> On Sun, 23 Dec 2018 20:04:28 +0100 > Matus UHLAR - fantomas wrote: > > I have tried to create rule that will match names "ján" and > > "jano" (john and johnny in slovak languages). > > > > I have created rule: > > > > body LOCAL_JANO /\bJ[aá]no\b/i > > > > however, it does not match. > > > > Apparently the [á] does not match even when normalize_charset is set > > to '1'. > > > > any idea what can cause this?
On Sun, Dec 23, 2018 at 11:11:39PM +0000, RW wrote: > normalize_charset converts to UTF-8 but the tests are still done on > bytes, so á isn't a character, it's a string. You need (?:a|á) instead > of [aá].
yes, luckily variation of this works as expected...
On Mon, Dec 24, 2018 at 10:16:58AM +0200, Henrik K wrote:Nope, that makes no difference. If message is UTF-8, neither of them will match anyway. The only correct solution: (?:[aá]|\xc3\xa1) works for any message in latin1 or utf8, regardless of normalize_charset setting. Config contents are never converted to anything, you need to make sure regex contains raw bytes for both cases. perl -MEncode -e 'print unpack("H*", encode("UTF-8", "á"))' No one should use normalize_charset until things are possibly simplified for 4.0.0, it's a terrible mess. Also it was broken for html as per bug 7656.
On 24.12.18 10:18, Henrik K wrote:
Oh and SpamAssassin assumes config files are latin1/bytes, so do not use UTF-8 encoding on config files itself, it will only create more confusion.
i assumed normalize_charset to expect rules in utf8, otherwise it would be a little strange ... -- Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/ Warning: I wish NOT to receive e-mail advertising to this address. Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu. 2B|!2B, that's a question!
