Recent versions of gnupg, like 1.0.4-9, includes translated messages. Recent versions of gettext, like that from glibc 2.2, translates fetched messages to the used character set. If no character set has been choosen with setlocale(LC_CTYPE, ...), a plain US-ASCII is assumed. Any characters not in this set is replaced with question marks. This happens to the Swedish messages for gpg. The 1.0.4-9 SRPM includes a patch for this, gnupg-1.0.2-locale.patch (patch #0). But for some reason this patch isn't applied. Why not?
GnuPG depends on collation data for the C locale, so the patch would cause things to break. We kept it in the package for future reference. *** This bug has been marked as a duplicate of 16222 ***