Re: [GHC] #8780: abs for IEEE floating point is slightly wrong.