[Kst] Solaris compile problem...

[Brisset, Nicolas]

I think Barth recently introduced a non-portable change:
********************************
r503057 | netterfield | 2006-01-28 01:36:35 +0100 (Sat, 28 Jan 2006) | 5
lines

BUG: 120823

Read past EOF results in a vector of 1 sample = NaN.
**********************************

On my Solaris box (Solaris 2.8, without any manual tweaking and not in a
freshly checked out tree) this does not compile. I think NAN is in a
different header on Solaris, but probably the way it's done is
kstvector.cpp (lines 39 to 43) is the cleanest...