[Scilab-users] bug in fftw

[grivet]

Hi,
The bug that affected the dct transform in Scilab 5.4.0 is still present 
in Scilab 5.5.0. To evidence the bug,
please run the following lines:

x = 0:127;
xx = 0:255;
y = cos(2*%pi*x/32).*exp(-x/32);
yy = cos(2*%pi*xx/64).*exp(-xx/64);
w = dct(y);
z = fft(y);
//ww = dct(yy,"dct1");

where I define a damped sinusoid and attempt to compute an fft *AND* a 
dct from it.
The output is:

Warning !!!
Scilab has found a critical error (EXCEPTION_ACCESS_VIOLATION)
with "fftw" function.
Save your data and restart Scilab.

On the other hand, if I compute 'ww' instead of 'w', i.e. dct and fft on 
vectors of different lengths,
the problem disappears.
It would seem that one of the numerous fftw parameters is not reset 
between two calls.

JP Grivet