agree with that last suggestion:
< bartolin> r: say NaN ** 0
<+camelia> rakudo-{moar,jvm} d8ce4c: OUTPUT«1»
As far as I understand it, that is in line with the IEEE 754 Standard 2008
(e.g. http://stackoverflow.com/questions/17863619/why-does-nan0-1).
I'm inclined to change the test accordingly (expecting 1 as the result) and
unfudge it. Any opinions?
