[01:15:00] addshore: Got a couple fixes for the mw-docker-dev awaiting your review :) [01:15:17] Also got a Travis CI test working :) [01:25:44] https://travis-ci.org/Krinkle/mediawiki-docker-dev/builds/303315675 [16:36:15] bd808: Meeting? [18:12:02] https://www.mediawiki.org/w/index.php?diff=2622469&oldid=2612190&title=Extension:Scribunto&curid=86223 [18:12:07] anomie: ^ :) [18:12:11] greg-g: When would be a good day to set the comment schema migration flag to write-both on Beta Labs? Now, Monday, or wait until after Thanksgiving? Unless there's bugs, the only user-visible change would be that some things might have longer edit summaries (e.g. API edits or maybe some thing with auto-summaries that are currently truncated). [18:12:56] https://wikitech.wikimedia.org/wiki/Labs_labs_labs [18:13:28] Ok, "Beta Cluster" then. [18:13:43] anomie: on beta cluster it should be fine "whenever" as long as the code is all there/ready. Just test it and revert if it's not working :) [18:26:34] Anyone want to sanity-check https://gerrit.wikimedia.org/r/#/c/392082/ for me? [18:29:27] anomie: it should be in InitialiseSettings.php [18:29:42] ok, fine. [18:32:05] legoktm: New PS up [18:32:33] lgtm :) [19:08:55] legoktm: I found a bug already. Review https://gerrit.wikimedia.org/r/#/c/392088/ for me quick? [19:15:28] woot [19:16:40] anomie: CommentStore is 1.30+ right? should we backport this? [19:16:58] legoktm: Yeah, probably. [19:17:02] On the good side, check out https://en.wikipedia.beta.wmflabs.org/wiki/User:Anomie/Sandbox2?action=history [19:17:44] 🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥 [19:22:21] hm, doesn't cherry-pick cleanly [19:25:29] why not? [19:25:41] that code hasn't changed in years... [19:25:52] oh, duh [19:26:17] https://github.com/wikimedia/mediawiki-extensions-FlaggedRevs/commit/8804e7c7561029425147347e6092ca37b25c8e8b [19:30:52] legoktm: Huh, apparently the patch that needs fixing didn't make it into FlaggedRevs 1.30 after all. [19:31:06] It must still be using back-compat code. Which is probably fine. [19:40:20] Yay, unit tests are passing for the actor table change! https://gerrit.wikimedia.org/r/380669 [19:54:08] legoktm: I see usernames starting with lowercase in centralauth.globaluser - is that expected? [21:27:52] MaxSem: no...that should not happen [21:28:17] bd808: re https://phabricator.wikimedia.org/T96705 was the task just supposed to be extensiondistributor tarballs or MW core tarballs to? the dashboard is only the former [21:28:56] MaxSem: globaluser should always be the canonical form [21:29:33] legoktm: select gu_name from globaluser where gu_name>'d' order by gu_name limit 10; [21:30:09] we...have a problem :< [21:32:50] I have a script for invalid usernames somewhere [21:37:16] legoktm: I'll create a bug [21:38:07] thanks, I'm trying to figure out what I did 2 years ago for https://phabricator.wikimedia.org/T5507 [21:40:09] https://phabricator.wikimedia.org/T180851 [21:49:34] MaxSem: the accounts look like they're from 2013, so probably not a ongoing issue [21:52:05] probably not super urgent, I could look at this on monday maybe? I put some notes on the ticket, but I have some school projects I need to finish up this weekend [21:53:28] yeah, nothing's on fire [21:54:08] (as far as I can tell:P) [21:54:44] I checked gu_registration for those accounts and they're all 2013 except for ~5 in 2015 [21:59:35] Today I learned: When phabricator is being useless and not providing the "not-in" or "in-any" variant of the top-level result (e.g. when typing "wikidata" and first seeing 500+ other tags and even then still not showing "Not-In: Wikidata"), you can also use not(wikidata) to get it. [21:59:38] https://phabricator.wikimedia.org/typeahead/help/PhabricatorProjectLogicalDatasource/ [22:07:36] legoktm: you know how we were talking about code coverage time series? it looks like jenkins only has builds going back a month, there's no obvious way to get coverage statistics older than that [22:07:51] The API doesn't auto-capitalize the first letter when given a username. Am I right in thinking that this should be fixed? [22:09:21] Krinkle: helpful! [22:13:16] TimStarling: we could backfill the data I guess? go back a week in git, generate coverage, repeat until it can't generate coverage? [22:13:58] Niharika, which API? [22:17:32] yes\ [22:19:56] legoktm: oh. I guess I assumed your dashboard was both. That's was a crazy assumption though. [22:25:26] tgr: Revision api. [22:25:46] the API should output canonical usernames, if not, that's a bug [22:26:42] https://meta.wikimedia.org/w/api.php?action=query&list=globalallusers&origin=*&agufrom=d [22:26:53] Is that a bug? [22:27:23] that's list=globalallusers [22:27:43] And yes, that's the bug that MaxSem just filed [22:29:10] legoktm: My question was different - if given a username with a lowercase first letter - shouldn't the API auto-capitalize the first one before doing the query? [22:29:41] oh, you mean aguform=d to aguform=D [22:29:48] from* [22:30:16] I'm not sure, maybe? [22:30:26] Yes. [22:31:01] seems like a Dont Do That Then thing? [22:31:57] might have surprising effects with non-ASCII characters, not sure if the lowercase and uppercase versions are always in the same order [22:32:09] We do it for PrefixSearch [22:32:19] then again, not doing it would definitely have a surpising effect too [22:32:19] tgr: Well, this is a tool which asks the user to input a few usernames. Sure we can do the capitalizing in the tool but it doesn't hurt to do it with the API once and for all. [22:32:20] seems reasonable enough to do listusers and globalallusers [22:32:32] Alternative is to require the client caller to know about these semantics, which is also wrong [22:37:44] Niharika, sounds like you'd want to use globaluserinfo in that case, not globalallusers? [22:38:34] Ah, sorry, I misunderstood. I assumed all usernames start uppercase, and this would be part of normalisation (like with other titles). [22:38:35] but yeah, asking for list=globalallusers&agufrom=foo&agulimit=1 and getting Bar back because it gets sorted before Foo seems stupid [22:38:46] But it seems there actually *are* entries that start with a lower case entry [22:38:51] I did not expect that to be the case? [22:38:51] Krinkle: all usernames do start uppercase. the lower case ones are bugs [22:39:23] User::getCanonicalName() uppercases so names just need to be sent through that [22:40:09] I'll poke davidwbarratt because it was actually him who ran into the problem in the first place. David - see tgr's comment above about using globaluserinfo. [22:42:39] tgr: "asking for list=globalallusers&agufrom=foo&agulimit=1 and getting Bar back because it gets sorted before Foo seems stupid"- I don't understand. Can't we turn foo -> Foo before doing the query? [22:43:57] Niharika: I meant, we should do so, because otherwise that happens [22:44:27] Yeah. [22:44:47] wanna write a patch for it? [22:44:52] Yup! [23:19:46] tgr uhh, we use globalallusers because the user is typing in a username and getting results (i.e. an autofill) [23:20:19] tgr so we can either fix the API to capatizlie the first letter, or I can fix the app, but if every app needs to do this, it makes more sense to me that the API should do it for you [23:20:58] oh OK, didn't think of that use case [23:22:13] tgr I mean unless there's a better endpoint for that, but this seemed like the best one [23:22:24] tgr Niharika should I create a ticket? [23:22:57] davidwbarratt: Yup. [23:23:35] kk [23:28:37] Niharika tgr https://phabricator.wikimedia.org/T180858 [23:29:03] a patch, rather :) [23:31:59] Meh, setting up CA without vagrant is such a pain. [23:50:32] if the patch is simple enough you can just write it blind and say so in the gerrit comments [23:50:57] the reviewer will probably have a working instance anyway [23:51:00] Does anyone know the colours to http://planet-hotdog.wmflabs.org/Wikimedia_Community_Logo.svg ? [23:51:02] please [23:51:08] (why not set it up with vagrant though?) [23:54:31] My vagrant gets "kernel soft-locked" in minutes after I start it. I was forced to do a manual install after years of being a loyal mwvagrant customer. [23:55:30] Yeah, I'll submit it blind. I wanted to ensure it doesn't screw up with non-latin usernames.