[14:21:03] 10MediaWiki-Core-Team, 10MediaWiki-Page-editing, 7I18n, 5Patch-For-Review: Long edit comments get entirely removed instead of truncated (error in cutting multibyte chars?) - https://phabricator.wikimedia.org/T85700#1251861 (10MaxBioHazard) https://ru.wikipedia.org/w/index.php?title=Теперь_его_дозор_окончен... [14:21:18] 10MediaWiki-Core-Team, 10MediaWiki-Page-editing, 7I18n, 5Patch-For-Review: Long edit comments get entirely removed instead of truncated (error in cutting multibyte chars?) - https://phabricator.wikimedia.org/T85700#1251862 (10MaxBioHazard) 5Resolved>3Open [16:21:35] <^d> ApiQueryAllPages::run is slow :\ [16:22:00] <^d> even with a limit 6?? [16:26:04] ^d: Can you get the actual query? [16:26:10] <^d> Yep [16:26:28] <^d> I'll pastebin the explain [16:26:33] Which wiki, too [16:26:59] <^d> https://phabricator.wikimedia.org/P593 [16:27:16] <^d> Oh I ran explain against enwiki, lemme see where the error originated [16:28:42] <^d> ugh I can't find the wiki [16:29:26] ^d: if the user is Pywikibot-test, that means its our integration tests [16:29:47] https://github.com/wikimedia/pywikibot-core/blob/master/.travis.yml#L81 [16:29:47] <^d> Some of them don't have a user at all [16:29:50] :o [16:29:56] <^d> But the rest are the pywikibot-test user [16:30:19] <^d> They're all timing out at 60s [16:32:00] ^d: It's choosing a good index and not filesorting. The only bit that looks suspicious to me is the join with langlinks, if langlinks is mostly empty on that wiki, which would make it have to go through a lot of rows. [16:32:51] <^d> Ahhh, found it [16:32:57] <^d> It's fast on enwiki, sure. [16:32:59] Or, I suppose, if it's being stupid on the other wiki. [16:33:02] <^d> Run it on wikidatawiki [16:33:39] Not surprising that langlinks would be mostly empty there. [16:33:50] <^d> There's our worst case. [16:34:44] <^d> Only ~60k langlinks on wikidatawiki [16:36:53] And none in namespace 0, probably. [16:37:52] <^d> They aren't normalized into ns/title pairs [16:38:09] Since that's all their "Q" pages which are non-wikitext, IIRC. [16:38:16] <^d> Oh, where il_from in page_id, nvm [16:39:05] <^d> How the heck do we optimize for that? [16:40:27] No idea. Removing the straight join might work for wikidatawiki since langlinks is so small, but still wouldn't help the general case where other namespaces have lots of langlinks. Could add a "ll_from_namespace" column, I suppose. [16:40:58] <^d> That could work, but would require backfilling to be useful. [16:54:48] <^d> anomie: I'll file a task for this. I don't think there's a quick answer we'll figure out on a friday [16:55:46] <^d> YuviKTM: Is the wikidata test important? [16:57:14] ^d: kind of...we have wikidata specific code and it also checks to make sure we're not making any bad assumptions about wikitext content model. but if breaking stuff we can disable it [16:57:27] <^d> It's not breaking really, just timing out [16:59:36] well, the tests are failing https://travis-ci.org/wikimedia/pywikibot-core/jobs/60845284 :( [17:00:37] <^d> anomie: https://phabricator.wikimedia.org/T97797 filed [17:00:46] <^d> YuviKTM: Then perhaps disable for now, pending some fix for ^? [17:07:26] ^d: asking in #pywikibot [17:15:45] YuviKTM: are labvirt100{7..9} actual hosts? [17:15:50] i see them in hieradata but not in salt [17:23:55] AaronSchulz: yt? [17:40:40] Hey anomie, do we currently have any special password recommendations for bot accounts? It looks like we say if the account is shared, those users should have "non-trivial" passwords.. [17:41:11] csteipp: None that I know of. [17:42:09] csteipp: BTW, is there a PS6 coming on https://gerrit.wikimedia.org/r/#/c/206156/? Two things were said done that aren't in PS5. [17:42:23] anomie: Yep [17:42:53] anomie: I was just debating about the 8-char min lengths... if those should be + or - for any of those groups. [17:48:47] ori: wrong person ;) [18:06:04] bd808: https://github.com/wikimedia/mediawiki-extensions-CentralAuth/blob/master/includes/GlobalRename/GlobalRenameRequest.php#L432 do you know why we check listUnattached() twice? [18:07:53] legoktm: I bet that was an oversite when changing things during review. The check at 446 will never be reached [18:08:25] yeah, that's what I was thinking :) [18:08:39] do you want to submit a patch or should I? [18:08:50] I'd love to make a patch :) [18:27:49] ori: I dont think labvirt1007 exists. Andrew Bogott would know for sure. I know that the ganglia cluster graph is accurate [18:27:54] * yuvipanda makes way to office [18:44:19] AaronSchulz: https://phabricator.wikimedia.org/T97818 [21:14:10] Deskana: do you remember the phab ticket (or what I should search for) to find the discussion of a server side solution for face recognition like the mobile apps use internally? [21:15:41] Deskana: nm. found it -- https://phabricator.wikimedia.org/T88633