[00:00:07] "staff" ? [00:00:11] RoanKattouw, we're not invited then [00:00:12] :P [00:00:15] haha [00:00:22] engineering staff [00:00:22] Come to our continent [00:00:23] Well why would there be a staff meeting in Germany? [00:00:25] Not invite us [00:00:26] PFFFT [00:00:31] whaha [00:00:33] RoanKattouw, cause a lot of the tech staff will be there [00:00:36] We can crash it [00:00:37] and yes, I believe you'll be invited [00:00:42] We'll happen to be in town, you know [00:00:52] And some of us speak the language ;) [00:01:06] Hopefully my exams fall alright [00:01:21] that's my very vague understanding of the plan [00:03:25] I totally dropped the ball on making a techblog post about the category collation rollout. I discussed it with Tim last week, and then I think I got the timing a week off or something [00:03:40] Tim is still not here [00:03:44] It's 11am [00:03:49] Tim-away: you around? [00:04:07] I would like him to show up before I go to bed [00:26:04] http://probablyinteractive.com/url-hunter Abusing the HTML5 history to the extreme (needs a modern browser) [00:27:28] also, if you open that, do it in a new tab as it will generate several hundred lines of history [00:38:23] neilk_: Awesome :D [00:38:33] Hashtag Pacman :P [00:39:07] holy crap, that just made my day [00:39:16] only 1 history entry in Safari though [01:15:08] damn. a copy of amazing fantasy #15 just went for 1.1M [06:25:32] hey there [06:25:44] i get this error while installing mediawiki [06:25:46] Fatal error: Maximum execution time of 30 seconds exceeded in C:\wamp\www\Mediawiki\includes\db\DatabaseMysql.php on line 2 [06:26:31] do i increase my max_execution_time var or is there some other error possible? [07:11:24] worked on changing max_exec_var [07:12:10] installation doesn't make changes to any file on the mediawiki folder except generating a LocalSettings.php file? [07:18:13] in 1.17 i believe even that doesn't happen [07:20:14] then what does installation do in 1.17? [07:21:55] i am using 1.18alpha [07:22:19] i think it generates a file that you then have to upload to the server yourself [07:22:24] and it generated a LocalSettings.php file which i had to upload [07:22:28] exactly [07:22:44] oh ok [07:23:06] but it doesn't actually make any changes itself then *shrug* [07:26:20] lol [07:26:50] really cool for the development environment [10:23:48] robla: Re blog post: we didn't actually switch the site to UCA because the required libraries weren't installed on the cluster. AFAIK there's no ETA on setting this up, Tim-away would know [15:28:35] robla: I am it seems successfully convincing CiviCRM to do GSoC, as long as I can help them recruit students [15:28:57] oh, cool [15:29:00] http://forum.civicrm.org/index.php/topic,18844.0.html [15:38:12] RoanKattouw: ^demon: (and anyone else)....so, a lot of the category problems we're having stem from the job queue problems (?). Is Tim the only one positioned to investigate this, or are there others that can help? [15:38:48] How did you come to that conclusion? [15:39:14] (Also, did you see my ping about how we didn't enable UCA yet?) [15:39:41] Job queue issues explain a lot of other problems that are about [15:40:01] But I don't see how category problems would be caused by them [15:40:15] There are only 5 wikis that are currently running or have not yet run the population script [15:40:22] enwiki, commonswiki,. frwiki, jawiki, ruwiki [15:40:28] Indeed, problems are from not updated, and the script is doing direct db inserts, not job queue delays [15:40:50] Category-related weirdness on those wikis are almost certainly intermediate-state-during-update issues [15:40:57] *weirdnesses [15:41:01] RoanKattouw: I did get your ping, thanks [15:41:36] RoanKattouw: as far as the conclusion I drew about the dependency, that was from ^demon closing a bug that I had filed as a dupe of the job queue problem [15:41:40] If there are any other wikis where reproducible weirdness happens, I'd be very interested [15:41:40] *robla looks for the number [15:42:13] <^demon> The bug was about category changes in templates not being reflected yet. [15:42:42] <^demon> Which as best I could tell was job-queue related. [15:42:47] Ah [15:43:04] Yes that's related to deferred LinksUpdate jobs [15:43:26] The sortkey changes and associated issues are a different breed entirely [15:43:30] oops....I was getting two bugs mixed up [15:43:37] sorry ^demon [15:43:45] <^demon> It's ok :) [15:44:26] this is the problem on commons with at least one big category: https://bugzilla.wikimedia.org/show_bug.cgi?id=27956 [15:45:53] RoanKattouw: any way for me to track the progress of the population script, other than sitting here? [15:46:16] robla, keep poking roan :D [15:46:16] Not really [15:46:24] I put ETAs in the server admin log [15:46:32] Looking into the commons cat now [15:48:52] OK so [15:49:06] The code tries to fetch 200 files, 200 pages and 200 subcats [15:49:28] It gets 167 files, and 200 things that claim to be pages but are really files, and 0 real pages [15:49:39] So it doesn't put a (next 200) link on the pages section because there were no pages in the result [15:50:02] And it doesn't put a (next 200) link on the files section either because there were fewer than 200 files in the query result [15:50:18] oops [15:50:55] So, bad news: category paging is broken due to an edge case in the intermediate state we're not handling well [15:51:10] Good news: the repopulation script will fix this [15:52:22] oh, cool [15:53:31] ok...on the UCA front: is it technically accurate that we're "switching" to UCA (just not "switched")? [15:54:01] Don't know, still need to ask Tim [15:54:30] I want to, but Tim spoke some ambiguous words that could be construed as opposition to installing UCA [15:54:45] oh...hrm [15:54:59] Which would be weird cause he wrote it, maybe he meant that it was a good idea to not introduce the new code /and/ a different collation at the same time [15:55:28] So I've been planning to hunt him down and ask, but I've been forgetting to ask in the mornings when he's still around [15:55:52] ok, well, he reviewed the blog post (and the UCA bit in particular), so I'm assuming he's planning on it [15:56:15] OK [15:56:27] Was that before or after we deployed? [15:56:44] after [15:56:50] OK [16:04:05] so, the category maintenance script explains some of the problems. is Tim the only one working on the job queue problem? [16:05:04] Ariel was poking at it on and off [16:05:08] But AFAIK, yes [16:06:05] RoanKattouw: ping [16:09:15] It looks like we have a problem with either our cache or bits server. I received several complains from users who have their page loaded without CSS [16:09:42] Just being discussed in -tech ;) [16:10:34] Aha [16:10:36] Thanks [16:28:49] *RoanKattouw rages at MySQL [16:29:03] ORDER BY cl_type yields page, subcat, file, because that's the order the enum is defined in [16:29:25] But cl_type > 'page' only yields subcat, because 'file' < 'page' [16:29:41] I'm wondering how this behavior is even *possible* with an index [16:32:52] Surprise surprise, enums suck [16:32:56] *RoanKattouw smacks AryehGregor [16:33:58] <^demon> We killed the enum in CR for a reason :p [16:34:15] :D [16:34:19] *an enum [16:34:22] there's still others ;) [16:35:53] It seems neither the range scan nor the order are indexed [16:37:37] This is really, really, annoying, it probably means we need to change the enum to a varchar asap [16:37:52] Gonna do that in trunk now [18:07:18] RoanKattouw_away, are you saying that for ORDER BY, MySQL treats enums as integers, but for equality comparisons, it treats them as strings? [18:07:24] I didn't realize enums were that horrible. [18:07:44] Er, inequalities, I mean. [18:07:51] But why do you need to do cl_type > something? [18:08:10] Can't you just do cl_type != 'page' or cl_type IN ('subcat', 'file') or something? [18:08:40] Wait, the ordering isn't indexed either? [18:08:42] That's . . . wtf. [18:15:01] AryehGregor: I need to page through catlinks in API lis=categorymembers [18:15:09] Hmm. [18:15:18] I figured paging on (cl_type, cl_sortkey, cl_from) (which is indexed when cl_to is fixed) was reasonable [18:15:24] And ORDER BY cl_type, cl_sortkey doesn't use the index? [18:15:32] That sounds really improbable. [18:15:50] Although yeah, maybe we should just use a tinyint for cl_type. [18:16:01] Or a varchar, but that seems excessive when we'll probably never have more than three types. [18:22:44] *TrevorParscal stands by while RoanKattouw hacks away [18:23:17] AryehGregor: Which ever is fine. Do you have time and motivation to make that change on trunk? [18:23:39] And ORDER BY cl_type, cl_sortkey doesn't use the index? [18:23:40] That sounds really improbable. [18:24:06] I guess that one does use the index, my bad [18:24:09] But range scans don't [18:30:08] I bet they would if you translated the string to a number. [18:30:17] If 'page' is 0, do > 0 instead of > 'page'. [18:30:19] Stupid, but . . . [18:30:37] Anyway, no, I'm probably not going to be changing trunk, I have way too many things right now to do in life. [18:44:48] TrevorParscal: OK, edit section link deployment [18:44:56] yes [18:45:11] The rev list I had was complete, right? [18:45:57] yes [18:46:01] OK [18:46:05] lets try that on test [18:46:13] Lemme merge that first [18:46:25] cool [18:46:55] Merge running [18:57:42] i wonder how many people in the us are watching charlie sheen and thinking that he's clever or something, or have absolutely no experience with addicts of any kind, and thus don't know what's really going on. [18:58:25] RoanKattouw: sorry networking issues [18:58:27] how's it going? [18:58:31] i've heard some people say that he's just a "misunderstood genius" and others say that he's bipolar and other people think it's an elaborate comedy act. [19:00:34] someone like my dad would never be able to recognize post-addiction hypomania. [19:00:37] it's depressing. [19:02:09] Mobile redirect still broken, working on that [19:02:18] Almost ready to turn on SEL on test [19:02:46] ok, well I have a patch :) [19:02:51] about to commit it [19:05:19] it's 2 characters [19:05:35] basically JS was evaluating '0' as true [19:06:04] so I used if ( bucket > 0 ) as the condition, rather than just saying if ( bucket ) [19:06:22] RoanKattouw: see r83602 [19:09:34] Merging [19:10:04] :) [19:12:21] http://pastebin.com/ggQLdJev [19:13:49] Needs global true, user false it seems [19:14:43] It's not loading [19:14:55] Oh wait [19:14:57] you used the settings I gave you right? [19:15:00] I managed to end up on mw.org [19:15:17] RoanKattouw: http://jsperf.com/jquery-selector-perf-select-right-to-left [19:15:36] Just created it to test an article I read. [19:16:53] TrevorParscal may be interested as well [19:19:23] Yay, it works [19:19:26] Grr [19:19:33] *RoanKattouw yells at Trevor's conection [19:19:47] Isn't he at Kuls today? [19:20:11] He is [19:20:11] dude, networking issues [19:20:13] sorry about that [19:20:17] yes, I'm at Kul's :) [19:20:20] Yay, it works [19:20:23] Grr [19:20:23] * RoanKattouw yells at Trevor's conection [19:20:38] SectionEditLinks WFM on test after hacking the cookie myself [19:20:47] Now unsetting cookie, see if I win the lottery with odds=100% [19:21:36] it's not setting a cookie for me... [19:22:03] It is for me [19:22:15] you used global = false, user = true? [19:22:31] >>> $.cookie('ext.vector.sectionEditLinks-bucket'); "2" [19:22:33] Yes [19:22:35] No [19:22:38] global=true, user=false [19:23:04] that might screw with the way the noexpereiments setting changes it [19:23:18] hmm [19:23:25] might need to patch this [19:23:25] Let me try that by logging in and setting it [19:23:44] ah, yes for anons it's working [19:23:49] you need user = true [19:24:24] strange, I'm in bucket 2 but it's not actually changing the page... [19:25:44] That's very weird [19:25:52] As an anon? [19:26:04] either [19:26:10] did you actually see the link move? [19:26:56] Yes, WFM now [19:27:05] I changed it back to user true, global false [19:27:09] The pref toggles it for me while logged in [19:27:19] As an anon, you might need to purge the page [19:27:57] WFM as anon [19:28:21] on http://test.wikipedia.org/wiki/SectionEditLinkTest [19:28:38] yeah... hmm [19:28:45] Does it work for you now? [19:29:04] no [19:29:10] some sort of caching thing? [19:29:15] Possibly [19:29:19] Works here [19:29:22] Same URL? If so, Ctrl+F5 [19:29:58] Hm.. the darker version of the pencil seems to be loaded only when I hover it (half-a-second delay between hover and seeing the image) [19:30:08] sprite perhaps :) [19:30:25] or jQuery.fn.opacity() [19:30:43] I @embedded it I think [19:30:48] Oh, d'oh, didn't merge that rev [19:30:48] hmm [19:31:48] Merging now [19:32:44] Should be fixed nwo [19:32:49] TrevorParscal: OK we ready to go live? [19:33:18] still isn't working for me... [19:33:22] going to try another browser [19:33:38] http://test.wikipedia.org/wiki/SectionEditLinkTest right? [19:33:48] ok [19:33:50] worked in FF [19:33:55] must be a rediculous caching issue [19:34:15] Alright, let's take this ive [19:34:16] Hnm.. why do they jump from right [edit] to left "/ edit" ? Afaik in earlier prototypes it was done serverside. [19:34:24] It was [19:34:28] But we want to bucket test now [19:34:47] Now setting odds back to 0.1% and enabling on enwiki [19:34:47] okay, so when final it will be back to server side again ? [19:34:51] this is a client-side experiment [19:34:59] yes, the real version would do this in the HTML [19:35:04] k, np. [19:35:10] Cool [19:35:29] RoanKattouw: we're looking good now [19:35:53] Can it also be put to 100% in user preferences ? [19:35:57] good call on the @embed thing btw [19:36:06] Krinkle: ? [19:36:10] There is no pref for it right now [19:36:16] k [19:36:24] Krinkle: you just have to change your bucket to 2 if you want it [19:36:31] most browsers let you edit cookie values [19:36:33] There's just a "don't give me this experimental crap" pref [19:36:38] yeah [19:36:45] OK we're live [19:36:47] it's supposed to be a random selection [19:36:51] Let's look at some clicktracking data [19:37:16] RoanKattouw: Which pref is that ? Is the bucet test now only for anons ? [19:37:41] No, for everyone [19:37:45] RoanKattouw: why does wikipedia only show view source for everyone? [19:37:53] everything I mean [19:38:03] the whole thing is locked? [19:38:06] For users there's hthe "Exlcude me from experimental features" pref in Appearnce IIRC [19:38:12] TrevorParscal: Try vising a non-protected page [19:38:23] I'm just clicking random article [19:38:31] ah, got it RoanKattouw. I was looking at a different tab. [19:38:38] this is locked? http://en.wikipedia.org/wiki/Babson_Park_Woman%27s_Club [19:38:48] whaha [19:38:52] I'm anon, and see view source instead of edit [19:38:56] everywhere [19:39:09] WFM [19:39:10] I see Edit [19:39:20] Maybe your IP is blocked? [19:39:31] Click on view source, see what reason it gives yuo [19:40:05] yeah, my IP is blocked [19:40:06] wtf! [19:40:16] Loading Michael Jackson is bloody slow [19:40:24] dude, this is Kul's house [19:40:25] TrevorParscal is a vandal. [19:40:26] wtf [19:40:31] Kul is a vandal, then. [19:40:48] wtf! [19:41:00] he was probably editing while on vicodin [19:41:12] Dynamic ip? [19:41:30] dude, it's a whole range that's blocked 75.36.128.0/17 [19:41:36] lol [19:42:06] 75.36.128.0 [19:42:10] Checkuser block [19:42:13] WHOOPS [19:42:17] [19:42:19] (expires on 20 April 2011 at 07:11, anon. only, account creation blocked) [19:42:49] API fail? [19:43:20] why's it html encoded the '? [19:43:40] look at the whole thing [19:43:42] Or it's behaving that the module doesn't exist [19:43:50] it's a url, that contains a url as a param [19:43:56] and the contained url has params [19:44:00] it works on test [19:44:01] ClickTracking was disabled [19:44:04] On en [19:44:07] i see [19:44:14] you can set the click tracking rate to 0 btw [19:44:19] Fixed [19:44:26] they probably disabled the whole thing rather than turning it down [19:44:29] What's it default to? [19:44:36] Oh, that [19:44:43] I just reenabled it on en [19:44:46] Will set rate to 0, good idea [19:45:40] 'wmgClickTrackThrottle' => array( [19:45:41] 'default' => 100, [19:45:41] -1 is supposedly the special value for off for everyone [19:45:43] 'strategywiki' => 1, [19:45:45] 'testwiki' => 1, [19:45:46] ), [19:46:12] Yeah [19:46:16] Just set -1 on all wikis [19:46:36] yes [19:46:38] awesome [19:47:03] that will ensure we don't send out the clicktracking extension to people who don't need it [19:47:09] "Unrecognized value for parameter 'action': clicktracking" [19:47:36] Hm... Edit section fails on testwiki [19:47:37] I just fixed that didn't I [19:47:43] nope [19:47:48] atleast not on testwiki [19:48:00] http://test.wikipedia.org/w/api.php?action=clicktracking&eventid=ext.vector.sectionEditLinks-bucket2-click&token=null&redirectto=%2Fw%2Findex.php%3Ftitle%3DSectionEditLinkTest%26action%3Dedit%26section%3D1%26clicktrackingsession%3Dnull%26clicktrackingevent%3Dext.vector.sectionEditLinks-bucket2-save [19:48:04] WFM on enwiki [19:48:33] indeed, manaully edited to use enwiki, works there [19:49:06] RoanKattouw: r83607 might be needed to actually enable tracking on en [19:49:09] API module not registerred [19:49:50] http://test.wikipedia.org/wiki/Special:Version doesn't show click tracking... [19:50:02] Hm.. weird, I can apparently edit en.wiki's main page.. Uh ? [19:50:08] http://en.wikipedia.org/w/index.php?title=Main_Page&action=edit [19:50:18] I have no sysop right there, and it's sysop protected [19:50:19] Fixing [19:50:29] Fine for me as anon [19:50:43] ClickTracking is now enabled on test [19:51:05] that's more sne [19:51:28] TrevorParscal: OK we seem to be live now, without problems and with 0.1% tracked [19:51:36] yeah? [19:52:25] When did you want to go to 1%? [19:52:54] let's make sure it's working as expected for an hour [19:53:00] also, did you see r83607 [19:53:15] it's going to be needed for tracking how many people are getting thrown into buckets [19:53:33] !r 83607 [19:53:33] --elephant-- http://www.mediawiki.org/wiki/Special:Code/MediaWiki/83607 [19:53:33] http://www.mediawiki.org/wiki/Special:Code/MediaWiki/83607 [19:54:05] heh [19:54:56] and, quickly would be good [19:55:06] quickish :) [19:55:21] it's just to help verify that the buckets are evenly distributed [19:55:34] so, you are leaving now, and I'm supposed to be on a call [19:55:40] so, will you be around later at all? [19:56:59] Not actively [19:57:10] As in, watching a movie [19:57:44] well, I was hoping to peek at the clicktracking table, then change the rate [19:57:59] could you push that rev out thoguh? [19:58:14] r83607 that is [19:59:47] I did [20:22:13] when will the date of the berlin meeting be announced? [20:23:13] Bryan, it has been [20:23:26] hmm [20:23:30] did I miss a mail? [20:23:32] http://www.mediawiki.org/w/index.php?title=Berlin_Hackathon_2011&diff=389733&oldid=389713 [20:23:34] *Bryan checks the wiki page [20:23:38] Bryan, ^ [20:23:43] they haven't sent an e-mail yet [20:23:51] ah thanks [20:55:09] Hm.. could someone help me figure out why I can edit the en.wiki Main page ? Seems to be a bug though I can't narrow down why or what. [20:56:58] I can't in chrome incognito mode [20:57:28] Or as my bot account [20:57:58] I tried logged out and under bot account too, doesn't appear there [20:58:19] It's not just the label of the Edit-button being "Edit" or "View source", I can actually save the page [20:58:37] I'm getting standard edit only by admin, and on my main account i am admin [20:58:59] Krinkle, does the save actually go through? Try something simple, edit a comment or something [20:59:06] I did [20:59:36] lol, you did [21:00:49] I have globalinterface right to edit MediaWIki-namespace but that shoulnd't grant me right to edit sysop-protected pages in general. [21:01:05] I'll ask a steward to revoke it for a minute [21:02:24] Lol, now I can't save the page [21:02:29] So thats causing it. [21:02:37] Weird. [21:03:18] ['editinterface'] = true causes ability to edit any sysop protected page ?? [21:03:20] *Krinkle files bug [21:03:26] Hey, at least it's a limited bug [21:03:52] Sure, no vandalmania [21:04:16] I mean I'm the only one in that group anyway :P [21:04:27] I think kaldari is/was too. [21:05:06] heh [21:05:17] Kaldari is well known for his vandalism [23:46:34] TrevorParscal or RoanKattouw, some RL questions...? [23:47:02] yo [23:47:13] how many places can a module be defined? [23:47:32] is there some cache for module definitions that isn't Resources.php or the Hooks script that I'm loading? [23:47:53] because I'm getting a "module redefined" error now that I've moved one around [23:49:41] no, that error is a PHP exception iirc that occurs durring module registration [23:49:54] yes [23:50:12] I assume you've looked to make sure you aren't actually registering something 2x [23:50:17] what's the module name? [23:50:23] mediawiki.language.parser [23:50:36] now, I used to define it in Resources.php, but now it is definitely gone [23:50:45] and I'm defining it instead in UploadWizardHooks.php [23:51:02] interesting [23:51:04] however, I was getting similar errors even before that, so I'm not sure where this other registration is coming from. [23:51:32] someone may have added the global way somewhere? [23:51:52] Tim added the ability to append $wgResourceModules the way we append $wgHooks [23:52:08] the original design is using a ResourceLoaderRegisterModules hook [23:52:12]