dillo-i18n-misc had a bug in handling entities like > in websites: on PPC and most probably other big-endian platforms, entities do not appear (it is an accidental double endianness conversion). It's fixed upstream in dillo-i18n-misc-20051207 or perhaps earlier (I don't know, it was me who submitted bug and patch to the maintainer of the patchset, but forgot when it was fixed); later I installed Gentoo and wondered why I still see my old bug ;) The bug can be fixed by either upgrading the version of the patchset (which I'd prefer) or by adding this small change to html.c (quoting my own mail to the patchset maintainer): > Given the entity &x1234;, isocode is 0x1234. Then html.c will do the following: > > 1167 guint ucs2le = GUINT_TO_LE((guint)isocode); > 1168 gchar wc[3] = {((gchar)ucs2le<<8)>>8, ucs2le>>8, 0}; > > On a little endian platform, GUINT_TO_LE does absolutely nothing and given (for > example) 0x1234 in isocode, this will calculate wc to {0x34, 0x12, 0}. > > However, on a big endian platform, given 0x1234 in isocode, this will result in > ucs2le being 0x34120000 and therefore wc will end up being {0x00120000, > 0x00341200} which collapses to {0, 0, 0} given 8bit chars. I think line 1167 > should be removed altogether since line 1168 does the actual conversion to > little endian. > > And I think the two shifts are not necessary, but don't hurt. I'd have written > instead: > > 1168 gchar wc[3] = {(gchar)isocode, (gchar)(isocode>>8), 0}; > > Now it works here and should still be equivalent on x86,
Sorry for the delay. I updated i18n-misc patchset to 20060104, so it should be fixed now. Thanks for the bug report.
