[00:00:33] >>Pages blanking after completion of loading, noted at en.wp's Village Pump (technical) [00:00:56] not being able to search in Help:Contents seems like a big issue, but I had a friend try it with a modern IE and it was OK [00:01:16] BarkingFish: document.write() after page load [00:01:19] BarkingFish: That's probably a document.write problem [00:01:36] I think there's a bug for that but any gadget could do it [00:02:14] it'd be easy enough to override document.write to throw an exception after page load [00:02:24] https://bugzilla.wikimedia.org/show_bug.cgi?id=34542 is for document.write [00:02:29] Yeah, someone's told the reporter to disable their gadgets and try again, also to note their browser when they reply. I'll let you know what comes back from it TimStarling [00:02:44] thanks BarkingFish [00:02:47] I remember a presentation from wikia about doing that to allow asynchronous loading of partner ad scripts [00:03:01] better still, since there's a bug already for it, I may note it there instead. But I'll still let you know [00:03:11] ah one instance of [00:05:05] TimStarling: got a Collection test environment? Should I push it on my brand new trunkwiki? [00:05:11] !log tstarling synchronized php-1.19/extensions/Collection/Collection.body.php 'r112745' [00:05:16] Logged the message, Master [00:05:16] too late [00:05:26] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 0 C: 2; - https://gerrit.wikimedia.org/r/2903 [00:05:31] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2903 [00:05:37] wow, 265MB of mariology [00:05:42] * robla goes to try it on Tim's test environment :) [00:05:50] fwiw, just looked at a histogram of enwiki slow parse times from the 22:00 hour, and >23:10 and they're identical to the nearest 100ms (50% 4.7sec, 99% 24.8sec) and hitting the slow log threshold at the same rate [00:05:52] it's downloading for me, I think it's working [00:06:02] no wonder it went OOM [00:06:52] Ouch [00:07:13] Just updating you, turns out the page blanking WAS a gadget, revisionjumper. The guy who reported is on Opera. [00:07:26] heh [00:07:27] The code for that gadget is on de.wp [00:07:37] files that size could easily cause exhaustion of /tmp space [00:08:51] Another reason to get the /tmp issues sorted out [00:09:19] maybe we can delete /a and carve out a new /tmp partition from it [00:09:58] robla: just fyi I'm running a test of the swift deletion script for about 10 minutes. It shouldn't affect anything you're doing. [00:10:56] any ops people here who can comment on that idea? [00:10:59] thanks for the warning [00:11:30] maplebed: ^ TimStarling's question is kinda apropos of our conversation earlier this week(?) [00:11:45] woot, 743 pages of Mariology open in my Document Viewer [00:11:48] yeah, there's email talking about it. [00:11:55] think I should print it just to make sure that works too? [00:11:59] :p [00:12:25] impressive that the whole system can handle files that large except for that one line of MW code [00:12:48] try printing the Wikipedia docs [00:12:53] 1 PHP Warning: parse_url(://incubator.wikimedia.org/wiki/W/?goto=mainpage) [function.parse-url]: Unable to parse URL in /usr/local/apache/common-local/wmf-config/missi [00:12:53] ng.php on line 122 [00:13:01] 1 PHP Warning: parse_url(http://mobile.wikipedia.org:80, mobile.wikipedia.org/?useformat=mobile) [function.parse-url]: Unable to parse URL in /usr/local/apache/common-local/wmf-config/missing.php on line 52 [00:13:37] * Reedy wonders what some people are taking [00:13:43] good stuff [00:14:05] Forward? [00:14:20] yes [00:14:27] I say so... TimStarling? [00:14:52] <^demon> Evening folks. Do I get a vote? [00:14:52] Reedy: Whatever it is, right now, I could use some :) [00:15:02] ^demon: no [00:15:17] You weren't here for registration [00:15:37] Roan doesn't have a vote either [00:15:41] <^demon> Ok, I'll just go back to Voyager reruns and cake. [00:15:42] ^demon: you represent DC [00:15:51] ^demon: do you like that show? [00:16:03] <^demon> robla: DC only gets non-voting observer status :( [00:16:25] virtual representation, yes, it's not like that didn't work out before [00:16:57] let's keep going [00:17:00] Segfaults.. [00:17:01] de? [00:17:17] well....explain your Segfaults...comment first [00:17:30] Just there at the top of the fatal log [00:17:42] we get it from time to time anyway, and it's not flooding [00:18:00] * chrismcmahon BZs 3 IE7 issues and leaves the Teahouse one alone.  [00:18:13] <^demon> e_warning: parse_url() doesn't seem to like protocol-relative urls. [00:18:20] TimStarling: which revision of Collection is now sync'd? (checking my Special:Version problem in bug 34796) [00:18:25] OOM causes segfault [00:18:35] chrismcmahon: you may want to ping heatherw on one of the other channels about the Teahouse problem [00:18:49] ^demon: or with ports, spaces and commas [00:18:52] r112745 [00:19:05] !r 112745 [00:19:06] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/112745 [00:19:11] robla: my guess is that IE7 is not critical to Teahouse. [00:19:16] I was replying to saper [00:19:18] <^demon> Reedy: Well spaces are just weird...you shouldn't do that. [00:19:28] chrismcmahon: worth checking [00:19:29] Sorry [00:19:32] I won't do it again [00:19:39] robla: will do [00:19:43] <^demon> The collective "you" :) [00:19:52] * Reedy stops trying to break wikipedia [00:20:08] en garde! [00:20:44] Auf Deutsche [00:20:54] ^demon: you're not watching Threshold are you? [00:21:06] TimStarling: am I the only one to see (r112130) there in https://en.wikipedia.org/wiki/Special:Version (logged in) [00:21:19] <^demon> AaronSchulz: Oh lord no. I made Mike watch that the other day though. [00:21:40] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: dewiki to 1.19wmf1 [00:21:43] Logged the message, Master [00:21:53] Doesn't the version depend on the .svn dir? [00:22:03] yes [00:22:19] saper: "1.19wmf1 (r112130)" [00:22:25] It's fine [00:22:33] well, partly, there is a global in defaultsettings with the MW version [00:22:57] I was meaning the svn rev-ness [00:23:06] yep [00:24:01] admin report of broken user script on enwiki: http://en.wikipedia.org/wiki/User:PleaseStand/userinfo.js <- RoanKattouw, if there's something really obviously broken to drop as a comment, I'm sure the authors+users would appreciate it [00:26:12] robla: Tried it just now from the console in Chrome on [[User:Krinkle]] (the script only acts on NS_USER root pages), working fine [00:26:21] need more exact report [00:26:52] It adds "A reviewer and rollbacker, 3 years old, with 1,258 edits. Last edited 3 days ago." to the tag line below the page name' [00:27:58] Reedy: mass tag removal is broken in CR [00:28:12] oh, nvm, I see [00:28:22] I was going to say, it worked yesterday [00:28:36] it helps to use the checkboxen [00:28:38] Krinkle: Fluffernutter was the original reporter [00:28:45] :D [00:28:46] New patchset: Ryan Lane; "Seems wikidev *is* required" [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2904 [00:28:53] hallo [00:29:04] Fatal log is just 1.18 stuffs. Onwards? [00:29:19] Fluffernutter: (04:26:11 PM) Krinkle: robla: Tried it just now from the console in Chrome on [[User:Krinkle]] (the script only acts on NS_USER root pages), working fine [00:29:26] more more more [00:29:34] Reedy: yup [00:29:47] fr [00:29:47] fr, ja or ru? [00:29:48] hm, robla/Krinkle, let me try it in chrome instead of FF [00:30:04] sure, fr [00:30:10] fr next? cool [00:30:13] nope, still getting nothing for it [00:30:33] possibly interaction with other gadgets and stuff? [00:30:44] Fluffernutter: Any errors ? [00:30:51] Fluffernutter: Works in Firefox for me as well [00:31:05] Fluffernutter: Do you know how to access to the console in Chrome or Firefox ? [00:31:26] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: frwiki to 1.19wmf1 [00:31:29] Logged the message, Master [00:31:31] Krinkle, no error, just it doesn’t do what it, you know, does (“should be loading a sub-headline on userspace pages showing me user age, status, etc - instead I just see the usual “from wikipedia, the free encyclopedia”) [00:31:43] A reviewer and rollbacker, 3 years old, with 1,258 edits. Last edited 3 days ago. From Wikipedia, the free encyclopedia. [00:32:07] Krinkle, alas no. Reporting that the script doesn’t work is pretty much the extent of my technical/script knowledge [00:32:16] Fluffernutter: OK [00:32:43] Fluffernutter: Could you ask the author of the script to look at it and if he needs help, point him to [[mw:Talk:RL/MGU]] and [[mw:RL/MGU]] [00:33:28] shall do [00:33:34] Th [00:33:36] Thx :) [00:33:50] Feel free to send me the link to your message so I can add it to my watchlist [00:35:08] Sorry guys, just a friendly note - You might want to look at the color scheme on diffs again, or sort out some way round the changes, user in #wikipedia-en with colorblindness issues reporting that the new color scheme is causing issues [00:35:23] "colors are blending into one another" [00:35:53] Krinkle: http://en.wikipedia.org/wiki/User_talk:PleaseStand#Your_userinfo_script . Cheers! [00:36:48] BarkingFish: Interesting, as that was exactly what the report said that requested to make the change in the first palce [00:36:50] place* [00:37:15] BarkingFish: Krinkle IIRC there was a bug report about that [00:37:21] Yes, and it was fixed [00:37:27] and now someone else is having issues [00:37:41] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 0 C: 2; - https://gerrit.wikimedia.org/r/2904 [00:37:45] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2904 [00:37:50] * Krinkle joins #wikipedia-en  [00:38:05] Wasn't that a specific colour blindness? [00:38:13] to be honest, you might be wise to remove the colors from diffs altogether, and go to the +++ and --- for additions and removals, and just use clear black text against a white background [00:38:58] * AaronSchulz detects sarcasm [00:39:08] None intended [00:39:38] you know colourblind people are able to see things that are coloured [00:39:49] lolz [00:39:50] they just can't tell the difference between the different colours [00:40:05] right, but they can't be symbol blind, TimStarling :) [00:40:06] the bolded black text on dark blue background of changed text really does make it hard to read. [00:40:12] no colors - no issue :) [00:40:15] ;) [00:40:30] TimStarling: some people actually do have "monochrome" vision, but that's super rare [00:40:58] Diffs already have + and - [00:41:00] that aside, this change violates w3c accessibility guidelines in a few ways [00:41:02] sure, and then they would have trouble reading text where the background and foreground only differs in hue [00:41:12] <^demon> TrevorParscal: Is there actually a combination of colors that doesn't screw with anyone? [00:41:14] but everyone finds such text difficult to read, that's why we don't do that [00:41:17] http://screencast.com/t/KWo8EdW9dXPR [00:41:31] ^demon: the rule is not to rely on color alone [00:41:44] that's broken rule #1 [00:41:45] in our last episode: https://bugzilla.wikimedia.org/show_bug.cgi?id=33335 [00:42:04] #2 is that the contrast ratio between black and the background color applied to changed portions is too low [00:42:19] personally I think it was not really necessary to optimise the colour scheme for r/g colourblind people in a case where the colour carried no useful information [00:42:49] actually, here's the monster debate: https://www.mediawiki.org/wiki/Special:Code/MediaWiki/105280 [00:43:00] you can tell which one is the old version and which is the new version by looking at what side they are on [00:43:17] good point, TimStarling [00:43:21] maybe someone can use the collections extension to create a nice printout of https://www.mediawiki.org/wiki/Special:Code/MediaWiki/105280 [00:43:39] TimStarling: i think the problem is that it used to be better [00:44:03] we used to use bold + colored text (not using only color) and the contrast ratio was good enough [00:44:25] now we've broken those 2 accessibility rules, and people are complaining legitamately [00:44:43] TrevorParscal: next step, breaking them all [00:44:52] New patchset: Ryan Lane; "Add a way to avoid installing apache" [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2905 [00:44:52] Reedy: srsly [00:45:13] <^demon> TrevorParscal: Would've been nice if you'd shared those tidbits when it was getting bikeshedded to hell and back. [00:45:19] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 0 C: 2; - https://gerrit.wikimedia.org/r/2905 [00:45:21] <^demon> (not being sarcastic) [00:45:22] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2905 [00:45:26] here's what we need to do imo - revert this change, and then have someone who knows what they are doing with accessibility issues come up with a better patch [00:45:46] ^demon: I'm sorry I missed this issue [00:46:06] ^demon: i agree, I would have been more useful early on [00:46:28] Reedy: AaronSchulz: TimStarling: should we move on? [00:46:31] 'we used to use bold + colored text " [00:46:36] has that actually changed? [00:46:36] Actually, there is a user change someone's done for css which was pointed out on our VP, that might allow users to customize it, TrevorParscal [00:46:41] robla: sure [00:46:43] <^demon> Only 5 revs to back out, and they're all the latest. [00:46:48] <^demon> TrevorParscal: ^ [00:46:48] robla: yes [00:46:57] ru or ja? [00:47:01] ja [00:47:02] robla: is it a corollary to Godwin's Law - whoever invokes Star Wars comparison, automatically wins? [00:47:17] ^demon: seems easy to do [00:47:50] saper: not sure I can get behind that [00:48:22] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/105280#c27840 [00:48:29] rainman-sr: :) hello [00:48:30] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: jawiki to 1.19wmf1 [00:48:33] Logged the message, Master [00:48:48] New patchset: Ryan Lane; "Removing non-ldap accounts from formey and manganese" [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2906 [00:48:49] saper, hello! :) [00:48:57] TrevorParscal: The red on yellow/green dark ground was a genuine problem though [00:49:09] no need to use bright red [00:49:31] Krinkle: I'm not defending the old version beyond saying it didn't obviously break W3C Accessibility guidelines [00:49:37] yeah [00:49:44] Krinkle: You and I could come up with something better that also follows those rules [00:49:49] There's also the matter of detecting whitespace [00:49:55] previously impossible, now possible [00:49:57] we used them when we did the Vector colors [00:50:00] but contrast is too low, for sure [00:50:09] Yep [00:50:20] yeah - so can we just back these out and then let Krinkle and I sort it out tomorrow? [00:50:43] I'm sure we could come up with a patch that could address the real problems without adding new ones [00:50:46] reverting that will get a tail in CR probably, but hey.. [00:50:54] New patchset: Ryan Lane; "Checking for no-apache gerrit config properly" [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2907 [00:50:58] maybe, but I think you should try to be tactful [00:51:04] <^demon> Krinkle: 109932, 108975, 107127, 106884, 105280 [00:51:17] brandon seemed to be pretty sensitive about it on the ML [00:51:25] TimStarling: who are you advising to be tactful? [00:51:40] you, if you're going to back it out [00:51:49] ^demon: what's that ? [00:51:58] Do I also have to be the one who backs it out now? [00:52:03] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 1 C: 2; - https://gerrit.wikimedia.org/r/2906 [00:52:05] or if you're thinking about posting to wikitech-l with a lengthy rant about how awful the new colours are ;) [00:52:06] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2906 [00:52:07] istr that the blue/yellow scheme was brought over from frwiki, someone might be attached to it [00:52:09] Oh, I missed a number in the rev you pasted ^demon [00:52:13] now I got it [00:52:14] Thanks :) [00:52:24] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 0 C: 2; - https://gerrit.wikimedia.org/r/2907 [00:52:28] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2907 [00:52:33] or Krinkle if he's doing it [00:52:47] <^demon> Someone who's willing to own it should do it. [00:52:52] * robla +1's Tim's call to be tread lightly [00:52:52] TimStarling: the last thing I would want to do is post anything about design on that list - omg no [00:52:57] chrismcmahon: not quite frwiki [00:53:01] I was about to revert all diff style changes from 1.19 back to 1.18, and re-open the bug for Trevor and I to look at this week and come up with a better one. [00:53:03] they still had green [00:53:04] <^demon> I couldn't care one way or the other what the colors are, so I'm not going to revert. [00:53:20] * AaronSchulz is not reverting [00:53:24] not it [00:53:37] I'm just offering advice here, this is not my crusade against anything - it's a bad patch that slipped through the cracks and needs to be reverted and redone - I'm offering to come up with a proper patch, that's it [00:53:38] * AaronSchulz checks that jawiki is not on fire while we are arguing [00:53:43] Krinkle: it's ok, just phrase your commit message carefully [00:54:20] AaronSchulz: Looks ok, page loads are a little slow, but that may just be a peak from the switchover [00:54:30] I wouldn't say it slipped through the cracks, lots of people looked at it [00:54:31] TimStarling: thx [00:54:55] yeah, it got a lot of attention [00:55:02] nvm, ja.wp is picking up speed now [00:55:12] <^demon> And now we're reverting because Trevor told us we painted the bikeshed the wrong color :) [00:55:52] omg - it's not my "preference", I'm talking about how it violates clear accessibility guidelines [00:56:01] this is not bike shedding [00:56:02] ^demon: hey bikes are important to some cultures ;) [00:56:20] <^demon> Yeah like those crazy Dutch. [00:56:22] New patchset: Ryan Lane; "I hate our apache manifests, so, so much." [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2908 [00:56:35] You know what? Forget I said anything. [00:56:50] and yes, it slipped through the cracks in that, nobody with the right knowledge caught it, and that's unfortunate - I should have been more attentive I think [00:56:51] * AaronSchulz first read the 'omg' like how a giddy valley girl says it...force of habit :) [00:56:57] <^demon> TrevorParscal: I'm being sarcastic because I don't care if they're fuchsia and cat poo colored ;-) [00:57:09] I didn't intend to start a war on color, I mentioned one small thing. Jeez... [00:57:12] ^demon: you mean those Roan-y people? [00:57:14] TrevorParscal: brandon used color-blindness tools when doing this. where's the accessibility guidelines you are referencing? [00:57:16] * robla throws in a "tubular!" for Aaron [00:57:17] * BarkingFish bangs his head on his keyboard [00:57:19] i don't expect TimStarling to know about UI accessibility, I'm not hating on people here [00:57:24] Most of our development team are visually blind [00:57:29] ^demon: lolz [00:57:29] I wonder if this is a combination of colorblindness and poor vision? [00:57:30] e.g. CodeReview is fine [00:57:53] I know people with poor vision have issues with pale colors [00:57:56] hey, don't bring me into it [00:57:59] <^demon> Actually the puke yellow/green of gerrit is growing on me. [00:58:25] sorry I wasn't involved earlier, I'm happy to lend a hand to fix it up [00:58:26] anyway, on the non-diff color front, anyone notice any performance probs or weirdness in logs? [00:58:32] I have had absolutely nothing to do with this except for my comments on IRC in the last 10 minutes [00:58:33] TimStarling: Just keep your head down and mind the bullets, I'm trying to do the same thing :) [00:58:47] I wish I'd never have spoken, frankly [00:58:55] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 0 C: 2; - https://gerrit.wikimedia.org/r/2908 [00:58:58] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: Meanwhile, on wikipedia.... Hello ruwiki! [00:59:01] Logged the message, Master [00:59:04] New review: Ryan Lane; "(no comment)" [operations/puppet] (production); V: 1 C: 2; - https://gerrit.wikimedia.org/r/2908 [00:59:06] Change merged: Ryan Lane; [operations/puppet] (production) - https://gerrit.wikimedia.org/r/2908 [00:59:10] TimStarling: sry, using you as an example of a non front-end person didn't seem too insane, wasn't trying to "drag" you into anything [00:59:29] robla: looks ok [00:59:40] robla: we can probably look at doing the remains of s2 followed by s7 remains [00:59:44] * TrevorParscal moves on to something else [00:59:49] Reedy: sounds reasonable [00:59:54] * ^demon goes back to his git bunker [01:00:00] So is enwiki 1.19 now? [01:00:07] yup [01:00:09] TrevorParscal: too late, it's on like donkey kong! [01:00:10] * Reedy follows ^demon to our glorious git future [01:00:15] * BarkingFish moves onto the nearest railway line and waits for a train :P [01:00:16] not really ;) [01:00:20] Reedy: https://en.wikipedia.org/wiki/MediaWiki_talk:Remembermypassword [01:00:27] Reedy: Can you update that, please? [01:00:45] Though I'm really not sure it was a good idea to add the value at all. [01:00:51] I guess people kind of expect it now. [01:00:53] doned [01:00:59] ty sir [01:01:35] robla: fatal log is all quiet on the 1.19 front [01:01:48] sweet....what all is on s2? [01:02:10] I wonder if there's been a measurable decrease in the number of new redirects after removing the button from the editing toolbar. [01:02:17] nm...found the list [01:02:35] zhwiki will be the interesting one I guess [01:02:40] yup [01:03:16] yeah, let's do that one, and then maybe start doing them in bulk [01:03:23] Reedy: are you switching fr.wp or has that already been done? [01:03:31] already been done [01:03:31] * BarkingFish lost track [01:03:35] ok, cheers [01:03:35] and ja, and ru [01:04:51] I've just updated wikiversions.dat to move the rest of s2 [01:04:53] * Reedy waits [01:05:09] zhwiki is constantly spamming the logs with OOMs, you wouldn't expect that to change in 1.19 [01:05:44] TimStarling: is there a way we can get fatalmonitor to output the dbname too? [01:06:12] Or is it going to be need redoing from scratch to be less simplistic? [01:06:24] robla: Just to let you know, there's a community notice up on tpi.wp for the update, posted in the local language, I passed it to the stewards earlier to stick up :) [01:06:31] just use /var/log/mw/fatal.log , it has the URL [01:06:48] dbname would be nice for cli fatals [01:07:05] We have too many log files all over the place [01:07:07] brb [01:07:11] It gets somewhat confusing [01:07:16] gn8 folks [01:07:17] s2 then? [01:08:19] sure [01:08:26] I asked for a server for log files, where they would all be consolidated, but mark delayed it, probably by months, by tying it to some other request for getting rid of fenari [01:09:20] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: s2 to 1.19wmf1 [01:09:23] Logged the message, Master [01:10:04] Yeah [01:10:12] I have a half-baked patch in Gerrit for this [01:10:24] We have a server for this now and fatal.log is the first thing I want to move to it [01:10:49] Also because that means it's put in puppet (Mark wouldn't let me puppetize it on fenari), and I wanted to put it in puppet so I can add log rotation easily [01:10:56] Right now fatal.log is freaking huge [01:11:12] RoanKattouw: So you're in SF full-time now? [01:11:20] Yes [01:11:23] what is the server called? [01:11:23] I need to update my mental timezone for you. [01:11:26] heh [01:12:00] Ahm, let's see [01:12:06] I have this /somewhere/ [01:12:42] Oh, fluorine, see RT #2350 [01:13:24] why do you want the fatal logs on the same server as deployment? [01:13:31] that's what I didn't get about mark's comment on my request [01:14:10] Well I want them /somewhere/ [01:14:20] That devs have ssh on [01:14:25] Deployment seemed to make sense [01:14:36] The typical workflow is you deploy a change, then you watch the log [01:15:24] but a deployment server doesn't need any resources [01:15:29] and log storage does [01:15:52] Right [01:15:55] log storage is potentially high bandwidth and has the potential to compromise the performance or stability of other services running on the same server [01:16:02] Hmm, that's true [01:16:06] I'd be happy to store it somewhere else [01:16:20] As long as it's easily accessible from the deployment server, and to people that only have ssh on the deployment server [01:16:41] or well people that have deploy access only I should say, so that means ssh on the Apaches as well [01:16:57] here's my ticket: http://rt.wikimedia.org/Ticket/Display.html?id=2400 [01:17:00] I want apache access logs [01:17:13] I don't really want apache access logs on the same server as anything else though [01:19:02] so nobody talked to you about your request being linked to mine in this way? [01:19:50] RoanKattouw: anyone here who knows wikilove? [01:20:09] Somewhat [01:20:09] * Saibo has never heard of wikilove ;) [01:20:16] robla, AaronSchulz: cu's were complaining about WikiLove storing only IP and no UA in CUlog back in Oct 2011, now we have 127.0.0.1 and no UA. Got it reproduced on my local test trunk wiki. It's an API call. [01:20:18] Oh, there's another one too [01:20:34] TimStarling: https://rt.wikimedia.org/Ticket/Display.html?id=623 [01:20:34] it should use DerivativeRequest or use WikiPage functions [01:20:39] None of these tickets are linked [01:20:48] Oooh [01:20:52] That might be an internal API call? [01:20:59] probably [01:21:02] yes, it's ugly [01:21:04] http://tools.wikimedia.pl/~saper/lovereq.log <-- here it is [01:21:06] ApiWikiLove.php [01:21:07] Probably to do with DerivativeContext [01:21:17] it just uses FauxRequest [01:21:22] two levels of FauxRequest [01:21:31] but the same stuff is on meta which has no WikiLove, must be more generic problem [01:21:33] since the API uses FauxRequest itself to call back to EditPage [01:21:47] :) [01:22:00] Right [01:22:04] All internal requests without UserAgent should be banned! [01:22:13] :) [01:22:16] But if the one in the API was broken, we'd have bigger problems [01:22:22] the API does add some value for WikiLove though, it does make sense to do it that way instead of using EditPage directly [01:22:41] the edit api uses DerivativeRequest as it should [01:23:01] !log reedy synchronized php-1.19/extensions/CategoryTree/CategoryTreeFunctions.php 'r112754' [01:23:06] Logged the message, Master [01:23:58] There's a few more different OOM appearing now [01:24:31] from zhwiki, as expected [01:25:34] Hello [01:25:35] [0125][tstarling@fenari:/home/wikipedia/common/wmf-config]$ ssh root@fluorine [01:25:35] Permission denied (publickey,password). [01:25:39] i have problem [01:25:51] haven't we all [01:26:28] itmard: what's up? [01:26:50] tanks Fixed [01:26:57] that was easy [01:27:19] \o/ [01:27:20] We should assign all bugs to robla from now on [01:27:30] another happy customer^G [01:27:49] s7? [01:28:00] sure....let's press on [01:28:50] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: s7 to 1.19wmf1 [01:28:54] so this is the one tpi.wp falls in? [01:28:55] Logged the message, Master [01:29:01] robla: is there a map somewhere that shows s? to wiki? [01:29:03] BarkingFish: nope [01:29:13] http://noc.wikimedia.org/conf/highlight.php?file=db.php [01:29:17] Everything else is s3 [01:29:19] hexmode: ^ [01:29:25] ty [01:30:21] 265 wikis to go [01:30:36] let's do them in 3 hour intervals [01:31:08] let's not ;) [01:31:27] Reedy: so s7 says centralauth ... wasn't that done with meta? or is CA only happening now? [01:31:39] hexmode: it's just a database, there's no CA wiki [01:31:45] Well guys, it's quite late here for me now. I'm gonna shoot off and get some sleep. Good luck with the rest of the switchover, I'll check the other wikis I work on in the morning in case of any issues :) [01:31:58] Reedy: ah, confusing it with meta [01:32:02] heh [01:32:06] BarkingFish: ty for your help [01:32:09] no probs [01:32:20] hexmode: yeah, we've done all non-Wikipedia sites already [01:32:35] Good night guys, don't work too hard :) [01:32:38] see ya [01:32:41] thanks! [01:34:13] robla: I see that I need to learn more about CA [01:35:15] Reedy: we might as well finish it all off now, I think [01:36:27] Sure [01:37:21] !log reedy rebuilt wikiversions.cdb and synchronized wikiversions files: Last 265 wikipedias over to 1.19wmf1 [01:37:25] Logged the message, Master [01:38:16] OOM in fss_prep_replace indicates that either the tables are too big or the memory limit is too small [01:38:17] * robla pulls up SiteMatrix and starts clicking [01:38:45] http://meta.wikimedia.org/wiki/Special:SiteMatrix for those that want to join the fun [01:39:55] 1 [error] [client 208.80.152.86] script '/usr/local/apache/common/docroot/wikipedia.org/w/INDEX.PHP' not found or unable to stat [01:39:56] 1 [error] [client 208.80.152.74] script '/usr/local/apache/common/docroot/wikipedia.org/w/INDEX.PHP' not found or unable to stat [01:40:09] Someone requesting rubbish? [01:41:57] 1 PHP Warning: Division by zero in /usr/local/apache/common-local/php-1.19/includes/ImagePage.php on line 315 [01:41:57] 1 PHP Warning: Division by zero in /usr/local/apache/common-local/php-1.19/includes/ImagePage.php on line 310 [01:47:31] AaronSchulz: thanks for r112758, testing now... I think it's not only a WikiLove problem though [01:47:52] Looks like it should be [01:47:58] Reedy: yeah, looks like a 404 [01:48:42] * mys_721tx turns on walkie talkie  [01:48:48] #wikpedia-tech, we've got a problem. [01:49:06] Care to enlighten us? [01:49:34] zhwp's zh-hant-zh-hans conversion goes wrong [01:50:02] Goes wrong really doesn't help us [01:50:51] !log aaron synchronized php-1.19/extensions/WikiLove 'deployed r112758' [01:50:54] Logged the message, Master [01:51:08] it only seems to use 55MB to load all the FSS resources [01:51:18] FSS? [01:51:33] fast string search [01:51:33] iirc [01:51:46] well, finally find how to say that.. [01:51:49] the PHP extension that does language conversion [01:52:08] So are we deploying r112750 (diff style revert) ? [01:52:14] TrevorParscal: [01:52:20] (needs review) [01:52:38] Krinkle: looking for someone to share the blame? [01:52:48] well, I can't review and deploy myself [01:53:03] I'll probably have to drag 2 people into it :) [01:53:54] Reedy: more CT errors [01:53:54] Krinkle: lookin [01:54:28] AaronSchulz: Yeah... I'm not sure if we can just if null return array() or something [01:55:02] Presuambly we've got GIGO [01:55:49] mys_721tx: what URL? [01:56:17] TimStarling: it's here http://zh.wikipedia.org/wiki/Special:RecentChanges [01:56:28] the -{}- markup doesn't work anymore [01:56:38] at the top? [01:57:03] Krinkle: http://www.mediawiki.org/w/index.php?title=Special:Code/MediaWiki/112750&path=#c31669 [01:57:05] is it also broken in articles? [01:57:07] PHP Warning: Division by zero in /usr/local/apache/common-local/php-1.19/includes/ImagePage.php on line 316 [01:57:32] AaronSchulz: and on another line. Doesn't help when there's 2 divisions on the same lione [01:57:56] ok zhwiki looks too broken for my liking, I'm going to revert it [01:58:08] TimStarling: I doesn't find it in article [01:58:11] $this->displayImg->getHeight( $page ) is 0? [01:58:11] Are other ZH projects presumably alright? [01:58:35] maybe the page does not exist [01:59:11] I'm reverting them all, I don't know if anyone has looked at the others but there's no reason to think it would be isolated [01:59:14] !log tstarling rebuilt wikiversions.cdb and synchronized wikiversions files: all zh wikis back to 1.18 [01:59:17] Logged the message, Master [01:59:52] TimStarling: you think we need the i18n folks to look at them? [02:00:03] * robla wonders how we're going to fix this [02:00:16] they might have some insight [02:00:24] I will also work on it [02:02:02] wikitech.wikimedia.org cert expired in Jan 2011 :( [02:02:25] !log reedy synchronized php-1.19/resources/mediawiki.action/mediawiki.action.history.diff.css 'r112750' [02:02:28] Logged the message, Master [02:02:55] TimStarling: I'm filing a bug and getting more detail unless you already have? [02:03:27] it was just about every message, look at their edit page: http://tstarling.com/stuff/zhwiki-converter-msgs.png [02:03:32] (for language converter) [02:04:57] TimStarling: could you get one for the same page now? [02:05:33] you mean a screenshot of 1.18? [02:05:38] yes [02:06:03] why do you want that? [02:06:31] so we can have a direct comparison between wrong and right [02:06:52] $opt = FormatJson::decode( $options ); [02:06:56] Reedy: so that fails? [02:07:31] hexmode: I can get that [02:07:37] AaronSchulz: returns null.. so yeah, I think so [02:07:52] really need to see what it's trying to decode [02:07:56] Reedy: want to add a trigger_error fatal? [02:08:02] robla: ty [02:08:12] I'm about to call it a night [02:08:19] that json thing might be a client-side error [02:08:31] yeah, it's possibly we're just being passed crap [02:08:36] you might dump it and find out it's "undefined" instead of proper JSON [02:08:36] yeah [02:08:49] client-side as in JS, not necessarily user fault [02:09:03] in that case it would be useful to have a URL [02:09:09] or reproduction procedure [02:10:12] robla: bug is https://bugzilla.wikimedia.org/34832 [02:11:04] evening all. im trying to open a wikiquote page with python (urllib2.urlopen()). im guessing the 403 that is being returned is popped there on purpose? [02:11:30] AbstractBeliefs: are you giving it a UA? [02:11:31] maplebed is starting the cleanup script now [02:11:41] hexmode: UA? (ie, no) [02:11:45] That's batty [02:11:46] almost now. I've gto to get container listings first. [02:11:51] let's say 10 minutes. [02:11:52] AbstractBeliefs: user agent [02:12:10] !log aaron synchronized php-1.19/extensions/CategoryTree/CategoryTreeFunctions.php 'debug logging' [02:12:14] Logged the message, Master [02:12:34] hexmode: it'll be whatever urllib2 passes by default. this modules is not known for being particularly sane, so it may be the case that no, its not [02:12:53] Place to start then [02:13:00] TimStarling: it's lo behold..."undefined" [02:13:29] AbstractBeliefs: I'm not familar with urllib2 -- google "how do I set the user agent for urllib2" [02:13:44] hexmode: no probs. but that is the issue, yeah? [02:14:08] AbstractBeliefs: pretty sure, yes. if you still have trouble, ask again :) [02:14:12] hey, I know how to set the user agent for urllib2... [02:14:12] unfortunately my psychic debugging skills don't extend as far as actually telling you what line of code the issue is in [02:14:39] TimStarling: im just saying, does no UA = 403. but thats fine [02:14:56] I don't think Tim was replying to you [02:15:09] oops. [02:15:13] AbstractBeliefs: but yes, that sounds right (we give some error for no UA) [02:15:20] maplebed: currently RTFMing over here. if you want to chime in and save my time, please do [02:15:27] AbstractBeliefs: Yes, User agents are required for Wikimedia pages, that's not new in 1.19 [02:15:35] https://meta.wikimedia.org/wiki/User-Agent_policy [02:15:40] Krinkle: thanks [02:15:48] You're welcome [02:16:49] AbstractBeliefs: req = urllib2.Request(url) \n req.headers['User-Agent'] = 'my-string' \n respons = urllib2.urlopen(req) [02:16:52] !log LocalisationUpdate completed (1.19) at Thu Mar 1 02:16:52 UTC 2012 [02:16:55] is one way to do it. [02:16:55] Logged the message, Master [02:17:47] maplebed: great, ill give that a bash. the docs say it has a value of "Python-urllib/2.6", but ill try with something more sane. if not, ill end up working sockets [02:17:51] thanks again, guys [02:23:09] Krinkle: does ext.CategoryTree.js look ok? [02:23:19] * Aaron|away looks at rsargs: [$link.data( 'ctTitle' ), $parentTag.data( 'ctOptions' ), 'json'] [02:23:37] Aaron|away: In general or a revision in particular ? [02:23:48] in general, just in HEAD [02:23:55] $parentTag.data( 'ctOptions' ) must be undefined most likely [02:24:03] at least sometimes [02:24:07] ext.categoryTree.js [02:24:15] yes [02:25:08] Aaron|away: I'm not sure [02:25:12] it's set a few lines above it [02:25:16] although conditionalluy [02:25:25] .data( 'ctOptions', mw.config.get( 'wgCategoryTreePageCategoryOptions' ) ); [02:26:38] when ack 'ing the ./extensions/CategoryTree directory I see only 2 occurrences of that [02:26:51] 1) the data() get when making the ajax request [02:26:58] 2) the data() set on the new element [02:27:07] but that set only does it if the element doesn't exist on the page already [02:27:24] TimStarling: I think Aaron|away is hinting he's about to go away, and I probably will take advantage of a window of quiet to get home too [02:27:34] 'the element' = $parentTag = $link.parents( '.CategoryTreeTag' ) [02:27:44] ok, bye [02:27:51] I isolated that zhwiki issue btw [02:28:00] TimStarling: excellent! [02:28:04] Aaron|away: So if that element does not exist on the page, ctOptions will be wgCategoryTreePageCategoryOptions [02:28:07] * Aaron|away installs CT [02:28:11] Aaron|away: and if it does, it's undefined indeed [02:29:31] TimStarling: I know I probably don't have to tell you this, but assuming you get a fix in, don't hesitate to redeploy to zh* [02:29:52] ok [02:30:01] Krinkle: can you have a look at it? you approved the revision [02:30:45] Krinkle: can you commit a CT fix then? :) [02:31:29] I approved it, yes. [02:31:35] Aaron|away: I don't see a bug report [02:31:39] What's the problem ? [02:33:12] CategoryTree::decodeOpti ons: bad json `undefined` in /usr/local/apache/common-local/php-1.19/extensions/ CategoryTree/CategoryTreeFunctions.php on line 187 [02:33:22] * Aaron|away kicks formatting [02:33:36] I haven't look a while into CT, I'm happy to digg it up but is there a bug in the front-end, or do you want me to fix is so that that logged error goes away [02:33:53] !log LocalisationUpdate completed (1.18) at Thu Mar 1 02:33:53 UTC 2012 [02:33:56] Logged the message, Master [02:34:03] huh, just saw this: [02:34:04] CategoryTree::decodeOptions: bad json `{"mode":20,"hideprefix":20,"showcount":false,"namespaces":false}` in /usr/local/apache/common-local/php-1.19/extensions/CategoryTree/CategoryTreeFunctions.php on line 187 [02:34:34] Krinkle: the JS should not give 'undefined' [02:35:46] * Aaron|away goes off [02:36:11] that js code doesn't looks good. it makes an element and assigns .data() to it, only to retrieve it a few lines later [02:36:19] the element is never inserted into the dom [02:36:28] fixing [02:40:16] Aha, ack didn't pick it up because JS data is named ctOptions and php outputs data-ct-options [02:40:21] not as bad as I thought [02:41:55] just took down centralnotice for everything except zh*, which will expire at 4:00 UTC [02:42:17] * robla heads out for reals this time [02:42:19] see ya! [02:42:27] see ya [02:45:13] none of my monobook modifications (js or css) are working [02:45:18] what the heck [02:47:22] Are you actually using the monobook skin too? [02:47:44] yes [02:47:56] Hmm, this /might/ be caused by something I wrote [02:47:56] I have Monobook in use, but none of my custon JS or CSS items seem to be working [02:48:04] *custom [02:48:19] might have gotten censored... ;) [02:48:21] Wait, your vector.js / css also don't work? [02:48:43] > We are now up to four endemic HTML validation errors per page. [02:48:48] I wish everyone wrote like that. [02:49:13] RoanKattouw: I think he means the code in his subpages. [02:49:24] Ryulong: Does your browser have an error console? Have you looked in it? [02:49:40] Joan: Is that our favorite reporter? [02:49:42] It'd be unusal for both CSS and JS to be failing. [02:49:49] well m aybe it's jsut js [02:49:56] RoanKattouw: Heh, I wish. It's from https://en.wikipedia.org/wiki/Wikipedia:VPT#HTML_validation [02:49:56] Joan: also noticed the validation errors... really shouldn't happen. I checked because of recent problems with parsing a page with mozilla's DOM parser. Yes, other sites are much worse :D [02:50:20] none of this stuff is showing up for me http://en.wikipedia.org/wiki/User:Ryulong/monobook.js [02:50:44] Let me copy that to my user page [02:51:27] Saibo: I told the person to file a bug. :-) [02:51:39] that's a good idea.. [02:51:48] however I think I've blocked the difference in page size numbers when I view my contribs in my css [02:51:57] akeytt is not defined [02:52:25] https://en.wikipedia.org/w/index.php?title=User:Voice_of_All/Addtabs/monobook.js&action=raw&ctype=text/javascript [02:52:32] Saibo: https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)#Edit_section_for_old_revisions and the section below. [02:52:38] I wonder what happened to that function [02:53:05] You might have just found an old problem. [02:53:36] It disappeared [02:53:44] Ryulong: /monobook.css and /monobook.js both aren't working for you? [02:54:00] It's working for me [02:54:02] As in the code is loaded [02:54:07] hm [02:54:10] But there's a JS error in one of the things he's importing [02:54:21] what do I have to fix [02:54:22] * RoanKattouw takes the liberty to edit that page to fix the problem [02:54:28] oh [02:54:32] It's another user's subpage, you probably can't edit ti [02:57:00] Whoa, wtf [02:57:04] Now I get 'ta is not defined' [02:57:21] But that's right, it's being used but it's actually not defined anywhere [02:57:29] Was this a global exposed by wikibits or something? [02:57:32] * RoanKattouw digs up 1.16 version of wikibits [02:58:02] Aha [02:58:06] 376 /* Dummy for deprecated function */ [02:58:08] 377 window.ta = []; [02:58:10] 378 function akeytt( doId ) { [02:58:11] 379 } [02:58:13] And that's in 1.16 [02:58:39] OK [02:58:52] So the reason this guy's code breaks is because he's using an interface that was deprecated in 1.16, possibly earlier [02:59:10] That was before we even started /writing/ ResourceLoader [02:59:24] * Krinkle did not read in here [02:59:28] Apparently we kept dummies around and someone decided to remove them after 2 versions of deprecations [02:59:34] Krinkle: akeytt and ta in wikibits.js [02:59:39] mw:RL/MGU mentions this RoanKattouw [02:59:43] Oh? [02:59:47] remove and link to it in the edit summary [02:59:53] Well they seem to have been deprecated pre-1.16 [02:59:56] redundant, is now in core and i18n'ied [03:00:22] Aha: [03:00:24] https://www.mediawiki.org/w/index.php?title=RL/MGU#MediaWiki_1.16_and_before [03:00:26] "Scripts like ta['pt-userpage'] = new Array('.','My user page'); are ignored. It was deprecated in 2009. The function akeytt() no longer exists. " [03:00:29] https://www.mediawiki.org/w/index.php?title=RL/MGU#ta.5B.5D_Tooltip_and_Accesskeys [03:00:59] RoanKattouw: some wikis created ta manually though [03:01:13] so the removal of ta = {} isn't always breaking [03:01:21] e.g. var ta = new Object(); is around a lot too [03:01:25] Well it's not working on this one [03:01:38] I'll just make it conditional on the existence of ta then [03:01:43] Well, or not [03:01:44] I mean [03:01:52] Does a locally defined ta actually /work/ ? [03:02:32] RoanKattouw: Who's tracking 1.19 deployment bugs? [03:02:37] Is anyone watching the village pump? [03:02:40] RoanKattouw: No, nothing works with ta in anyway [03:02:41] Where is hexmode. [03:02:49] here [03:02:54] hexmode: Are you watching the village pump? [03:03:03] It's 10pm on the East Coast [03:03:06] Robla just went home [03:03:13] https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)#Watchlist_time_.28in_the_box_area.29_shows_UTC [03:03:13] RoanKattouw: long long ago (in a galaxy far away), there was addOnloadHook(akeytt), (function akeytt(){} was in wikibits), which took window.ta and handled it [03:03:17] And I do believe they are watching the village pups [03:03:21] adding title="" attributes on the IDs [03:03:27] Joan: I was poking around, thanks for helping! [03:03:28] But those are in core now and outputted by PHP [03:03:40] there is i18n messages if wikis want to override this message [03:04:05] but most likely their messages are copied from elsewhere, similar to what once was merged into core [03:04:34] Joan: I don't think there were major issues other than the languageconverter thing that caused Chinese to be rolled back [03:05:07] The HTML validation stuff is just bogus errors [03:05:17] They're stupid brain-dead rules in HTML4 that were removed in HTML5 [03:05:22] RoanKattouw: The watchlist thing seems like a bug. [03:05:31] The validation errors should at least be logged and looked at. [03:05:36] RoanKattouw: example https://simple.wiktionary.org/w/index.php?title=MediaWiki:Monobook.js&diff=next&oldid=4974 [03:05:38] Such as empty
    s and a class attribute on the tag [03:05:46] We know about at least three of them already [03:05:54] dir=auto is probably also an HTML5 thing [03:06:11] Joan: Basically we will follow standards unless they're clearly brain-dead and just get in the way. These four errors are all like that [03:06:20] Joan, I'm pretty sure that non-techie users aren't going to be particularly interested in creating a Bugzilla account to report all these issues [03:06:32] RoanKattouw: Well, then someone should say so. I thought two of them were a lack of
, though. [03:06:43] No, it's
    without a
  • [03:06:43] > end tag for "ul" which is not finished​; two instances [03:06:58] It means it encountered an end tag for 'ul', but the list isn't finished [03:07:01] Well, that's an awfully awkward way of saying that. [03:07:04] Yeah [03:07:09] But that's most likely what it is [03:07:11] Because it has no list items? [03:07:17] Yeah [03:07:25] Empty lists are legal in HTML5 [03:07:27] A
      without list items is kinda brain-dead, isn't it? [03:07:35] A lot of things are legal, but that doesn't mean you should do them. [03:07:37] (Of course we don't actually output HTML5 yet, because of other JS-related problems) [03:07:41] Empty uls make sense [03:07:45] Do they? [03:07:50] If you want PHP to output the list and JS to conditionally fill it [03:07:58] RoanKattouw: btw that whole /Addtabs script section of his looks redundant with addPortletLink, which was added to core also several years ago [03:08:08] And JS can't add the
        ? [03:08:11] Or it's just easier? [03:08:19] anyway, I see you fixed it already [03:08:26] I mean, you want to insert
      • , surely you could also insult
          . They're a package deal. [03:08:27] It would have to know where to add it, it's much easier this way [03:08:42] Err, not insult. Insert. :-) [03:09:23] I think there's already a bug for this, then. [03:09:45] If it's just the standard validation complaint until en.wiki uses HTML5. That's being tracked somewhere. [03:10:16] It's a violation of XHTML/HTML4, but HTML5 allows it and no browser that I know of ever had a problem with it [03:10:21] only validators [03:10:22] lol, I only realized just now that User:Voice of All == Aaron Schulz [03:10:28] Heh. [03:10:37] Yeah, it's been filed in places [03:10:44] I WONTFIX those when they pop up [03:10:54] Do you know of the bug? I can't find it in Google.. [03:11:01] There are several [03:11:04] RoanKattouw: Hm.. I just looked at that history of that monobook.js, I was like… "Why did Aaron spend 2 hours on this script a few years ago making dozens of edits" [03:11:07] It is him? [03:11:08] I'm not sure if it's ever been reported on its own [03:11:16] Yes, VoA == Aaron [03:11:20] He had a rename a few years ago [03:11:23] https://en.wikipedia.org/w/index.php?title=User:Voice_of_All/Addtabs/monobook.js&action=history [03:11:42] His edits were reattributed and all of his .js subpages were renamed [03:11:46] ah, his rename also moved that script [03:11:51] Then Ryulong got him to rename the script back [03:11:54] but moved back since JS doesn't understand #REDIRECT :) [03:11:55] because people were using it [03:11:56] indeed [03:12:20] Anyway it's kind of weird that the ancient script causing problems was written by I guy that just left here a few hours ago [03:12:21] That'd be fixed if redirects 301'd, right? [03:12:37] Well [03:12:44] action=raw on a redirect page, what do you expect it to do? [03:12:51] I think currently it just spits out the #REDIRECT [03:12:59] yes, as expected [03:13:07] it is requesting the raw contents [03:13:14] so it should do that [03:13:16] Right. But in general, a browser would follow a redirect with