Hi Alexey,
On Mon, 22 Apr 2024, at 7:43 PM, a.lobanov@xxxxxxxxxxxx wrote:
Hello all.Looks like, Squat search engine for non-ascii texts is heavily broken in recent Cyrus. The logic of damage is completely unclear for me:- search for single cyrillic character returns nothing (while nearly all messages definitely contain this character)- search for longer random seguence of cyrillic characters returns lots of findings (while messages definitely cannot contain this non-existent word)- search for a real word returns both real and false findings.
I don't suppose you know of a version that it used to work correctly in?
Can anyone confirm?Cyrus IMAP 3.6.1 Debian 12, default encoding UTF-8. Client programs Thunderbird and Sogo.
3.6.4 came out a little while ago with a handful of fixes to some long-standing Squat bugs (among other things). The Squat bugs weren't specifically about non-ascii texts, but they were about index corruption and bad search results, so it's possible there was some overlap that's been fixed.
Having a look, there doesn't seem to be a Debian package for 3.6.4 yet (but maybe I'm not looking correctly)
Resolved by upgrade to Xapian search engine.
Good choice. The Xapian engine is actively maintained, whereas Squat has been neglected for a while...
Cheers,
ellie
