QUOTE:

I tried to make some comparisons between the services in Leiden and
Paris-9. Some examples suggest me that the error may be a wrong indexing
of the colors in the gif files ?

See as an illustration the images img-leiden.gif and img-paris9.gif :
they come from the same exercise (optics/oefdescartes.fr), and we see
that in Leiden, the yellow area does not appear. However, the OEF source
does contain the code "frect , ... , yellow". One possibility would be
that the yellow color is changed to transparent during some conversion
of the GIF file. However this hypothesis does not hold : the background
is plain white in the file img-leiden.gif

We should check the sources of flydraw which were recently modified to
meet Gang's needs (DXF export) to try to discover the problem.
Unfortunately I have few time to do it precisely.

I think that you may try some attempts to compare the results of various
versions of Flydraw if you run it on more than one computer. Paste some
flydraw program in one file, for example sketch.txt, then launch the
command:

cat sketch.txt | /var/lib/wims/other/bin/flydraw > /tmp/sketch.gif

and open the file /tmp/sketch.gif to see flydraw's output.

I tried this command with a file sketch.txt containing the samples
visible in the help page for slibs (graphpaper/millimetre), and it works
for me.

Such a test may isolate the bug.