2D fft

[grivet]

Hello,

I have run into a problem using fft for nxn matrices. Here is a simple 
example:

x = linspace(0,10,512);
y = exp(-x/2).*cos(5*%pi*x);
m = y.*.y';
n = fft(m);
nr = real(n)

When I plot any line or column of n, I get a nice damped sinusoid; in 
other words, fft returned the original data.
The same happens if I do n = fft2(m).
Some bugs were reported for Scilab 5.1 and 5.2, but not for 5.3.2, which 
I use under WinXP.

Thanks for any suggestion or workaround,
JP Grivet