Hi all,
Dilyan kindly refreshed my memory about the search engines off-list. As I currently understand it, the state of things is:
* The "sphinx" search engine was removed a while ago; the "squat" engine was not.
* The "squat" search engine was broken in 3.0 and maybe 3.2 as well. I believe it to be fixed as of 3.4, so you _should_ be able to keep using squat when upgrading to 3.4. If you're on 3.0 or 3.2 and squat is broken, upgrading to 3.4 _should_ fix it.
* BUT, it sounds it's not working for people who've upgraded to 3.4.
* We only have the barest test coverage of "squat" search, but the tests we do have pass.
* It might be that there is still something broken about squat in 3.4. If so, it would be good to figure out what, fix it, and make a regression test so we don't break it again.
* It could also just be that there is some step missing from the upgrade process, and that figuring out what this step is and doing it will resolve the problem. If we figure out such a step, I'll add it to the upgrade documentation.
* You could switch to the "xapian" engine instead, like I previously advised. But unlike my previous advice, you _shouldn't have to_.
* Or you can disable search indexing by configuring "search_engine: none", but you _shouldn't have to_.
On Wed, 26 Jan 2022, at 9:20 AM, Chis Walker wrote:
For the past few days (mostly in the background) I've been looking to see if this incident already exists. Personally, I don't like duplicate issues so I'm trying to avoid reporting a duplicate to the Cyrus IMAP team.
I think if you can't find a duplicate within a reasonable amount of searching (say, 10-15 minutes) it won't hurt to just create a new issue. If we figure out it's a duplicate later, we can link them and close one; if we don't, then it never mattered anyway.
I also found this, https://github.com/cyrusimap/cyrus-imapd/issues/1614, which kind of looks like what I was experiencing. Looking at the labels for this it's a P3. I don't know about the Cyrus IMAP team but I know on my team a severity 3 or higher classification is basically purgatory unless someone really gets the hankering for some bug fixin' in their free time.
For whatever it's worth, those "Px" labels are from some historical tracking system that we used prior to Github. I think not even the most recent historical tracking system -- they predate my involvement, anyway. They're there because they were in the import, but we don't really use them for anything.
I can't tell whether that issue is the same thing as is being discussed here or not, but if you want to bump it up and use it for the current issue, please feel free to do so.
Cheers,
ellie
