[unicode] Syndication framework cannot handle unicode description

[bugs@…]

I have object with content attribute, where I have non-ascii data. For both cases (either specifying {{ obj.content }} in description template or by adding method

    def __unicode__(self):
        return smart_unicode(self.content)

), I got UnicodeDecodeError when trying to display feed:

UnicodeDecodeError at /feeds/wiki/
'ascii' codec can't decode byte 0xc3 in position 0: ordinal not in range(128)
Request Method: GET
Request URL: ​http://rpgpedia.cz/feeds/wiki/
Exception Type: UnicodeDecodeError
Exception Value: 'ascii' codec can't decode byte 0xc3 in position 0: ordinal not in range(128)
Exception Location: /usr/lib/python2.5/codecs.py in write, line 303

Local variables show object codecs is trying to decode:

u'\xdasp\u011bch zna\u010d\xed zd\xe1rn\xe9 zavr\u0161en\xed akce, kter\xe1 je p\u0159edm\u011btem testov\xe1n\xed. Je pot\u0159ebn\xfd zejm\xe9na tehdy, kdy\u017e se n\u011bkter\xe1 ((rp postava postava)) nebo jin\xfd element v ((rp rolova_hra rolov\xe9 h\u0159e)) sna\u017e\xed n\u011bco ud\u011blat, n\u011bco zd\xe1rn\u011b zavr\u0161it, nebo n\u011bjak\xfdm zp\u016fsobem zvr\xe1tit situaci ve sv\u016fj prosp\u011bch.\r\n\r\nNakl\xe1d\xe1n\xed s \xfasp\u011bchem z\xe1vis\xed od ((rp pravidla pravidel)) hry. V n\u011bkter\xfdch hr\xe1ch je d\u016fle\u017eit\xfd tak\xe9 po\u010det \xfasp\u011bch\u016f (pokud jich m\u016f\u017ee hr\xe1\u010d v testu dos\xe1hnout v\xedce), v jin\xfdch hr\xe1ch je podstatn\xe9 jenom to, jestli hr\xe1\u010d v ((rp test testu)) usp\u011bje, nebo ne.\r\n\r\nV prvn\xedm p\u0159\xedpad\u011b m\u016f\u017ee nav\xedc p\u0159i v\xfdsledku konfliktn\xed akce mezi dv\u011bma nebo v\xedce postavami (nebo elementy) b\xfdt rozhoduj\xedc\xed i po\u010det \xfasp\u011bch\u016f jednotliv\xfdch postav a ta s nejvy\u0161\u0161\xedm po\u010dtem \xfasp\u011bch\u016f pak v dan\xe9m konfliktu zpravidla v\xedt\u011bz\xed.\r\n\r\nV n\u011bkter\xfdch hr\xe1ch existuje t\xe9\u017e """tot\xe1ln\xed \xfasp\u011bch""" (jak\xe1si zes\xedlen\xe1 varianta \xfasp\u011bchu obvykle s \u0159\xe1dov\u011b ni\u017e\u0161\xed pravd\u011bpodobnost\xed) vedouc\xed zpravida k v\xfdkon\u016fm \u010di ud\xe1lostem, kter\xe9 by za norm\xe1ln\xedch okolnost\xed byly (t\xe9m\u011b\u0159) nemo\u017en\xe9.'

( = normal unicode string, which has no problem when encoding with s.encode('utf-8')