[00:21:52] i'd love for some performance engineer types to look at this and provide hate: https://www.mediawiki.org/wiki/Author_Thanks [00:26:26] jorm: I like how you include words like "sadistic" in your proposals :) [00:26:36] i call it like i see it. [00:26:53] i can't imagine being someone like Kaldari and getting THOUSANDS of notifications per day. [00:27:16] it would literally be sadistic for us to do that, knowing how awful it would be. [00:29:51] For denial of service, you could maybe piggyback on the existing rate limiter code, and have the DoS prevention be separate from the implementation of this thing [00:29:56] jorm: Depending on the amount of data and whatever (I'm tired...) you probably want to have both tables, on to restrict double endorses and another one to get per article+user rollups [00:30:39] bawolff: Well, we have two issues here: DoS which we can kill using ratelimits and users just endorsing the same edit/page 5 times [00:30:45] right. [00:30:50] hoo: right [00:34:16] I wonder how many people would actually click "thanks" - if it ends up being a large portion of anons, one might even have to take into consideration the higher write load on the db [00:34:21] * jeremyb looks up @ robla [00:34:37] but if it came to that, that could be worked around [00:35:22] bawolff: Yes, that can indeed be icky [00:36:01] If it was an important feature, and that was an actual issue, it could be put in a separate db I suppose [00:36:12] or even use non-db storage [00:36:35] That sounds like a rather typical workload for Cassandra [00:36:57] and the rollops could then be aggregated into SQL... but you wouldn't want to do that unless oyu have to [00:37:16] lots of work... and even MySQL can do some quite heavy lifting here [00:37:18] If we really didn't care about it being reliable, we could stick it in memcache [00:37:32] in a similar fashion to how we do rate limitting [00:37:52] bawolff: That's not queryable [00:38:37] right, that would be a problem [00:39:19] jorm: To sum that up for you: No matter how smart your MySQL table layout is: If a lot of people keep clicking on these a lot, stuff will die (painfully) [00:40:02] awesomesauce. [00:51:36] Preparing to replace hhvm 2.4.0~saucy (using .../hhvm_2.4.1~saucy_amd64.deb) ... [00:51:37] yay? [00:59:25] Reedy: where is this? vagrant? [00:59:31] locally [00:59:43] so... not vagrant then [00:59:43] i installed the package and never did anything further [00:59:44] * Reedy grins [00:59:46] No [00:59:47] lolno [01:20:55] Hmm line 338 of TimedMediaHandler/libs/getid3/module.audio-video.matroska.php looks kind of sketchy [01:21:29] (03CR) 10Daniel Friesen: "Both shared.css (mediawiki.legacy.shared) and commonPrint.css (mediawiki.legacy.commonPrint) are loaded implicitly by SkinTemplate." [core] - 10https://gerrit.wikimedia.org/r/113169 (owner: 10Mattflaschen) [01:26:16] (03CR) 10Alex Monk: "Almost forgot about this: 2014-02-02, #mediawiki-visualeditor, [01:18:33] Krenair: I'd note BTW that that patch really needs a R" [core] - 10https://gerrit.wikimedia.org/r/32700 (owner: 10Nischayn22) [01:30:42] oh I guess its fine [01:48:39] (03PS1) 10Brian Wolff: Add ID3 dependencies to getid3 library [extensions/TimedMediaHandler] - 10https://gerrit.wikimedia.org/r/113760 [02:40:08] (03PS3) 10Brian Wolff: Fix comment for $wgUseImageResize [core] - 10https://gerrit.wikimedia.org/r/113659 (owner: 10TTO) [02:40:30] (03CR) 10Brian Wolff: [C: 032] "Thanks for fixing that up." [core] - 10https://gerrit.wikimedia.org/r/113659 (owner: 10TTO) [02:44:31] (03Merged) 10jenkins-bot: Fix comment for $wgUseImageResize [core] - 10https://gerrit.wikimedia.org/r/113659 (owner: 10TTO) [02:50:55] (03PS1) 10Gnosygnu: v1.2.3.1 [xowa] - 10https://gerrit.wikimedia.org/r/113761 [02:58:24] (03PS2) 10Ori.livneh: Add scap.ssh module [tools/scap] - 10https://gerrit.wikimedia.org/r/113660 [02:59:00] (03PS3) 10Ori.livneh: Add scap.ssh module [tools/scap] - 10https://gerrit.wikimedia.org/r/113660 [03:01:48] (03CR) 10Brian Wolff: [C: 032] "PagedTiffHandler uses it, but it sets it in its own config anyways" [core] - 10https://gerrit.wikimedia.org/r/113653 (owner: 10TTO) [03:05:11] (03Merged) 10jenkins-bot: Remove unused setting $wgImageMagickIdentifyCommand [core] - 10https://gerrit.wikimedia.org/r/113653 (owner: 10TTO) [03:14:35] (03PS1) 10Physikerwelt: WIP: Generate SVG from LaTeXML output [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113762 [03:14:50] (03CR) 10jenkins-bot: [V: 04-1] WIP: Generate SVG from LaTeXML output [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113762 (owner: 10Physikerwelt) [03:15:36] (03PS4) 10Ori.livneh: Add scap.ssh module [tools/scap] - 10https://gerrit.wikimedia.org/r/113660 [03:15:49] (03PS5) 10Ori.livneh: Add scap.ssh module [tools/scap] - 10https://gerrit.wikimedia.org/r/113660 [03:27:21] (03PS1) 10Physikerwelt: Push math nodes onto mathArray individually to avoid problem in IE < 9 where push.apply doesn't work on a nodeList. Also, avoid duplicate elements in the array by checking for namespace prefixes. This needs to work for both HTML and XHTML. Resolves issue [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113763 [03:28:18] (03CR) 10Physikerwelt: [C: 032] Push math nodes onto mathArray individually to avoid problem in IE < 9 where push.apply doesn't work on a nodeList. Also, avoid duplicate el [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113763 (owner: 10Physikerwelt) [03:28:25] (03Merged) 10jenkins-bot: Push math nodes onto mathArray individually to avoid problem in IE < 9 where push.apply doesn't work on a nodeList. Also, avoid duplicate elements in the array by checking for namespace prefixes. This needs to work for both HTML and XHTML. Resolves issue [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113763 (owner: 10Physikerwelt) [03:29:45] (03PS1) 10Physikerwelt: Push math nodes onto mathArray individually to avoid problem in IE < 9 where push.apply doesn't work on a nodeList. Also, avoid duplicate elements in the array by checking for namespace prefixes. This needs to work for both HTML and XHTML. Resolves issue [extensions/Math] - 10https://gerrit.wikimedia.org/r/113764 [03:36:20] (03CR) 10SPQRobin: [C: 04-1] "This patch doesn't seem to change anything in practice. When you're viewing a certain page, that title is automatically capitalized. The p" [extensions/WikimediaIncubator] - 10https://gerrit.wikimedia.org/r/113650 (owner: 10Ebe123) [03:54:44] (03PS2) 10Bartosz Dziewoński: Remove a HTML backwards-compatibility hack from 2011 [core] - 10https://gerrit.wikimedia.org/r/113548 [03:54:50] (03CR) 10Ori.livneh: [C: 032] Remove a HTML backwards-compatibility hack from 2011 [core] - 10https://gerrit.wikimedia.org/r/113548 (owner: 10Bartosz Dziewoński) [03:57:04] (03PS3) 10Bartosz Dziewoński: Vector: Simplify text paths in tiny icons [core] - 10https://gerrit.wikimedia.org/r/112234 [03:57:09] (03CR) 10Ori.livneh: [C: 032] Vector: Simplify text paths in tiny icons [core] - 10https://gerrit.wikimedia.org/r/112234 (owner: 10Bartosz Dziewoński) [04:00:04] (03Merged) 10jenkins-bot: Remove a HTML backwards-compatibility hack from 2011 [core] - 10https://gerrit.wikimedia.org/r/113548 (owner: 10Bartosz Dziewoński) [04:01:17] (03Merged) 10jenkins-bot: Vector: Simplify text paths in tiny icons [core] - 10https://gerrit.wikimedia.org/r/112234 (owner: 10Bartosz Dziewoński) [04:17:34] (03CR) 10Gnosygnu: [C: 032 V: 032] "approving own commit" [xowa] - 10https://gerrit.wikimedia.org/r/113761 (owner: 10Gnosygnu) [04:21:56] (03PS2) 10Physikerwelt: WIP: Generate SVG from LaTeXML output [extensions/Math] (dev) - 10https://gerrit.wikimedia.org/r/113762 [04:26:01] Am I not supposed to touch the changes made by the translation bot? There are some strange line breaks - https://git.wikimedia.org/blobdiff/mediawiki%2Fextensions%2FPopups/75ea4566e68da7d4614c6cd31e49dacf2f723d37/Popups.i18n.php [04:28:06] (03PS12) 10Prtksxna: Add `redirects to` message at the top of the popup [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111983 [04:29:33] (03PS13) 10Prtksxna: Add `redirects to` message at the top of the popup [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111983 [04:34:25] (03PS1) 10Prtksxna: Consistency correction [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113766 [04:35:45] (03PS2) 10Prtksxna: Consistency correction [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113766 [04:37:25] (03CR) 10Prtksxna: Add `redirects to` message at the top of the popup (032 comments) [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111983 (owner: 10Prtksxna) [04:37:48] (03CR) 10Prtksxna: Add time ago in words at the bottom of the popup (031 comment) [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111977 (owner: 10Prtksxna) [05:11:15] (03CR) 10Raimond Spekking: [C: 032] Consistency correction [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113766 (owner: 10Prtksxna) [05:14:03] (03PS14) 10Prtksxna: Add `redirects to` message at the top of the popup [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111983 [05:21:48] (03PS7) 10Prtksxna: [WIP] Add EventLogging [WIP] [extensions/Popups] - 10https://gerrit.wikimedia.org/r/112661 [05:39:28] (03CR) 10Ori.livneh: [C: 04-1] "Could you call #load with an explicit 'true' for the optional async argument?" [core] - 10https://gerrit.wikimedia.org/r/113540 (owner: 10Bartosz Dziewoński) [05:40:11] When did we start hashtagging commit summaries? [05:41:38] Gloria: #old [05:41:55] So #old. [05:42:04] Can someone please fix Moiz? [05:42:08] it's JSDuck notation [05:42:20] object#method, I mean; not Moiz. [05:42:43] , I mean. [05:43:04] see [05:46:32] https://www.mediawiki.org/wiki/Manual:MediaWiki_Application_Programming_Interface_API_Guide [05:46:43] is there an API for the API, or is that title just saying the same thing twice? [05:47:02] Recursive APIs. [05:47:04] #inception [05:47:08] * ori sails away on the rio grande river [05:47:21] * Gloria bets Wikinaut. [05:50:09] (03PS8) 10Prtksxna: [WIP] Add EventLogging [WIP] [extensions/Popups] - 10https://gerrit.wikimedia.org/r/112661 [05:50:55] is that patch a work-in-progress that is itself a work-in-progress, or is the commit message just saying the same thing twice? [05:51:15] ori: :P [05:51:32] It's to not show bias between RTL and LTR. [05:51:41] ooooh, good answer. [05:51:42] Symmetry is beauty. [06:54:15] (03CR) 10Ori.livneh: [C: 032] Improve pydoc [tools/scap] - 10https://gerrit.wikimedia.org/r/113645 (owner: 10BryanDavis) [06:55:56] (03CR) 10Mdale: [C: 032 V: 032] "tested and gives correct duration for file in bug 61441" [extensions/TimedMediaHandler] - 10https://gerrit.wikimedia.org/r/113760 (owner: 10Brian Wolff) [07:01:07] Gloria: I am now! [07:01:40] legoktm: I was going to bother you about GlobalCssJs. [07:01:57] Dan G. might get in the way over the global pages. [07:02:03] Though I think it's quite silly. [07:02:09] Otherwise, it seems obstacle-free to me. [07:04:36] (03PS4) 10Ori.livneh: Improve logging output and consistency [tools/scap] - 10https://gerrit.wikimedia.org/r/113642 (owner: 10BryanDavis) [07:04:40] the MediaWiki:Global.js/css ones? [07:05:08] Given that we already let admins do it via CentralNotice which is a lot more obscure, I would think centralizing it in one place is a good thing. [07:05:16] (03CR) 10Ori.livneh: [C: 032] "Made some small fixes, see diff b/w PS3 & 4" [tools/scap] - 10https://gerrit.wikimedia.org/r/113642 (owner: 10BryanDavis) [07:05:18] (03Merged) 10jenkins-bot: Improve logging output and consistency [tools/scap] - 10https://gerrit.wikimedia.org/r/113642 (owner: 10BryanDavis) [07:05:32] legoktm: The arguments for even adding a config variable seemed super-weak to me. [07:05:37] But we'll see how loudly people complain. [07:05:44] (03PS2) 10BryanDavis: Improve pydoc [tools/scap] - 10https://gerrit.wikimedia.org/r/113645 [07:05:46] legoktm: I bumped the security bug. [07:05:48] (03CR) 10Ori.livneh: [C: 032] Improve pydoc [tools/scap] - 10https://gerrit.wikimedia.org/r/113645 (owner: 10BryanDavis) [07:05:50] (03Merged) 10jenkins-bot: Improve pydoc [tools/scap] - 10https://gerrit.wikimedia.org/r/113645 (owner: 10BryanDavis) [07:05:52] legoktm: Ball's in your court on that one. [07:05:55] yeah [07:06:04] I'll talk to Chris when I get the chance [07:06:19] been pretty busy IRL the last few days [07:07:19] I forgive you. [07:07:41] If you can get Chris to review before the end of this month, deploying to the wikis by the end of March 2014 seems feasible. [07:08:55] mhm [07:09:06] I need to figure out how different the wmf's setup with bits.wm.o is [07:09:23] Because I've only tested it with normal load.php's [07:10:11] I'm not sure why that would matter. [07:10:38] I want to make sure it doesn't :P [07:16:12] (03CR) 10Ori.livneh: "The decorator isn't needed, IMO. The one place you use it could have used the context manager instead (inside the function, or outside its" [tools/scap] - 10https://gerrit.wikimedia.org/r/113652 (owner: 10BryanDavis) [07:19:12] jorm: In the long term I want to restructure Echo to make it trivial to send a notification to a large number of users (eg: sitenotices) which I think would alleviate your performanceconcerns. [07:20:07] Well, some of them. [07:28:02] My name is Guillaume, and I approve of this plan. [07:44:15] (03PS9) 10Prtksxna: Add EventLogging [extensions/Popups] - 10https://gerrit.wikimedia.org/r/112661 [07:47:31] (03PS10) 10Prtksxna: Add EventLogging [extensions/Popups] - 10https://gerrit.wikimedia.org/r/112661 [07:48:32] (03PS1) 10devunt: Added missed database patch file [extensions/GlobalBlocking] - 10https://gerrit.wikimedia.org/r/113767 [07:50:00] Hey legoktm! Got a minute to discuss https://gerrit.wikimedia.org/r/#/c/112661/6/resources/ext.popups.core.js? [07:52:36] prtksxna: what about it? [07:53:28] legoktm: Is there something else that you suggest I can do? [07:53:44] like...not logging the page id? :P [07:53:53] hmm [07:54:07] legoktm: Do you think I misunderstood the message on top here - https://trello.com/c/Sbd8Q180/99-events-for-navigation-popups-beta-feature [07:54:55] which message? [07:55:06] I see that Dario wrote "As per the previous comments, associating a unique user to a page viewed has major privacy implications" which is basically my point [07:55:35] +++ [08:00:17] legoktm: http://pastie.org/8741313 this one [08:01:14] Nemo_bis: o/ [08:02:08] legoktm: Should I store a hash of the user id or something, that'll keep the user anon and we'll get the required data. Is that helpful [08:02:09] prtksxna: just logging user id should do that right? [08:02:23] if its a beta feature, anons won't be able to use it... [08:03:02] legoktm: Does user id only uniquely identify users and can't map back to an actual user? [08:03:19] anyone can map user id <--> user name [08:03:21] its public info [08:03:36] https://en.wikipedia.org/w/api.php?action=query&list=allusers [08:03:59] legoktm: Right, so if I hash it and save it it'll all be fine? [08:04:28] legoktm: We don't want to map it back to a real user, we just need it for analytics, right? [08:04:34] * Nemo_bis waves at prtksxna [08:04:41] eh, [08:04:58] You'd need to rotate key [08:05:06] if you just do a hash it would be simple for someone to create a table of the however million users and map it [08:05:16] /o\ [08:05:29] I think AFT used a temporary cookie or something [08:05:35] ori might know. [08:05:45] *uses [08:05:56] let's not get ahead of ourselves here [08:06:16] Nemo_bis: Was my reply helpful, I don't know much about mentorship programs and stuff… [08:06:47] ori legoktm: If we get rid of the pageId, everything is fine? https://meta.wikimedia.org/wiki/Schema:Popups [08:07:05] why get rid of pageId? [08:07:23] ori: don't want to link user to articles they're reading [08:07:29] see talk 13 min above [08:07:46] why log user id? [08:07:58] ori: Take a look at the Trello board? [08:08:02] ori: https://trello.com/c/Sbd8Q180/99-events-for-navigation-popups-beta-feature [08:08:52] prtksxna: Now I'm confused, which reply? I only have https://gerrit.wikimedia.org/r/#/c/111977/13/Popups.i18n.php from you in my inbox right now [08:09:14] I think getting rid of user_id is a better idea. [08:09:19] Nemo_bis: Erm, this one -> https://www.mediawiki.org/wiki/Thread:Extension_talk:Popups/Patrolling_and_maintenance [08:09:54] legoktm: Can you please leave a comment here -> https://meta.wikimedia.org/wiki/Schema_talk:Popups [08:12:09] prtksxna: https://gist.github.com/atdt/9046669/raw/6e4a6506ebeab2659061189b5b87e52f01bb0358/getId.js [08:12:15] * prtksxna clicks [08:12:31] errr wait [08:12:31] need to fix that [08:12:59] there [08:12:59] https://gist.github.com/atdt/9046669/raw/a5c05f7bc8d03c82ba00e3f225f1cc306de980a8/getId.js [08:13:08] (different url) [08:13:39] ori: Right, use this instead of userId? [08:13:43] this gives you a persistent unique token that is not relatable to the user id [08:13:50] ori: Also, why does this need to be in a try catch? [08:13:58] because not all browser support localStorage [08:14:37] prtksxna: ah ok, I'll look that later; mw.o Special:NewMessages is another queue ;) [08:14:38] ori: And in that case we won't be able to store anything? [08:14:38] the proper way to feature-detect localstorage is....... [08:14:39] https://github.com/Modernizr/Modernizr/blob/master/feature-detects/storage/localstorage.js#L20 [08:15:12] ori: Ah ok :) [08:15:16] prtksxna: yeah, the function will return null. but: why do you need this data anyhow? [08:15:31] ori: Did you take a look at the Trello board? [08:15:35] Nemo_bis: Alright :) [08:15:43] prtksxna: i did, but there were a lot of words there [08:15:49] lol [08:15:51] so i got scared and closed the tab [08:16:07] Nemo_bis: I just wanted to give you a heads up that I might not have understood your question properly :) [08:16:18] ori: Do you mind reading? [08:16:27] ori: I don't mind giving you a summary here :D [08:17:47] * ori is reading [08:18:06] it took me a moment to realize that trello threads are reverse-chron [08:18:12] once i did i became a lot less confused [08:18:18] ori: Yeah, thats a bit confusing… [08:19:28] re: > It would be great if we could filter out accidental displays of the popup, e.g. if they are displayed under a certain threshold of time <2 seconds… [08:19:47] I know it's not what you're asking about, but just by-the-way, that is eminently doable [08:20:25] setTimeout( function () { if ( stillVisible() ) { logEvent(); } }, 2000 ); [08:20:35] ori: Right, but we don't know yet what might be `accidental` [08:20:56] presumably anything the user closes very rapidly [08:21:23] ori: Right, we don't what that might be yet. I've commented on it on the Trello board [08:21:25] oh, i see you followed up [08:21:26] yeah [08:21:33] sorry, i really should just read everything and stop being a lazy jerk [08:22:03] (03PS6) 10UltrasonicNXT: Lots of additions [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [08:22:55] prtksxna: is there anywhere where i can preview this functionality? [08:23:03] it would help me grasp the requirements [08:23:07] (03CR) 10UltrasonicNXT: "Added sounds, should work in all browsers that support ogg or mp3 (everything except Opera I think), and moved the settings to preferences" [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [08:23:09] ori: https://gerrit.wikimedia.org/r/#/c/112661/ [08:23:40] ori: Or should I deploy it somewhere? [08:23:43] (03CR) 10Frédéric Wang: [C: 031] Push math nodes onto mathArray individually to avoid problem in IE < 9 where push.apply doesn't work on a nodeList. Also, avoid duplicate el [extensions/Math] - 10https://gerrit.wikimedia.org/r/113764 (owner: 10Physikerwelt) [08:24:00] prtksxna: no need, there's a vagrant role! [08:24:47] ori: (y) [08:25:33] ori: There is one more message (in that thread) that I sent, but it was over email (I really should have sent it on Trello, but I was getting a bit bugged by "Board" and "Cards") :( [08:26:34] an Extension:KanbanBoard would make a lot of people very happy, I think [08:26:45] Extension:KanbanBoardv5 [08:26:54] * YuviPanda waves at ori [08:26:56] a wild YuviPanda appears! [08:27:00] * ori re-waves [08:27:11] * YuviPanda particles [08:27:12] * prtksxna throws his Pokéball at the YuviPanda [08:27:22] * prtksxna captures YuviPanda [08:27:38] \o/ [08:27:42] prtksxna: heh, that's all the actions of pokemon I know of :) [08:27:49] ori: saw that you'll be coming to Zurich. Yay :) [08:29:13] prtksxna: is ori taking care of your Eventlogging patch? [08:29:25] prtksxna: (Ori single-handedly wrote the entire EventLogging infrastructure :) ) [08:29:45] YuviPanda: Yes, he taking a look at it. legoktm and ori also suggest I don't use userId [08:29:49] ah [08:29:52] and username instead? [08:30:07] YuviPanda: No https://gist.github.com/atdt/9046669/raw/a5c05f7bc8d03c82ba00e3f225f1cc306de980a8/getId.js [08:30:21] oooh, sessions [08:30:45] hmm, they are distinctly different from users, but I suppose it is good enough for what we need to do [08:31:05] ori: Any ideas? I'll just switch out userId with sessionId? [08:31:09] YuviPanda: Right [08:31:28] prtksxna: sorry i got distracted by a mediawiki-vagrant wart and and filed a bug so i don't forget about it (https://bugzilla.wikimedia.org/show_bug.cgi?id=61456) [08:31:30] YuviPanda: Did you get a chance to test https://gerrit.wikimedia.org/r/#/c/113090/ [08:31:31] still looking [08:31:55] ori: Oh, ok (y) [08:32:13] prtksxna: no, haven't. Surprisngly I've not touched the laptop at all in the last two days [08:32:20] YuviPanda: :O [08:32:24] YuviPanda: What have you been up to? [08:32:38] prtksxna: went to a completely non-tech event and met lots of people :) [08:33:00] (03CR) 10Olenz: [C: 031] Added functions to access mime type and subtype separately [extensions/MediaFunctions] - 10https://gerrit.wikimedia.org/r/47438 (owner: 10Olenz) [08:33:51] YuviPanda: Oh! Which one? [08:34:11] prtksxna: http://www.tatafirstdot.com/ [08:34:27] prtksxna: I helped start it 3 years ago in Chennai when I was in 3rd year. Others have taken and run with it :) [08:34:28] YuviPanda: Also, LIAR! Your idle time was 1d 5h! :P [08:34:42] prtksxna: I changed nick saturday night :P [08:35:07] YuviPanda becoming well-adjusted will be devastating to the WMF's overall engineering output [08:35:09] Nice [08:35:55] ori: I know! My overall commit count has decreased over the past few months, and so has the diversity (almost nothing on ops/puppet in that time) [08:36:24] YuviPanda: i kid; having a life is better. [08:37:07] ori: not so sure :P but it's ok. Having a 'oh I might move out of the country anytime now' sword hanging over your head for about 14 months constantly kills most attempts at having a life :P [08:37:15] I should be back to normal soon! :D [08:37:54] YuviPanda: So, are you leaving the country? [08:38:13] prtksxna: answering the same way for the last 14 months, 'I might but I do not know when' [08:38:28] YuviPanda: But now you've cleared the one thing that was stopping you! [08:38:32] it's a lot more frustrating than I thought [08:38:34] prtksxna: true, true. [08:38:51] YuviPanda: The internal process or the external one? [08:38:59] ori: Do let me know :) [08:39:07] prtksxna: the unintended side effects, more like :) [08:40:09] YuviPanda: icy [08:41:41] (03CR) 10Yuvipanda: [C: 032 V: 032] "No jenkins setup for this yet :)" [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113766 (owner: 10Prtksxna) [08:42:28] prtksxna: if you just want to group together events generated within the context of a single page view, you don't need persistence at all [08:42:57] ori: That is true -_- [08:43:17] you can either just assign the return value of mw.user.generateRandomSessionId() to a variable and include it in all events [08:43:22] ori: Right [08:44:01] or you could accumulate popups views and only send a single, aggregated event on a beforeunload handler [08:44:11] *in a [08:44:19] ori: I think I'll go with the 1st one, any tradeoffs? [08:45:02] i can't think of any, but i'd be lying if i said i fully understood what question(s) you're trying to answer with the data [08:45:43] ori: Alright, but there is no technical difference in either of the two approaches, right? [08:46:03] well, the latter means fewer network requests overall [08:46:23] ori: Right… [08:46:36] but they're background requests which occur after the page's content has loaded [08:46:38] ori: is doing it in onbeforeunload reliable? [08:46:46] (just curious) [08:46:48] so as long as you're careful to retrieve the popup contents before you submit the event you should be ok [08:46:58] * hashar is upgrading Jenkins ... Half an hour downtime. [08:47:04] YuviPanda: it's been a while since i tried but i think so [08:47:21] ori: Thanks for all your help! I'll talk to Jared and Dario and let you know what comes of it :) [08:47:32] ori: Do you mind reviewing the code (once its done) [08:47:44] no, feel free to add me as a reviewer [08:48:33] ori: unrelated, but would layout thrashing occur if I'm doing lots of read/write ops on an unattached, just created DOM element? Googling didn't help. [08:48:46] ori: just a vague general question I'm interested in, no rush :) [08:49:04] (since we are doing lots of HTML transforms in JS in the app, but they are done before they are attached) [08:49:13] btw don't let lofty thoughts from analysts delay the initial roll-out; you can basically assume that the first round of data that is collected will be inadequate in some way (it is usually the case that some omission is realized only at the time of analysis) [08:49:56] YuviPanda: no layout thrashing [08:50:43] ori: hmm, sweet! [08:52:11] ori: Hmmm. Ok! [08:59:40] (03Restored) 10Hashar: tests to compare En, Qqq and messages.inc [core] - 10https://gerrit.wikimedia.org/r/5550 (owner: 10Hashar) [09:16:05] (03PS12) 10Hashar: tests to compare En, Qqq and messages.inc [core] - 10https://gerrit.wikimedia.org/r/5550 [09:16:27] (03CR) 10Hashar: "Rebased / fixed conflict. I have restored this change following a mail exchange with MarkTraceur" [core] - 10https://gerrit.wikimedia.org/r/5550 (owner: 10Hashar) [09:18:11] (03CR) 10jenkins-bot: [V: 04-1] tests to compare En, Qqq and messages.inc [core] - 10https://gerrit.wikimedia.org/r/5550 (owner: 10Hashar) [09:28:57] heh, I check wikitech-l now and then and all I see seems to be drama for now. [09:35:43] (03PS11) 10Prtksxna: Add EventLogging [extensions/Popups] - 10https://gerrit.wikimedia.org/r/112661 [09:37:45] (03PS15) 10Prtksxna: Add `redirects to` message at the top of the popup [extensions/Popups] - 10https://gerrit.wikimedia.org/r/111983 [09:39:48] (03PS3) 10Prtksxna: Position timestamp div from the bottom [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113090 [09:41:03] (03PS7) 10UltrasonicNXT: Lots of additions [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [10:00:17] (03CR) 10Daniel Kinzler: [C: 032] Add ParserOutput::addJsConfigVars [core] - 10https://gerrit.wikimedia.org/r/108333 (owner: 10Umherirrender) [10:00:23] ok, so the jenkins-bot build failed on my patch because of php 5.3 syntax. should i fix it now and "git review" again, or should i leave it like this (because the error is superficial)? [10:04:02] (03Merged) 10jenkins-bot: Add ParserOutput::addJsConfigVars [core] - 10https://gerrit.wikimedia.org/r/108333 (owner: 10Umherirrender) [10:06:33] (03PS4) 10Ori.livneh: ResourceLoader: pass jQuery twice to 'implements' module wrapper [core] - 10https://gerrit.wikimedia.org/r/85637 [10:06:49] (03CR) 10Ori.livneh: "krinkle, ping" [core] - 10https://gerrit.wikimedia.org/r/85637 (owner: 10Ori.livneh) [10:07:05] (03CR) 10Daniel Kinzler: "There's a test for rollbacks in WikiPageTest, please add a case that covers the new behavior." [core] - 10https://gerrit.wikimedia.org/r/102550 (owner: 10Umherirrender) [10:09:47] (03Abandoned) 10Prtksxna: [WIP] Truncate extract to 250 characters [WIP] [extensions/Popups] - 10https://gerrit.wikimedia.org/r/113063 (owner: 10Prtksxna) [10:35:12] (03CR) 10Krusher: "I ended giving you more work than enhacements, lol." (031 comment) [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [10:36:13] (03CR) 10Krinkle: [C: 031] "Don't forget the other commit." [core] - 10https://gerrit.wikimedia.org/r/85637 (owner: 10Ori.livneh) [10:37:35] Krinkle: have you seen mwgrep yet? [10:38:23] ssh to tin and run: mwgrep mw.user.anon [10:40:40] ori: ooowwwm, super handy... thanks :) [10:41:09] I've been using a hacky bash script for that until now :P [10:41:38] how did it work? [10:42:06] ori: It basically just pulled in all JS pages I wan interested in (mostly common.js) and ran grep on them :P [10:42:11] * was [10:42:14] ahh [10:48:49] ori: how performant is that (how much is search engine, how much is local script), you think we could expose this publicly somehow? [10:49:28] I wrote two separate toolserver tools over the past years to do essentially that, all of them extremely hacky and slow. [10:49:44] hoo@terbium:~$ time mwgrep 453894375dfs [10:49:44] [...] [10:49:44] real 0m0.176s [10:49:52] the local script is basically constructing the URL for a request; it's all elasticsearch [10:50:09] couldn't use any API or search backend for it at the time, as it mangled code quite badly (punctuation normalisation etc.), and no way to search multiple wikis [10:50:11] nik (manybubbles) said the load it generates is laughably small [10:50:28] yeah, looked at the source code already [10:50:41] wasn't sure where ""_source['text'].contains('%s')" % args.term" ran though [10:50:42] ori: Does that service used in there have a public host? [10:50:58] http://search.svc.eqiad.wmnet:9200 [10:51:09] you could change it to a regexp search and as long as it was restricted to the mw ns it would be fine (again, per nik) [10:51:53] it's not publicly-accessible; pathological queries will run it into the ground, so you need to have a layer in front that restricts the types of searches [10:52:02] in our case, the mediawiki search interface [10:52:43] ori: Hm.. it currently doesn't take regex though, right? Would need a minor change to the script. Just verifying I'm not using it incorrectly [10:52:56] yeah, doesn't take a regexp at the moment [10:53:29] because this particular query is not risky (because ns:8, title ends with 'js' or 'css' restricts the pool of candidates), we could have a public front-end for it, but chad's plan is to make the entire elastic index queryable from labs [10:53:40] hoo: does this help the non-protocol-relative scripts searches? [10:54:07] Nemo_bis: Not much w/o regexp support, yet [10:54:13] ok [10:54:29] But I already have quite a list of stuff to clean up which I gathered using my old bash script [10:55:52] Krinkle: http://p.defau.lt/?Yswucu3jR5S9QqZfFre7Vg is the query (first line is path, everything below is the json query, which should be sent as POST payload) [10:56:56] Krinkle, hoo: you can replace 'contains' with 'matches' to make it a regexp search [10:57:29] ori: what script language is that? [10:57:44] python [10:57:48] http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-script-filter.html [10:57:56] ok [10:58:20] (03PS8) 10UltrasonicNXT: Lots of additions [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [10:58:24] Krinkle: no, MVEL [10:58:26] [10:58:36] context: http://p.defau.lt/?EZ6aPeZRMWdjZtwpOnFgWQ [10:58:53] yeah, it says that [10:59:05] scripts -> http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-scripting.html [10:59:09] uses by default mvel [10:59:24] ah, yep [11:00:27] ori: mh... regexp don't seem to work after I did the change you suggested... [11:00:34] or aren't that perl-style regexp? [11:02:18] Im trying to find docs on mvel regarding available string methods [11:02:20] no luck [11:03:18] http://mvel.codehaus.org/MVEL+2.0+Operators shows a different syntax ( foo contains "x", and, foo ~= '[a-z].+' ) [11:03:47] Krinkle: "By default, like Java, MVEL provides all java.lang.* classes pre-imported as part of the compiler and runtime." [11:03:54] right [11:04:23] so since _source['text'] is a string, you can call http://docs.oracle.com/javase/6/docs/api/java/lang/String.html#matches(java.lang.String) [11:05:06] I don't know how that is functionally different from using MVEL's ~= operator [11:05:10] Ah, this is going to be important to document and protect against. Working 3 or 4 levels of abstraction deep to end up evaluating java. [11:05:22] java, mvel, json, http [11:05:52] elastic's http interface is designed to be a private api tho [11:06:00] yeah [11:06:10] but yeah, it is a bit like a matryoshka doll [11:06:29] https://upload.wikimedia.org/wikipedia/commons/4/45/7babushkadolls.JPG [11:06:33] except that its bigger on the inside [11:06:34] i'll let you decide which one is java :P [11:06:51] http < json < mvel < java [11:07:16] well, kinda sorta [11:07:17] nvm [11:07:41] don't forget shell-style argv parsing > json [11:08:06] yeah, but out final API isn't going to do that :P [11:08:31] at least i hope not [11:08:59] ApiAwesome.php, $apiResult = `"/bin/mwgrep escapeshellarg($_GET['q'])"` [11:09:25] I'm sure we could come up with something more hideous than that [11:09:33] it needs Puppet [11:09:41] and ideally also XML [11:10:00] and php serialise/wakeup [11:10:37] see also for the "proper" way to do code searches without resulting to this sort of trickery [11:11:05] (03PS9) 10UltrasonicNXT: Lots of additions [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [11:12:17] basically teach lucene to run against the text and index names [11:12:20] (03CR) 10UltrasonicNXT: "Ahah, no prob :) Good point there" [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [11:12:35] * ori chuckles at 'import com.google.caja.SomethingWidgyHappenedError;' [11:13:56] If ElasticSearch can be used for proper code searches, maybe gitblit can be convinced to use it instead of Lucene? [11:14:23] If yes please file a bug upstream, the request for gitweb-like grep search was rejected. [11:14:37] elastic uses lucene under the hood [11:16:04] I've been poking at its overall performance (), it's hideously slow atm [11:18:37] I know it uses Lucene, hence I'd think it would be slightly easier to convince them to adopt it. :) [11:19:21] Oh, thanks for the link, I thought I had starred that one. [11:22:47] And thanks for working on that [11:28:51] (03Abandoned) 10Hoo man: Allow using LuaEngineTestBase without having to set test modules [extensions/Scribunto] - 10https://gerrit.wikimedia.org/r/111108 (owner: 10Hoo man) [11:32:07] (03CR) 10Hoo man: "I can only agree with Daniel... amending the unit tests would be great :)" (031 comment) [core] - 10https://gerrit.wikimedia.org/r/102550 (owner: 10Umherirrender) [11:39:27] (03CR) 10Hoo man: [C: 04-1] "Needs a manual rebase." (032 comments) [core] - 10https://gerrit.wikimedia.org/r/109903 (owner: 10Lekshmi) [11:43:55] (03CR) 10UltrasonicNXT: "Re your comment on 7, sorry, forgot to publish my reply, but yeah, that's a good point. Probably something for another patch though :) I c" [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [11:45:31] (03CR) 10UltrasonicNXT: [C: 032 V: 032] Lots of additions [extensions/MediaWikiChat] - 10https://gerrit.wikimedia.org/r/113164 (owner: 10Gerrit Patch Uploader) [11:58:26] (03PS1) 10Hashar: Comment for parsoidsvc-php-parsertests [integration/jenkins-job-builder-config] - 10https://gerrit.wikimedia.org/r/113768 [11:58:35] (03CR) 10Hashar: [C: 032] Comment for parsoidsvc-php-parsertests [integration/jenkins-job-builder-config] - 10https://gerrit.wikimedia.org/r/113768 (owner: 10Hashar) [11:59:03] (03Merged) 10jenkins-bot: Comment for parsoidsvc-php-parsertests [integration/jenkins-job-builder-config] - 10https://gerrit.wikimedia.org/r/113768 (owner: 10Hashar) [12:09:27]