Andriy Gapon wrote: > I have discovered the following problem which at first appeared very > puzzling to me. I use uk_UA.KOI8-U locale with KOI8-U charset and, as I > discovered later, this chareset is not included into ICU standard data > library.
Are you sure about this? I think it is supported by ICU but it requires a different alias for the character set. I guess GTKG would need another alias translation table like there is already one for iconv. Though, you could be right, maybe ICU supports only the KOI8-R variant. > sent out. That was very mysterious to me. Then I noticed a warning(?) > message that ucnv_open() failed and only after that I started looking at > ICM stuff. It appears that in a situation like mine search text gets > "converted" to void somewhere, so no search was done. Well, yes. The code is definately wrong here. It ignores the error except for giving a warning and later when it comes to conversion it ignores all but expected errors. It should at least fall back to iconv(). > Please also note that search text I entered was pure ASCII. > And there was no such problem with any of the previous gtk-gnutella > versions I used (including 0.93.X). Your best bet is to disable ICU support at the moment. Matching of search results encoded in cyrillic will probably fail if characters have a different case but everything else should work fine. -- Christian
pgpgkqt40hU2f.pgp
Description: PGP signature
