[00:58:21] Anyone know why the replag on s4 is high? https://dbtree.wikimedia.org/ reports 148237s on db1121 and 10537s on db2091 [12:44:35] AntiComposite: see https://www.irccloud.com/pastebin/vbXVugqc/ [12:45:27] marostegui seems to be doing some changes at that host, and db1121 is depooled [12:55:14] hej everybody, does anyone know of a way to pass error messages via the onVerifyFileUpload hook to the visualeditor? whatever error message I declare in the hook gets displayed in VE as 'This file did not pass file verification.' (this comes from UploadChunks.php:201) [12:56:11] i also tried other hooks to intercept the error but to no avail (ApiMain::onException, APIGetAllowedParams, APIAfterExecute) [15:45:03] Is it possible to obtain the list of articles which don't exsit on enwiki of folklore theme? [15:45:18] Query? [15:55:57] Tulsi: How is "don't exist" defined? [15:56:07] Tulsi, how is "folklore theme" defined? [15:58:49] No idea.. that's complicated :/ [15:59:19] lol [16:00:51] Heh... that basically translates to "Someone knows how to fix my car? But I cannot provide more info currently." :) [16:02:38] andre__, Folklore can be defined using categories or sub categories? [16:32:33] something strange on Pt.Wp. I am able to go to any other address on that project but there is a specific one on which I get an error message [16:32:39] this one: https://pt.wikipedia.org/w/index.php?title=Filipinas&diff=57250777&oldid=57158884 [16:33:03] Request from 2804:14d:72b3:add3:1003:ad19:c056:8f32 via cp1077.eqiad.wmnet, ATS/8.0.5 [16:33:03] Error: 504, Connection Timed Out at 2020-01-22 16:31:02 GMT [16:43:53] Weird [16:51:16] Is there any documentation for how the replica database (i.e. what's behind quarry) differs from the production db? [16:51:51] there's the script that creates the view... [16:53:38] I discovered that archive.ar_comment_id is always null; I assume that's something intentionally stripped out of the replica? [16:54:41] where can I find that script? [16:57:13] I'm not 100% sure if archive comments have necessarily been migrated [17:01:12] Hmm, they have for enwiki (first and last row of archive has it) [17:02:13] https://github.com/wikimedia/puppet/blob/production/modules/profile/files/labs/db/views/maintain-views.py [17:02:34] https://github.com/wikimedia/puppet/blob/production/modules/profile/templates/labs/db/views/maintain-views.yaml [17:03:15] RoySmith: Seems it's purposeful [17:03:16] https://github.com/wikimedia/puppet/blob/production/modules/profile/templates/labs/db/views/maintain-views.yaml#L306 [17:03:21] https://github.com/wikimedia/puppet/blob/production/modules/profile/templates/labs/db/views/maintain-views.yaml#L315 [17:03:54] OK, thanks, and thanks for the puppet link. [17:04:31] on a different topic.... [17:04:48] I think... [17:04:49] https://github.com/wikimedia/puppet/commit/91e1d8f93da49c5d52f5d961000b54e6a0d8618b [17:04:53] Someone just needs to make the changes to expose it [17:05:02] I've been working with the k8s python3.7 container [17:05:22] https://phabricator.wikimedia.org/T158166 [17:05:38] There's no ssh on it, so I can't run "git pull" and the like. [17:06:32] I've been doing all my git work on the bastion host, then shelling into the container to actually work on the code, because there's no python 3.7 on the bastion. [17:06:36] this is kind of awkward. [17:07:00] Is there a way to set up an environment that has everything all in one place? [21:53:19] Any idea why Abusefilter's aser_age reports 0 for old, registered users? [21:53:28] At least on Commons [21:53:44] *user_age [22:30:58] ankry: Presumably for people who got their accounts before the "new" logging in ~2010? [22:32:24] James_F: [[phab:T243469]] [22:32:25] T243469: user_age abusefilter variable reports 0 for logged in, established users - https://phabricator.wikimedia.org/T243469 [22:33:47] test for user_age is commonly used to identify not logged i users, so changing this behaviour is very bad [22:33:52] Sure. [22:33:56] But I don't think it's a change? [22:34:06] I think you've found one of the "fun" edge cases in MediaWiki. [22:34:11] Aka everything is terrible. [22:35:10] ankry: E.g. https://commons.wikimedia.org/wiki/Special:AbuseFilter/examine/log/249473 for me. [22:36:28] well, so you are also not logged in now from abusefilter's point of view [22:37:04] If people are writing bad AbuseFilter filters assuming that age=0 guarantees that the account is new, they should fix their bad filters. [22:37:41] (it would be better to check for "user" in `user_groups`) [22:37:51] Indeed, that's what you're meant to do, I think. [22:42:30] James_F: people ale alse checking if an account is newer than one year cor some operations [22:42:59] user_age < 3600*24*365 [22:43:02] ankry: Sorry, not 2010, 2008. Account creation time was added in https://phabricator.wikimedia.org/rMWe8215b27e423c14f7db68ddd75550e526aba80c2 in September 2008, so deployed in ~late 2008. [22:43:23] any replacement suggested for this? [22:43:26] So all accounts created before 2009 will just be logged. [22:44:24] ankry: `user_age < 3600*24*365 && user_age != 0` or whatever the syntax is. [22:51:49] James_F: and efficiency which is on the edge already will go down... [22:52:02] fortunately, this is not my problem [23:00:00] James_F: my commons account was created in Sep 2007 and has non-zero user_age; so the problem is somewhere more in the past [23:03:15] ankry: Some accounts were back-filled, I believe. [23:05:25] Looks like ankry's right, there was a recent change to the code that generates user_age [23:06:08] https://phabricator.wikimedia.org/diffusion/EABF/change/master/includes/AFComputedVariable.php;87459ec679e9fc990bc264df6a7b8824882217ce line 419 [23:07:17] the question is, did the change actually change anything. I don't know enough php or mediawiki internals to know [23:08:10] James_F: the behaviour *has* changed; JuTa had user_age=user_age in August; now has user_age=0 [23:08:30] had user_age=1565522918 [23:08:36] bad paste [23:09:23] Hmm, yes, that looks like it shouldn't have changed. [23:09:51] Ah, I see [23:10:29] Previously, a user with no registration time did (now) - Null, which gave (now) [23:10:40] In unix time, seconds since 1970 [23:10:47] Yeah. [23:10:52] Which is why that user age is ~50 years [23:11:23] Yeah, it was wrong before, but it's differently wrong now. [23:11:24] Now, there's an explicit check for no registration time, which instead of returning something that's functionally infinite, returns the opposite, 0. [23:11:36] So it's wrong in the complete other direction. [23:11:57] Yeah. [23:13:03] AntiComposite: Nice find. [23:16:09] (Fixing now.) [23:24:41] Thx James_F [23:24:54] and AntiComposite :) [23:25:02] Fix posted, will get a colleague to merge and then I'll cherry-pick into production.