float_int():  Some systems raise an exception if a double is cast to
long but the double is too big to fit in a long.  Prevent that.  This
closes some recent bug or patch on SF, but SF is down now so I can't
say which.

Bugfix candidate.
1 file changed