bug of signif

["Molinari, Luciano" <Luciano.Molinari@kispi.unizh.ch>]

S+2000 under Windows 2000.
Notwithstanding past messages concerning a possible bug in signif, which was
not, in one case, a bug, I am puzzled by the following behaviour:
------------------
S+>signif(1)
[1] 1
S+>signif(1,dig=12)
[1] 0.9999999999999998
S+>signif(1,dig=19)
[1] 1
S+>signif(1,dig=20)
[1] 1e-019
----------------------
Using signif(1.,20) or signif(as.numeric(1),20) does not help.
R, of course, does not have this problem.
L. Molinari