bug in randperm?

[Juan Pablo Carbajal]

Hi all,

According to the docstring of randperm "all permutations are equally
likely". Therefore I was expecting that the resulting histogram from
the script below was more or less 1 until N-n and then went to zero

N = 100;
n  = 10;
x=zeros(1e4,1);
for i=1:length(x);
  x(i) = min (randperm (N, n));
end
hist (x,1:100)
set (gca, 'xtick', 1:9:100)
axis tight

Even with n=2 it seems that small indexes are more likely than larger
ones (seems exponential). Am I misunderstanding something here or is
this a bug?