JuliaExperimental: incorrect use of AbsInt and GcdInt; create_rational is dangerous

[fingolfin]

JuliaExperimental calls AbsInt and GcdInt with rationals as arguments, which cannot work and will lead to crashes and/or corrupt data.

Note there is a kernel function AbsRat which can be used instead of AbsInt. This function is also available on the GAP level as ABS_RAT, so it could also be called as GAP.GAPFuncs.ABS_RAT. However, this may be slower compared to the ccall, see also issue #68 . Another thing one could try is to simply use ccall(:AbsRat, ..., i.e. without setting up a wrapper in gap_macros.c

For GcdInt, it is unclear to me what the intended meaning is when computing the gcd of two rationals. I'd say it is always 1 (and GAP agrees), except of course if both inputs are 0.

Lastly, create_rational is rather dangerous, as it allows creating denormalized rationals, such as 2/1, 8/4, 3/0 or 2/-1. GAP is not prepared for this -- it expects rationals to always have a denominator > 1. If that rules is violated, then e.g. the rational 2/1 in GAP is not equal to the integer 2.

I suggest to instead call the GAP kernel function QuoRat, which accepts both rationals and integers as arguments.

Lastly, with the current master branch, the wrapper functions MyFuncLT in the kernel are no longer necessary, as LT etc. all were converted from macros into functions.