[Rd] trigamma for very large arguments (PR#14020)

[maechler at stat.math.ethz.ch]

>>>>> "JL" == Jerry Lewis <Jerry.Lewis at BiogenIdec.com>
>>>>>     on Thu, 22 Oct 2009 04:20:13 +0200 (CEST) writes:

    JL> Full_Name: Jerry W. Lewis
    JL> Version: 2.9.2
    JL> OS: Windows XP Professional
    JL> Submission from: (NULL) (96.237.55.233)


    JL> trigamma(x) returns 0 for x>1e152, yet
    JL> trigamma <- function(x) 1/x
    JL> gives machine accuracy for any x>1e16

Thank you, yes, and the problem is worse for digamma().
I will fix this, but not anymore for R 2.10.0 (we are in code freeze!)

Martin Maechler, ETH Zurich