If I'm not mistaken, Olivier Landemarre had identified that bug (which
was somewhere in Windom).
Well, currently it seems to be something in ldg_close( codec_pointer ); but who knows, maybe windom messed up something. If Olivier has some information, I welcome it of course :)