When using several connections with different charsets
(for example, UTF-8 and Latin-1), the converter dict is
not built correctly. It copies the dict, but not the
lists in the dict. Thus the .insert's further down in
the function init of connections.py insert into the
same list in all instances of the class, because it's
only referenced. The list then contains the almost same
converter function multiple times, but with different
charsets. The attached patch changes the code to copy
lists in the converter dict.
This happens with the latest release, 1.2.1, and is not
yet fixed in SVN. Running on FreeBSD, but this error is
not distribution specific.