[Rd] segfault in approx() (PR#6809)

[kjetil at entelnet.bo]

This is R1.9.0, windows XP

I run across the following:

It is admittedly silly, but it shouldn't segfault:

test <- list(y=1:100, x=rep(NaN, 100))
approx(test, xout=rep(NaN, 10))

Bombs!

Kjetil Halvorsen