[00:32:01] bd808: Watching your Kibana4 tech talk. Related videos: What's new in Kibana 5 :P [00:32:06] Now I'm spoiled. [00:32:25] heh. kibana 5 does look much nicer [00:32:30] I'm not watching it yet though. I'm happy with K4 [00:32:39] I think we have to upgrade elasticsearch to get it [00:32:40] We'll see K5 sometime next year. [00:32:44] yup [00:32:45] Yeah [02:24:29] bd808: Hm.. I'm struggling to add a simple table panel to a Kibana dashboard. I'm trying to re-create the "slow-parse" dashboard. screenshot: https://phabricator.wikimedia.org/T98563#1577217 [02:25:02] On https://logstash.wikimedia.org/app/kibana#/dashboard/slow-parse I basically want to add a table that shows 10 event entries in a table with 'time' and 'title', sorted by 'time'. [02:25:28] It seems when I do it as a visualisation, I can't do it unless I add some sort of aggregate metric like "count", which doens't make sense. [02:25:53] As an event list I almost got it, except that it shows a timestamp column and doesn't sorts "naturally". Perhaps I need to make it a number field. [02:25:56] Not sure how to do that though. [02:36:07] hmmm... I can try to play with it in a bit [02:36:07] Krinkle: ^ [02:37:03] https://logstash.wikimedia.org/app/kibana#/discover/slow-parse-top-list [02:37:04] bd808: thanks :) [03:34:06] Krinkle: I can't figure out how to get rid of the aggrigate column either, but this is pretty close to the old table -- https://logstash.wikimedia.org/goto/754ca82da92845c4649220fee69f269a [15:45:17] tgr|away: when you have some spare cycles (ha!) can you point florian towards a less totalitarian solution for his autocomplete complaint? -- https://gerrit.wikimedia.org/r/#/c/308353 [15:46:07] bd808: DAS IST VERBOTEN [15:46:27] still too soon Reedy ;) [15:51:52] I wonder does anyone know the difference between pygments 1.* and 2.* [15:52:26] Since i know syntax highlight geshi uses the 2.* but on phabricator it is using 1.* so i am wondering will there be any benefits in upgrading to 2.*? [15:52:35] http://pygments.org/docs/changelog/ [15:52:56] Lots of new languages etc [15:53:24] probably not [15:53:53] Oh thanks [15:54:05] use of syntax highlighting in phab is limited to the set of languages that the wmf uses [15:54:18] which are already well-supported [15:55:03] Oh [15:55:49] phabricator uses pygments but can default to using a php implementation if pygments isen't available. [16:00:14] I guess this https://bitbucket.org/birkenfeld/pygments-main/commits/e1abdf91c1fe18b9ab313f3dc298d7025af5869c?at=2.0 would make it show it in utf-8? [19:26:00] anomie: where does OOUI handle autocomplete? [19:26:56] tgr: oojs/ui repo, looks like in php/widgets/TextInputWidget.php and src/widgets/TextInputWidget.js [19:28:03] ah, didn't realize we are including PHP classes from ooui [19:28:20] the external repo I mean [19:30:37] It's somewhat of a weird situation [19:30:44] Some are in core. Some are in vendor [19:33:46] anomie: what do you think, should we just not support autocomplete other than on/off or support but downgrade it to a boolean for OOUI and file a bug for them to fix it in the future? [19:35:00] tgr: MatmaRex might know if there are any plans for OOUI to support the fancy autocomplete. [19:35:40] hm? [19:36:10] what fancy autocomplete? [19:37:22] MatmaRex: https://html.spec.whatwg.org/multipage/forms.html#autofill [19:39:01] Also, https://gerrit.wikimedia.org/r/#/c/322138/ [19:40:18] that honestly reads like an output of a markov chain [19:40:43] are we planning to have users input billing addresses anywhere? [19:41:11] anomie: i don't think it was considered. it's probably "in scope". but i'm not really the primary maintainer right now. file a task :) [19:41:29] tgr: ^ [19:41:32] Special:WikimediaShop [20:31:49] TimStarling: does SecurePoll follow the "use REL_xx branch for MediaWiki 1.xx" model or the "use master for all supported versions, as indicated in the infobox" model? [20:45:31] well, I wasn't the last one to add features to it, that was anomie [20:46:50] tgr: Also, extension registration... So master won't work with 1.23 [20:48:04] TimStarling: you should probably pass the hot potato on https://www.mediawiki.org/wiki/Developers/Maintainers then :) [20:48:16] anomie: want to call it? [20:48:18] I'm sure anomie would love me for that [20:50:06] once master can run on the LTS version then I guess we can make it "use master" [20:50:06] tgr, TimStarling: I'll pass the buck to Huji per T147935 [20:50:07] T147935: Ownership request for SecurePoll - https://phabricator.wikimedia.org/T147935 [20:52:19] Jees. 1.23 till May 2017!? [22:19:43] tgr|away: anomie I guess for LDAP stuff... [22:19:54] It'd make more sense rewriting from the ground up ontop of AuthManager? [22:21:47] Reedy: Ideally, yeah. https://phabricator.wikimedia.org/diffusion/ELDA/browse/master/LdapPrimaryAuthenticationProvider.php;770c89e36733e0ee3a38dd0e01cafc4a5e22b328$33-43 [22:22:06] I was looking at it... And seeing some of the updates [22:22:11] Then the deprecations... [22:22:42] For wikitech... Someone is going to have to bite the bullet before we can rip it out of core :( [22:29:42] Reedy the fix for your task https://gerrit.wikimedia.org/r/#/c/322202/ :) [22:50:01] Reedy: what needs to be ripped out of core? [22:50:15] tgr: AuthPlugin etc eventually [22:51:15] it's only a matter of deprecation schedule at this point, isn't it? [22:51:31] yeah [22:51:42] But thinking ahead... And how these things usually work [22:51:44] People don't update [22:51:49] We want to remove as it's long overdue [22:51:54] And all the things have to be updated [22:52:07] you can just move it to LDAPAuth if no one is willing to rewrite that properly [22:52:25] haha [22:54:05] well thats what we did with ORM and EducationProgram... [22:54:32] lol [22:54:38] no, seriously, the authprovider for LDAPAuth subclasses AuthManagerAuthPluginPrimaryAuthenticationProvider, just merge it in, use some other global than $wgAuth to communicate with it (or maybe that's done already), core is clean, LDAPAuth is not worse off than it already is [22:54:51] How many other extensions use it? [22:55:08] * Reedy looks [22:55:38] LDAPAuth is the only one that has been updated for AuthManager (for some value of updated) and still uses is [22:55:49] Oh, of course CentralAuth still uses AuthPlugin [22:55:50] I'm sure there are dozens of old ones [22:56:05] 3 other extensions after ldap and CA [22:56:13] that's probably just compatibility code [22:57:06] there are probably lots of auth extensions not in gerrit though, it's a problem domain that often requires custom solutions [22:59:38] LDAPAuthentication is the most popular one [22:59:48] It's actually one of the most popular extensions period [23:00:16] https://www.mediawiki.org/wiki/Special:ExtensionDistributor [23:00:18] #2 [23:00:27] Yep [23:00:31] I was just looking at grafana [23:03:09] amazing [23:03:53] legoktm: I'm trying to tweak the grafana dashboard to omit versions from that top visualization that have 0 entries. [23:05:17] Ah, hide series with only zeros [23:05:18] Done [23:05:33] metrolook is the #2 most downloaded on https://www.mediawiki.org/wiki/Special:SkinDistributor [23:10:52] legoktm: https://grafana.wikimedia.org/dashboard/db/extension-distributor-downloads looks a little nicer, did some minor tweaks [23:11:07] (omit entries with only 0s, make the second row a little taller) [23:14:20] wow, VisualEditor is the most frequent download? [23:14:32] anecdotally, no one seems to be using it [23:15:07] at least when it's asked on wikitech or phab [23:18:43] tgr: Either that or James_F is just downloading it all the time to make it look popular ;-) [23:19:09] legoktm: Do you remember where we're stashing the $wgPingback data? I want to start watching it now that 1.28.x is going out and includes it [23:19:19] Also: half tempted to backport to 1.27LTS [23:19:42] that would be nice [23:19:58] ostriches isen't $wgPingback that supposed to make it possible to offer updates to wiki's [23:20:03] like an in built updater [23:20:13] Nope [23:20:14] to get more wiki users to update? [23:20:15] oh [23:20:37] Mainly it'll let us get stats on who's running what versions [23:20:43] Also: what DB backends are being used [23:20:50] ^ That second one is Important To Me [23:20:53] Isn't it eventlogging? [23:21:00] Yeah but I dunno how to view that data [23:21:10] ori: ^^^ [23:21:22] Oh [23:21:33] ostriches why is the db backend importent to you? [23:21:43] So we can drop support for shitty ones no one cares about [23:21:48] Ding ding ding ^^^ [23:22:04] So far, in my last ~8 years of committing, I've only managed to get rid of DB2. [23:22:14] and/or move them out of core, as a minimum [23:23:10] LOL [23:23:11] I was wondering about submitting a "get rid of DB abstraction" session for WikiDev but then decided somebody else's problem [23:23:30] ostriches Reedy one problem with that not everyone is going to enable that config [23:23:40] and epecially there was already a debate on mssql [23:23:43] some people are really angry when e.g. postgres breaks, but don't consider it their responsibility to maintain it [23:24:04] Eh, I wouldn't characterize the mssql discussion as a debate. [23:24:04] and the WMF is not serious about it (no CI etc) [23:24:15] ostriches but they were trying to remove it [23:24:22] I know, I was the one who tried. [23:24:30] Then like 2 people showed up and said NOWAY I CARE A LOT [23:24:37] And then promptly did nothing to work on it [23:24:59] Yeah [23:25:01] It's stupid [23:25:03] ostriches see https://phabricator.wikimedia.org/T113831 [23:27:06] You're telling me things I know all too well :) [23:27:13] LOL [23:27:19] ostriches mssql comes to linux [23:27:20] too [23:27:26] I don't see how that help [23:27:26] s [23:27:35] I wonder how hard it would be to extensionalize DB backends [23:27:53] Reedy well with new support for linux other users can use it now [23:28:00] i find mysql/marriadb better [23:28:03] Why would they actually want to? [23:28:18] I doint know [23:28:51] tgr: Problem is you need an autoloading mechanism for them [23:28:54] different people like using different things, some like postgres, some mysql and some mariadb and now the new sql server for linux from microsoft [23:29:01] You can very easily end up needing a DB connection before you hit LocalSettings [23:30:23] paladox: Any many platforms say you can only use one DBMS [23:30:39] Oh [23:30:51] Phabricator doesn't support Postgres for example ;-) [23:31:21] I mean, unless you're using a 100% abstracted ORM that handles *all* of that logic for you, playing "support multiple DBs" is always a game of whack a mole. [23:31:27] It's fine if you've dedicated maintainers [23:31:34] But if you've got people that only care about one, maybe 2 [23:31:35] Gerrit doesn't write any manual SQL, which is why it supports like 8 different backends. [23:31:49] The others will get neglected [23:32:18] Reedy: Indeed. I'd also make the argument that if there was an RfC to *introduce* new DB backends in 2016, they'd probably get shot down pretty damn quick [23:32:29] lol [23:32:36] Indeed [23:32:39] ostriches what about sqlite? [23:32:50] ostriches: But what about mongodb? It's webscale. WMF could use it [23:33:01] lol [23:33:11] paladox: SQLite is pretty trivial because it's almost identical to Mysql in syntax, and makes developer's lives easier. [23:33:23] oh [23:33:29] Schema changes suck a bit [23:33:45] Schema changes suck broadly in MW... [23:33:51] True [23:33:54] Remember that abstract-schema branch I had a long time ago? [23:33:58] * ostriches goes down memory lane [23:34:08] But even more so in sqlite when you've got to copy the rows and stuff [23:35:57] ostriches you may get away with killing oracle since it is experimental according to the mw docs [23:36:08] but postgres dosent have the experimental tag [23:36:13] not sure about mssql [23:37:02] toolserver was using postgres [23:37:06] Then got stuck on an old version [23:37:09] As upgrading was hard [23:37:13] oh [23:37:52] Oracles maintainer... Has stopped doing Oracle for his work... So doesn't care about it in MW now [23:37:59] we could make all the other db's experimental as in you wont get much support with those db;s [23:38:09] Well, you don't anyway [23:38:25] so postgres, mssql [23:38:32] oracle could be experimental [23:38:51] that implies it's being worked on [23:38:53] Awww, Oracle. I miss freakolowsky [23:39:07] When at best, it gets drive by fixes when people touch other stuff [23:39:24] ostriches: how can you get a DB connection before LocalSettings? DB user/pass is expected to be set up there [23:39:44] tgr: ...this...is true... [23:39:56] Config order would matter, a lot. [23:40:01] But yeah I suppose you could do it! [23:40:23] Reedy: Jure's $DAYJOB doesn't involve Oracle anymore? [23:40:38] Not when I last spoke to him... [23:41:00] Trying to think when/where I last spoke to him though [23:45:35] Just asked him on FB [23:47:15] Seems his $DAYJOB does have Oracle stuff... [23:48:31] * ostriches shrugs [23:48:38] Hey, it's quittin' time. Later folks [23:49:11] * paladox updating windows 10 [23:52:30] tgr: Given I get a "help me install VE" request 2–3 times a week, I'm unsurprised.