[01:06:45] anomie_: yay! [01:08:29] legoktm: https://gerrit.wikimedia.org/r/#/c/204428/ easy [01:08:33] 6MediaWiki-API-Team, 10MediaWiki-API, 5Patch-For-Review: list=logevents tries to return both the creating user's and the created user's ids as "userid" - https://phabricator.wikimedia.org/T73020#1211166 (10Jdforrester-WMF) [01:09:07] 6MediaWiki-API-Team, 10MediaWiki-API, 5MW-1.25-release, 5Patch-For-Review: Clean up ApiResult and ApiFormatXml, create new formatversion - https://phabricator.wikimedia.org/T76728#819003 (10Jdforrester-WMF) [01:15:34] blazecat: LocalRenameUserJob: 0 queued; 2 claimed (0 active, 2 abandoned); 0 delayed <-- is there a way I can get rid of those two jobs? they're just going to keep failing [01:15:38] (on testwiki) [01:16:02] abandoned jobs never get retried [01:16:09] that disappear after 7 days [01:16:13] *they [01:16:16] ok [01:16:22] they were already retried a bunch of times [02:23:02] Keegan: https://test.wikidata.org/w/index.php?title=User_talk%3ALego-test~testwikidatawiki&diff=12542&oldid=12507 [02:23:52] All righty [02:30:39] Keegan: ok, I'm satisfied with everything on the testwikis...ready to start on aawiki? [02:32:09] I'm ready, legoktm [02:32:19] * Keegan fires up the email to wikimedia-l [02:33:24] legoktm@terbium:~$ ./DORENAMES.sh aawiki [02:35:55] https://aa.wikipedia.org/wiki/Special:RecentChanges [02:38:33] Keegan: ok, aawiki should be all done [02:38:53] Okay, sending email then [02:43:21] <^d> anomie_: All the tag-related stuff finally merged to rel1_25 [02:43:43] <^d> no thanks to jenkins, as usual [02:45:28] I should probably take the time to read that tag email. [02:46:21] (User rename log); 02:45 . . Maintenance script (Talk) renamed user Spam cleanup script (0 edits) to Spam cleanup script~aawikibooks ‎(SUL finalization) [02:46:26] was that a system account? [02:47:14] there's also (User rename log); 02:45 . . Maintenance script (Talk) renamed user MediaWiki spam cleanup (4 edits) to MediaWiki spam cleanup~aawikibooks ‎(SUL finalization) [02:53:10] TimStarling: ^ do you know? [02:54:24] probably maintenance/cleanupSpam.php [02:55:35] aha [02:55:43] that's not in $wgReservedUsernames [02:57:11] it probably should be [02:57:25] TimStarling: https://gerrit.wikimedia.org/r/204442 [02:59:09] thanks [02:59:54] wtf is ChangeTags.php not in /changetags? [03:01:50] move it? [03:35:49] legoktm: https://gerrit.wikimedia.org/r/#/c/204273/ also simple-ish [03:40:32] OK. this is it guys. We need to stop writing code for a month and start cleaning up our tests. [03:46:25] mwcore went up from 8 minutes on Zend to timing out after 30 minutes with 60% progress in the last 1-2 months. [03:47:12] it also switched from sqlite --> mysql and real hardware --> labs vms [03:47:52] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth, 10SUL-Finalization: Code review CentralAuth's GlobalUserMerge and related UserMerge code before SUL finalization mass usage - https://phabricator.wikimedia.org/T961#1211375 (10tstarling) a:5tstarling>3None >>! In T961#1208704, @Keegan wrote: > 1. A... [03:48:12] We already had various jobs running in labs and those slowed down the same way. And other jobs migrated to labs did not slow down noticably before/after the switch. [03:48:18] legoktm: I'm going afk for a bit but you can message me on hangouts or text or whatever [03:48:46] Keegan: ok, I'm waiting for jenkins right now so I can deploy the MediaWiki spam cleanup fix an then am going to have dinner [03:48:58] We also boosted test capacity to now dedicate a VM CPU to one test run (4 workers per quadcore vm now, instead of 6 concurrent workers previously) [03:49:15] Sounds good [03:52:07] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth, 10SUL-Finalization: Code review CentralAuth's GlobalUserMerge and related UserMerge code before SUL finalization mass usage - https://phabricator.wikimedia.org/T961#1211379 (10Keegan) I don't recall the circumstances at this point, I just recall talkin... [04:07:43] * legoktm is off for dinner, nothing should be running right now [04:09:43] Krinkle, any particular offenders re. tests? [04:10:02] Krenair: Not that I can see. Overall the database usage is just off the charta. [04:10:04] charts [04:10:30] Way too much uses the database at all. And most everything that uses the database overuses it (too much setup/teardown going on). [04:10:44] Perhaps some gotcha's with regardless to in-memory tables. [04:11:05] There are some comments in the code that say those are only supported in MySQL if used within the same query. [04:11:18] Maybe TimStarling knows more about that. [04:24:58] Also half the time it takes 4 minutes to create the mysql database tables during the installer. [04:25:06] Whereas other times it takes 5 seconds. [04:25:36] Something is off and I've got no clue. [04:26:43] shared filesystem? [04:27:52] I guess not, it is on a partition called vd-second--local--disk [04:28:00] use tmpfs? [04:28:19] actually I am not looking at the right server, am I? [04:29:25] did anything ever happen with that suggestion of mine to skip teardown if no write queries are done? [04:36:32] ori: Mysql on tmpfs? [04:36:52] TimStarling: This is the first I hear about that. Sounds interesting. [04:37:00] along the same lines, mentions libeatmydata, a library that you LD_PRELOAD to disable fsync() [04:37:35] TimStarling: It does use a second disk indeed. what server're you looking at :) [04:37:56] the test case ("a simple loop in the not at all exciting mysqltest language that will create 30,000 identical tables, insert a row into each of them and then drop them") went from timing out after hitting the 15m limit to 77 seconds [04:38:02] The integration-slave's now use an m1.large instance, with 4 CPU, and limited to only 4 concurrent jobs per slave. [04:38:12] tmpfs is used for all TMPDIR [04:38:48] each job gets a fresh TMPDIR subdir assigned to TMPDIR, and torn down afterwards. [04:38:53] 'eatmydata', packaged for debian [04:39:31] Krinkle: that's TMPDIR; but where is mysql's database files? [04:39:44] ori: I guess in /var/lib/mysql [04:39:46] Right? [04:39:56] Maybe we can disable disk syncing for mysql? [04:40:02] I don't know if that's an option and whether that's faster. [04:40:05] yes, /var/lib [04:40:09] I hear mixed reviews about people using memory tables [04:40:29] And some stuff about whether that's supported for cases like ours where the tables are used in multiple queries. [04:40:37] best to use innodb if you want to have correct locking behaviour [04:41:16] but mysql/innodb on tmpfs should be possible [04:41:22] That's the default I think (innodb). [04:41:34] Yeah, that's what it uses. [04:41:39] yeah I think TimStarling was thinking out loud that MEMORY storage engine would be simpler and then rejecting that idea [04:42:12] TimStarling: And you reckon that should improve performance? [04:42:32] I'm curious what's causing the installer to get stalled for 4-5 minutes just creating the tables [04:42:37] well, it did when we moved the previous sqlite setup to tmpfs [04:42:54] Yeah, but I don't know how mysql works. It doesn't write to disk for everything, right? [04:43:03] Not synchronously at least I assume [04:43:14] if that's reproducible then we could have a look at what is going on while it is waiting [04:43:17] Somehow in my crazy mind I actually thought mysql would make things every so slightly faster [04:43:42] sqlite on tmpfs versus mysql on spinning rust? [04:43:44] It's intermittent unfortunately. Though I'll try it in a loop to see if I can trigger it. [04:43:55] TimStarling: Well, if you put it that way :) [04:44:11] sqlite is a lot faster [04:44:11] I just thought that mysql does more stuff in memory and writes to disk from time to time. [04:45:00] libeatmydata looks worth a shot as well, IMO [04:45:20] I didn't quite get where libeatmydata would come in the picture. [04:46:37] when you tell the OS to write data to disk, it usually doesn't [04:46:40] It will still write to disk, but without failure guards? [04:47:10] it just acknowledges your request and buffers the data in memory [04:47:15] Right [04:47:21] And reads can get it from that memory? [04:47:22] this lets it batch disk access [04:47:46] fsync() is a way of telling the OS, "No, I really mean it, go write this to disk" [04:47:57] Right. Like waiting for slaves to catch up. [04:48:24] MySQL uses this a lot to protect data integrity [04:48:34] you don't want a whole bunch of data just gone if it happens to crash [04:48:35] tmpfs is actually just a few lines of code on top of the kernel's caching system [04:48:41] basically a kernel cache with no backend [04:49:15] so you might expect libeatmydata to have similar performance to tmpfs if it does its job properly [04:49:25] but I suspect the filesystem will also sync occasionally [04:49:37] libeatmydata makes those fsyncs a no-op, so in essence it's up to the OS again when to sync [04:49:38] Ah right. that makes more sense [04:49:51] the benchmarks here are promising: http://thelinuxjedi.blogspot.com/2011/05/libeatmydata-feed-me-seymour.html [04:50:02] Well need to increase tmpfs a little as its currently limited 512M [04:50:13] which may not be enough to host 4 tmpdirs and 4 databases [04:50:23] Though actually.. it should suffice. [04:50:33] with libeatmydata you might not need to use tmpfs [04:50:53] since you'd be hitting the kernel cache anyway whenever you were reading/writing from disk [04:51:12] ori: Would libeatmydata affect all users/procecces? [04:51:18] no [04:51:36] LD_PRELOAD=/usr/lib/libeatmydata.so mysqld [04:51:39] only affects mysqld [04:51:48] s/mysqld/$foo executable/ [04:51:49] Right, so we'll need to change msyql deamon in puppet [04:52:12] or hand-edit the init.d file that actuallly invokes it [04:52:48] TimStarling: OK. I'm able to hit the slow db creation even on an idle slave. [04:52:53] This might be usable for you. [04:53:02] what slave? [04:53:13] I'm on integration-slave-trusty-1012.eqiad.wmflabs [04:53:23] sudo -iu jenkins-deploy [04:53:28] cd /mnt/jenkins-workspace/workspace/mediawiki-core-qunit/src [04:53:43] export WORKSPACE=$PWD; export BUILD_TAG=debug; export EXECUTOR_NUMBER=999 [04:53:53] Then I alternate between these two [04:53:56] ; /srv/deployment/integration/slave-scripts/bin/mw-install-mysql.sh [04:54:02] ; /srv/deployment/integration/slave-scripts/bin/mw-teardown-mysql.sh [04:54:44] the scripts derive credentials and location of install.php from those env variables [04:56:40] ori: TimStarling: Thanks a ton for thinking along here. unfortunately, I'm really at the end of my day here (5AM). will catch up tomorrow. Feel free to keep looking if you can. All help is immensely appreciated. This is one of many slow downs lately. [04:56:43] mw-install-mysql.sh is I/O limited [04:57:12] bd808: does the sul rename warning do anything for API logins? [04:57:31] I find it usually stalls around "Creating tables". Sometimes it goes through in under 3 seconds. Other times it stalls for a minute or more. [04:57:34] bd808: just curious because I just saw a bot get renamed [04:57:37] takes about 20-30s [04:58:14] On Jenkins it takes either 1-5s or 2m30s-5m25s past 24 hours [04:59:11] I guess it has to write the schema to disk. That's not something it's willing to defer to later. [05:00:00] Hm.. would you say it's better to overload /var/lib/mysql with an extra tmpfs mount, or configure mysql to use a different directory? [05:00:52] probably better to configure mysql to use a different directory, if it's just as easy [05:01:38] It does mean we'll need to add some hooks in the puppet config to allow overriding for contint slaves. I think right now it's an indirect dependency a few levels deep from mediawiki::packages or something like that [05:02:26] mounting is ok too [05:02:42] I'm kind of scared to add mounts to existing directories :D [05:03:15] probably reboot mysql and it will auto create anything it needs? [05:04:12] no, I think you will have to manually run mysql_install_db while it is shut down [05:05:24] i.e. stop mysql, mount tmpfs, run mysql_install_db, start mysql [05:07:11] OK. noted. Will give it a try tomorrow. [05:07:53] TimStarling: one other thing while I have you. I noticed a huge difference between Zend/Precise and HHVM/Trusty. It seems the latter handles MySql much better. [05:08:04] Not sure whether to attribute that to hhvm or trusty. [05:08:33] Specifically for database tests and other interactions with mysql. [05:09:17] When we switched to mysql, mediawiki phpunit went from 6 to 8 minutes on hhvm/trusty, but from 8 to 15 on zend/precise [05:09:46] It varies but those are my rough average measures [05:10:31] mysterious [05:15:03] hmm, first entry on https://advisory.wikimedia.org/wiki/Special:GlobalUsers is broken [05:17:33] TimStarling: OK. I'll let you know how it works out. Thanks a bunch. [05:39:54] Keegan|Away: at the current speed, it took 4-5 minutes for 1,968 renames (afwiki) [06:32:22] legoktm: I... Don't know if it would work for app logins or not. I can try to trace the code and figure that out. I know I didn't consider api login when I wrote the patch though [06:50:24] legoktm: at that speed, that's 24.333 hours to complete the process >.> [06:50:57] * Keegan does the math again [06:53:22] Ah, 24x24=about a half a million accounts per day [06:53:31] So three days, at current speed [06:53:38] The optimal you suggested [06:55:06] which matches six days at 2,\.8 mil [06:55:11] 2.8 [06:55:55] I'm satisfied with that [10:07:19] 10MediaWiki-Core-Team, 15User-Bd808-Test: [draft] Performance Roadmap April - June 2015 (Q4 2014/2015) - https://phabricator.wikimedia.org/T93845#1211862 (10Qgil) [14:13:45] aude: Any tips on running your unit tests in mediawiki-vagrant? [14:14:27] anomie: I think with --group Wikibase or --group WikibaseAPI [14:14:50] and on the repo [14:14:56] jzerebecki: do you know? [14:15:18] i think php phpunit.php --wiki wikidatawiki --group Wikibase [14:15:49] aude: they don't work for me on vagrant [14:15:52] oh [14:16:03] thought i had them work at one point [14:16:43] there is probably something overriding the database settings [14:16:48] ok [14:16:53] maybe i hacked that in [14:17:55] maybe creating that dababase and the account in mysql [14:40:11] 6MediaWiki-API-Team, 10MediaWiki-API: Update ApiTag for I7b37295e - https://phabricator.wikimedia.org/T96251#1212190 (10Anomie) 3NEW a:3Anomie [14:40:55] 6MediaWiki-API-Team, 10MediaWiki-API: ApiTag doesn't output errors from the status - https://phabricator.wikimedia.org/T96252#1212199 (10Anomie) 3NEW a:3Anomie [14:50:11] bd808: mediawiki-vagrant won't provision for me, seems to be missing utfnormal from its "vendor" directory. :( [15:01:49] anomie: I've seen that before. Try 'vagrant git-update' and then 'vagrant provision'. Not sure if we have a bug filed for that yet [15:02:39] bd808: I wonder if it's because vagrant git-update was erroring out trying to update the MW checkouts (due to local changes) and never getting to updating composer? [15:03:23] Hmmm maybe? [15:03:49] "This time for sure!" [15:04:50] bd808: Seems to be working now, so I'm guessing that was it. [15:05:23] Think it is worth a bug report? [15:06:36] Not sure. Technically it was user error. [15:14:49] hmmm, my wikidata vagrant wiki has become a client :/ [15:32:52] https://meta.wikimedia.org/w/index.php?title=Special%3ALog&type=gblrename&user=&page=&oldname=Kajnorberg&year=&month=-1&tagfilter wha [15:41:29] I don't see how that could have happened [15:50:27] https://meta.wikimedia.org/w/index.php?title=Special%3ALog&type=gblrename&page=Special%3ACentralAuth%2FMdrnpndr [15:50:55] https://meta.wikimedia.org/w/index.php?title=Special%3ALog&type=gblrename&user=&page=&oldname=Mdrnpndr&year=&month=-1&tagfilter= [15:51:09] weird [15:51:19] also Pndrmdrn [15:51:49] who knows... [15:57:31] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth: Possible to globally rename the same user twice (race condition) - https://phabricator.wikimedia.org/T96267#1212535 (10Legoktm) p:5Triage>3High [15:58:05] something very weird going on here legoktm [15:58:34] Krenair: for which one? Mdrnpndr or Kajnorberg? [15:58:57] nvm, just labs replication lag [15:59:22] (log table rows for the Kajnorberg log entries are missing from labs but present in prod) [16:00:35] latest timestamp in the revision table for metawiki_p is 20150416022943 [16:03:31] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth: Possible to globally rename the same user twice (race condition) - https://phabricator.wikimedia.org/T96267#1212573 (10Legoktm) I've manually deleted the bad renameuser_status row and unlocked the account. [16:06:42] ^d: https://gerrit.wikimedia.org/r/#/c/204449/ [16:09:40] legoktm: https://gerrit.wikimedia.org/r/#/c/204451/ [16:09:49] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth: Possible to globally rename the same user twice (race condition) - https://phabricator.wikimedia.org/T96267#1212608 (10Jianhui67) Well, perhaps both Stryn and I were coincidentally on the same page and we were going to approve the ticket. Stryn was sligh... [16:12:36] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth: Possible to globally rename the same user twice (race condition) - https://phabricator.wikimedia.org/T96267#1212615 (10Legoktm) >>! In T96267#1212608, @Jianhui67 wrote: > Well, perhaps both Stryn and I were coincidentally on the same page and we were goi... [16:17:22] is there an easy way to copy/move files from fluorine to terbium? right now I'm downloading them locally and uploading them back to terbium with scp [16:20:34] legoktm: if you forward your agent to one or the other you can scp directly between them [16:20:45] isn't agent forwarding disabled now? [16:21:07] I don't think that has been merged yet, but I could be wrong [16:22:27] nope it hasn't. thanks :) [16:25:02] <^d> Soon, hopefully ;-) [16:25:19] <^d> Andrew asked to hold off for a tad, he needs it for some ongoing labs migration stuff [16:27:53] uhmmm, there are 8 MassMessageServerSideJob's queued on akwiktionary... [16:27:53] is there a way I can look at the parameters of those jobs? [16:28:31] --list [16:28:49] uhoh [16:28:50] MassMessageServerSideJob Odwumanyɛni_nkɔmmɔbea:Tehut subject=Your account will be renamed message=string(1411) title=User talk:Tehut (uuid=cbefbfc59af84b01a482e2b4dd9e8659) status=unclaimed [16:30:40] blazecat: can I use runJobs.php to force run the queued jobs ^ ? [16:30:52] "force run"? [16:31:06] if you them to run now, yes [16:31:12] nothing should explode [16:31:13] blazecat: those jobs have been queued for ~2 weeks? [16:33:21] legoktm: You can use the -3 option to scp, which automates the download locally then upload. [16:35:03] anomie: oh that's perfect, thanks :D [17:15:11] 6MediaWiki-API-Team, 10MassMessage, 5Patch-For-Review, 7Wikimedia-log-errors: Invalid parameter for message "logentry-massmessage-failure" - https://phabricator.wikimedia.org/T93110#1212839 (10Krenair) 5Resolved>3Open That commit was included in 1.26wmf1, but... ```61 Invalid parameter for message "log... [17:38:35] 6MediaWiki-API-Team, 10Analytics, 10MediaWiki-Authentication-and-authorization: Create dashboard to track key authentication metrics before, during and after AuthManager rollout - https://phabricator.wikimedia.org/T91701#1212974 (10Tgr) * Successful logins via Special:UserLogin - `LoginAuthenticateAudit` hoo... [17:40:37] Krinkle: https://gerrit.wikimedia.org/r/#/c/204450/ [17:43:41] good idea [17:45:41] ^d: any chance you could look at https://gerrit.wikimedia.org/r/#/c/202589/6 ? [17:46:30] <^d> I don't have the cycles to give it a good review [17:51:37] (that was in the wrong channel) [17:55:36] 6MediaWiki-API-Team, 10MediaWiki-extensions-OAuth, 5Patch-For-Review: OAuth permission screen needs redesign for better usability and comprehension - https://phabricator.wikimedia.org/T75062#1213120 (10Nirzar) Based on the feedback on the first iteration here are improved screens for oAuth - Not inside a... [17:57:47] ^d: ok, here is a simpler patch https://gerrit.wikimedia.org/r/#/c/204553/1 :) [17:58:40] <^d> much simpler :p [17:59:02] csteipp: any time to look at https://gerrit.wikimedia.org/r/#/c/200325/ this week? [17:59:42] blazecat: Not yet... let me finish this security bug and I'll take a look [18:07:34] tgr: Do you want to take over (or throw out and start over) the things I started on in https://gerrit.wikimedia.org/r/#/c/204209/ ? [18:08:00] I was trying to keep it as dumb and simple as possible to start but that may be the wrong direction [18:35:01] Krenair: hmm, do we know what specific log entries are causing it? [18:35:29] what do you mean? [18:36:22] like, are there bad log entries in the logging table and people are trying to view it? or are we actively creating new bad ones [18:38:16] bd808: TBH it feels a bit like we are trying to sneak in something that should have an RfC and proper discussion [18:39:46] tgr: that wouldn't hurt for sure. The "MVP" for hooking to EL in a soft way is pretty small, but a "real" event system is certainly RfC worthy [18:40:36] I threw up the patch to get feedback from you and Ori. I'm not wedded to anything there [18:41:10] no new global functions please :( [18:41:46] that said I'm probably underrating the value of keeping it simple because I did not work much on core and it's easy to come with pretty architecture when you work on a feature in a team of three [18:42:49] so I believe I should trust your instinct there [18:55:37] aude, jzerebecki: FYI, I managed to get the API tests working in mediawiki-vagrant by disabling all roles besides "wikidata", hacking /vagrant/settings.d/wikis/CommonSettings.php to call MWMultiVersion::initializeForMaintenance() instead of MWMultiVersion::initializeForWiki() if MW_DB is defined, and then doing ``php /vagrant/mediawiki/tests/phpunit/phpunit.php --wiki wikidatawiki --group WikibaseAPI /vagrant/mediawiki/extensions/WikidataBuildReso [18:55:38] urces/extensions/Wikibase`` [18:56:28] anomie: thx [19:10:57] legoktm, oh I understand [19:10:59] I have no idea [19:23:22] blazecat: seems weird that the jobs only got stuck on wiktionaries... http://fpaste.org/211980/2658142/raw/ [19:30:07] annoying that labswiki can't connect to the DB, but whatever [19:36:19] legoktm: are you sure it just didn't get to them yet? [19:37:11] 6MediaWiki-API-Team, 10MediaWiki-extensions-OAuth, 5Patch-For-Review: OAuth permission screen needs redesign for better usability and comprehension - https://phabricator.wikimedia.org/T75062#1213479 (10Anomie) >>! In T75062#1213120, @Nirzar wrote: > {F115139} This design seems to go against the requirements... [19:37:28] I see, it must not be in the aggregator hash [19:37:39] it shows with showJobs but not getJobQueueLengths [19:37:40] blazecat: I queued those jobs 2 weeks ago... [19:45:40] legoktm: is it fixed now? [20:38:41] anomie: So I told robla that I was delving into making HTMLForm work with OOUI, and he said you might be interested and/or able to help [20:39:20] anomie: Mostly I could use help with determining if things still need to be supported and how they should look - I'd love to tell MatmaRex to make me new widgets, but I don't want to force him to do too much [20:40:20] marktraceur: i'll be happy to work on anything that has nothing to do with icons for a while [20:40:21] we MUST have a datepicker that supports both Gregorian and Julian calendars. also, a pony [20:40:39] MatmaRex: The big one I had trouble with so far was CheckMatrix. [20:40:51] I have a feeling there's no good way to do that that isn't "Make a new OOUI widget" [20:41:33] MatmaRex: Also, the radio button field type has support for nested arrays with H1 labels separating sections, which isn't supported in OOUI, we can do it another way if you want, but it seems like a widget candidate [20:41:55] Apart from that I'm going to chat with anomie about things for now, I think [21:12:38] marktraceur, MatmaRex: Chances are anything in there needs to be supported somehow or other. And we need a date picker, but I don't care if it supports Julian calendars or not. [21:13:53] marktraceur: CheckMatrix will probably want T88250 to replace the mode that currently uses jQuery.chosen. The list-of-checkboxes mode should be ok as a list of checkboxes, I'd think. [21:15:37] anomie: Yeah, list of checkboxes is fine, but I think CheckMatrix has a literal matrix, so two-dimensional, which is trickier. [21:15:39] marktraceur, MatmaRex: And once we have a date picker, we need a date range picker. And it needs to support both start+end dates or start date + duration, because Design. [21:16:00] Oh, does it? I thought it was just a list. [21:16:01] * anomie looks [21:16:39] Maybe I read it wrong [21:16:43] It keeps talking about columns [21:16:51] checkmatrix is indeed a matrix, yes [21:17:03] Funny name for a list if not :P [21:18:11] * legoktm gently pokes anomie with https://gerrit.wikimedia.org/r/204623 [21:18:13] marktraceur: you could still do that as just a HTML …
wrapping OOUI CheckboxInputWidgets. not sure how useful such a control is in general [21:18:23] Weird. "Operates similarly to HTMLMultiSelectField" which makes it sound like a list, but it is in fact a matrix. Only user seems to be Echo. [21:18:36] marktraceur: anomie: we don't have date pickers in OOUI yet, and in fact we don't have them in HTMLForm either, i think [21:18:38] anomie: And OAuth, which is why I can't ignore it. [21:19:12] marktraceur: you can always ignore anything by just falling back to non-OOUI form bits. :D [21:19:29] MatmaRex: I figure, but I don't like that option...I'd really like to have it all neat and fancy. [21:19:33] (OOUI is a pretty thin wrapper over HTML, so you can mix-and-match freely) [21:19:35] yeah, true [21:19:36] MatmaRex: The reason we don't have date pickers in core HTMLForm is because people whined about jQuery.ui when I tried to add one. So SecurePoll and ApiFeatureUsage both have copies of the same code to do it. [21:20:19] MatmaRex: https://gerrit.wikimedia.org/r/#/c/118111/ specifically [21:20:29] anomie: okay, but… not even just ? [21:20:43] (that's pretty widely supported by now) [21:21:26] http://caniuse.com/input-datetime doesn't look that wide: no FF, no IE, and no Safari. [21:22:26] hmm. okay. in my defence, all browsers i use support it. :) [21:22:48] legoktm: Replied [21:22:51] still, we can just do placeholder=YYYY-MM-DD for the rest [21:23:24] That's not all that user-friendly though. [21:23:31] anomie: perhaps https://gerrit.wikimedia.org/r/#/c/118111/ should be split in two parts to appease the whiners, one that just adds the PHP code, and the second adding jQuery UI? [21:23:45] anomie: more user-friendly than what we have now, i think [21:23:56] MatmaRex: The jQuery.ui was already in core, they just didn't want to use it. [21:24:02] (for example Contributions has separate dropdowns to choose year and month) [21:24:29] I feel like we put a different datepicker in UW. [21:24:35] I forget what it's called [21:24:37] But it's nice. [21:25:35] anomie: ah, whatever. if we implement the date picker in OOUI, we can just use that for all form types. we should do that. [21:26:02] anomie: responded (no, we don't) [21:26:04] MatmaRex: Steal marktraceur's date picker. :-) [21:26:29] MatmaRex: Or anomie's. :-) [21:26:41] Oh, I'm maybe incorrect, it might be jquery.ui [21:26:53] I could have sworn we changed it [21:27:49] anomie: I'm just going to deploy it now [21:31:03] James_F: Was there not a new datepicker in UploadWizard? It's using jqui now I think. I'm so confused. [21:34:25] Oh well, use another library, implement it in OOUI, {{done}} [21:34:26] Anyway [21:34:55] hoo: any idea on how https://phabricator.wikimedia.org/T96267 happened? that bogus renameuser_status entry seems suspicious to me [21:35:03] MatmaRex: I guess I'll do what I can in OOUI, then make it all into HTML up the line. [21:35:18] marktraceur: Oh. Boo. [21:35:35] legoktm: mh, I have an idea [21:46:24] legoktm: Got it [21:46:40] woot :D [21:49:47] hoo: err, are you going to say what it is? ;) [21:50:07] legoktm: Patch uploaded [21:50:22] Hope it's the right thing to do(tm) [21:51:08] https://gerrit.wikimedia.org/r/204637 [21:51:45] that makes total sense [21:53:19] I'll backport that during swat [21:53:26] :) [21:57:27] ... Why is Paladox going around updating random old patches that have been sitting with -1 or -2 for a long time? Considering what he did with T86570, I doubt it's to try correctly fixing them. [21:58:12] anomie: he's just rebasing them afaik [21:59:18] 6MediaWiki-API-Team, 10MediaWiki-extensions-CentralAuth: Possible to globally rename the same user twice (race condition) - https://phabricator.wikimedia.org/T96267#1213950 (10hoo) 5Open>3Resolved a:3hoo [21:59:31] legoktm: In the couple I checked just now, it seems like he's updating bugzilla bug numbers to phab, and sometimes rebasing too. Useless inbox spam either way. [21:59:45] :| [22:00:10] * legoktm spins the blame wheel [22:00:15] domas!!! [22:12:59] 6MediaWiki-API-Team, 10MediaWiki-extensions-OAuth, 5Patch-For-Review: OAuth permission screen needs redesign for better usability and comprehension - https://phabricator.wikimedia.org/T75062#1214013 (10Jaredzimmerman-WMF) perhaps we could look into how translatewiki does their main page (e.g. en extension th... [22:27:34] 6MediaWiki-API-Team, 6CA-team, 10MediaWiki-API, 10MediaWiki-User-login-and-signup, and 4 others: App says I'm logged in, but edits are saved from IP - https://phabricator.wikimedia.org/T75086#1214084 (10bearND) a:3Dbrant [22:30:59] hoo: https://gerrit.wikimedia.org/r/204652 easy :) [22:32:40] approved [22:32:43] thanks :D [22:33:12] anomie: I know it's late in the day for you, but do you have time for an urgent and quick-ish CR? [22:34:00] ah, he's idle, and thus probably already off. If any of you guys have time, it's https://gerrit.wikimedia.org/r/#/c/197034/ . I have to head into an interview in a few minutes so I can't look at it for another hour. [22:39:17] anomie: marktraceur: please email me, or file bugs and assign to me, if you need anything done. this looks like pretty useful work and i'll be happy to help. off for the night~ [22:39:33] MatmaRex: Will do! Cheers. [22:52:58] ori: +1'd, someone should probably review my changes though [22:53:43] legoktm: will do. thanks a ton [23:00:40] "but just "Type: string" is more useless than asking someone what they want for lunch and they reply "Food"." haha [23:48:05] 6MediaWiki-API-Team, 10MediaWiki-API, 5MW-1.25-release, 5Patch-For-Review: Clean up core API data formats for new formatversion - https://phabricator.wikimedia.org/T87053#1214330 (10Legoktm) [23:50:31] 6MediaWiki-API-Team, 10MediaWiki-API, 7Easy, 5Patch-For-Review: API help should always give parameters' type and default - https://phabricator.wikimedia.org/T47652#1214337 (10Legoktm) 5Open>3Resolved [23:57:30] anybody knows how I can enable Gadgets menu in preferences on my local Vagrant copy? [23:57:48] I enabled gadgets role, loaded the extension, but the prefs are not there [23:59:17] SMalyshev: you have to actually create some gadgets by editing MediaWiki:Gadgets-definition [23:59:42] legoktm: I can’t edit it… it gives me Permission error