On Mon, 2009-04-27 at 18:28 +0100, Chris Young wrote: > On Mon, 27 Apr 2009 00:49:19 +0100, John-Mark Bell wrote: > > > In which case, either the iconv filter in libparserutils is doing > > something odd that doesn't work with the iconv() implementation you're > > using, or the iconv() implementation itself is broken in some way. > > > > It's almost impossible to tell with the information available, I'm > > afraid. > > I'm testing with it disabled, if it doesn't noticeably break any sites > then I'm not particulary bothered about investigating further.
It will break sites. It rather depends on how many users you have that require the ability to read Russian, Middle-Eastern, and CJK languages. > > Does the csdetect test still fail? That's certainly odd, but likely > > unrelated (as, unless I've forgotten how that works, iconv() isn't > > involved) > > Yes, it does: > > 1: Detected charset windows-1252 (2252) Source 1 Expected (0) > FAIL - mibenum == parserutils_charset_mibenum_from_name( expected, > strlen(expected)) at line 133 OK. There's two problems here. Firstly, it's detected Windows-1252 and not UTF-8. Secondly, it appears that there's no expected value. Is the test data in the correct format? It should be: #data <data goes here> #encoding UTF-8 J.
