[00:16:32] Reedy: is there anything about REL1_19 particularly risky? [00:17:23] I don't think so [00:17:30] In comparison to 1.18, it's a relatively small release [00:17:42] (not that that actually means anything) [00:18:09] Ouch [00:18:21] The release notes file is half the size (in number of lines) [00:20:01] Reedy: I installed trunk+an extension+a gadget today and poked around. Tomorrow's busy but I can poke at 1.19 some as well. [00:21:23] You'll probably need 4 or 5 dev installs at minimum [00:21:40] trunk, wmf, current release, last release [00:22:23] Krinkle: hey, what do you know about addEvent? MediaWiki:Edittools.js on ar.wikipedia.org is trying to use it, and it's failing [00:22:47] That's probably an ancient thing that should be replaced with jQuery event binding [00:22:55] yeah, that's the easy answer [00:23:17] ok, just isolating this [00:23:30] Could be that they have a local function defining it [00:23:36] but it's no longer a global [00:23:45] or maybe it was in core once and got removed [00:23:55] I don't remember removing such function from wikibits though [00:57:03] Alphabet soup. [00:57:16] Do you all talk like that in-person? [00:58:15] What's alphabet soup? [00:58:45] https://en.wikipedia.org/wiki/Alphabet_soup_(linguistics) [00:58:51] I was skimming scrollback. [00:59:00] I appreciated some of your clarifications. [00:59:04] But it was still a bit of headache to read. [00:59:24] <^demon> Joan: You mean you don't use WP:SHORTCUTS in everyday speech? [00:59:42] VE, EE, ACW, NPT (now PT?), etc. [01:00:00] ^demon: Only WP:DICK and WP:OKAY. :-) [01:00:28] <^demon> I use WP:ILIKEIT, but not in the traditional sense :) [01:00:44] <^demon> Oh, and WP:NOT. That's very very important or I'd never get anything done. [01:00:45] I use {{like}} for that. [01:01:14] Joan is WP:NOT#NICE. [01:19:59] brion: just to make sure that I'm thinking right: How close is MathJax to deployment? [01:20:31] hexmode: in theory we could probably merge & deploy it as an experimental option any time now. i'd want a little more testing first though :) [01:21:05] k... [01:21:21] and then we've got the whole next product cycle to decide to make it default / improve it etc [01:23:47] how can i let users of my extension implement their own custom css? [06:58:02] hashar: Hi [06:58:39] hashar: question, testswarm still only has 1 tag in github (v0.1) how did you get Firefox 10 on integration.wm.o/testswarm ? that was added in v0.2pre and not released yet [07:27:47] Krinkle: :) [07:27:52] Krinkle: sorry was not really there earlier [07:28:08] Krinkle: I guess I have manually added Fifefox 10 in the database [07:28:31] ok [07:28:40] http://testswarm.wmflabs.org/testswarm/ [07:28:47] I just did apt-get install testswarm a minute ago [07:28:49] seems it's failing [07:28:52] the .ini file isn't read [07:29:02] do you remember having changed anything after apt-get? [07:29:08] or is the package good as it is [07:29:43] I suspect the .ini file might not be readable thus the override here: https://github.com/jquery/testswarm/blob/master/index.php#L6 isn't happening [07:30:12] I tried var_dumping it and indeed $swarmConfig is not being changed from the defaults [07:31:19] during 'apt-get install testswarm' there was 1 warning in the output: chown: invalid user: `testswarm:www-data' https://labsconsole.wikimedia.org/wiki/Nova_Resource:I-0000017b#Config, is that related? [07:31:40] It's probably very simple but I'm not sure what to do [07:35:07] hashar: [07:35:56] sorry [07:36:44] that chown: invalid user: `testswarm:www-data' is bad [07:36:53] yeah [07:36:59] the package should create the testswarm user [07:37:27] I didn't use puppet for most of the stuff though. As you can see on the nova resource page #Config I tried to do most stuff that the puppet manifest does. [07:37:33] oh the package should do that? [07:38:09] I am looking at what you did [07:38:54] * hashar looks at puppet conf [07:39:11] hashar: https://toolserver.org/~krinkle/tmp/psty/_out3/ [07:41:14] ok that was bug 33276 [07:41:43] fixed in testswarm-0.1.0-5 :-/ [07:41:51] we need to have the package updated on apt.wikimedia.org [07:42:26] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/107118 [07:49:22] didn't you do this already? [07:49:27] oh, it's from 2011-12 [07:49:37] so… you did this by hand on the production server? [07:50:40] I think puppet did it [07:50:58] at the start I had lot of configuration being done by puppet [07:51:17] then moved some bits to the testswarm package so we can eventually release it [07:51:37] Krinkle: the https://integration.mediawiki.org/testswarm/?state=wipe has been fixed this night :-) [07:51:46] pinged CT about it and someone fixed the innodb issu [07:51:48] issue [07:51:49] the puppet does it before or after installation of the package? [07:51:57] it depends [07:51:59] because I already ran apt-get and now it doesn't work [07:52:12] I have to check the puppet dependency, I think by default it is kind of a random order [07:52:16] so can I just create the user now and chown the file by hand? [07:52:30] yeah that will fix the issue you are facing [07:53:07] ok [07:53:13] can you help me with the exact commands? [07:53:28] sure sure [07:53:39] I can understand some of the debian language, but not enough [07:54:20] just before we do that, is the ?state=wipe testswarm URL supposed to output anything else than "done" ? [07:54:25] no [07:54:29] it's fine now [07:54:32] great :) [07:54:45] hashar: Hm.. I see you added the fetcher files to the debian package [07:55:16] I find that a little strange, what do you think? [07:55:52] hu [07:55:55] oh nevermind [07:56:01] I was looking at tag/ci [07:56:24] I thought I was looking at /debs/testswarm but this is tools/testswarm/debian [07:56:26] sorry I had two other conversations to handle. Now finished and I am fully available. [07:57:45] the fetcher is deployed by puppet [07:58:17] the code is in operation/puppet git repository in files/testswarm/ directory [07:58:47] yeah, its fine [07:59:08] It looked like you added files to the debian package for it, but I saw it wrong [07:59:17] so for the chmod issue [08:00:46] so you need two fix: [08:00:51] 1) create the testswarm user [08:00:54] 2) do the chmod :-)) [08:00:58] yep [08:01:22] testswarm user and group creation are in https://www.mediawiki.org/wiki/Special:Code/MediaWiki/107118 [08:02:02] what matter is the postin script : https://svn.wikimedia.org/viewvc/mediawiki/trunk/debs/testswarm/debian/testswarm.postinst?&pathrev=107118&r1=107117&r2=107118 [08:02:13] it is run by debian package manager after file copying [08:02:22] much like MediaWiki hook system [08:02:23] Hm.. I thought that was special debian code, but it's just plain sh ? [08:02:28] yeah [08:02:45] but I can't paste it in the terminal right ? since it does " if " [08:02:48] and debian offers some function to ease tasks. They are known as debian helper [08:03:13] the "getent group testswarm" is used to verify if a group exist [08:03:27] I know sh/bash a little bit, I just didn't realize this was it [08:03:38] that might be confusing [08:03:39] I'll create a file for in my home and then sh it? [08:04:02] you can just copy paste the diff from viewvc and then run that in shell [08:04:04] that should work [08:04:14] or just copy paste the relevant part [08:04:17] addgroup --system testswarm [08:04:23] right [08:04:33] adduser --system --ingroup testswarm --home /var/lib/testswarm testswarm [08:04:33] usermod -d "/var/lib/testswarm" testswarm [08:04:33] usermod -g "testswarm" testswarm [08:04:34] usermod -G "www-data" testswarm [08:05:40] Warning: The home dir /var/lib/testswarm you specified already exists. [08:05:41] Adding system user `testswarm' (UID 115) ... [08:05:41] Adding new user `testswarm' (UID 115) with group `testswarm' ... [08:05:41] The home directory `/var/lib/testswarm' already exists. Not copying from `/etc/skel'. [08:05:41] adduser: Warning: The home directory `/var/lib/testswarm' does not belong to the user you are currently creating. [08:05:42] usermod: no changes [08:05:42] usermod: no changes [08:07:14] doh [08:07:37] so all is OK ? [08:08:11] /var/lib/testswarm is the user home directory [08:08:14] seems to belong to someone else [08:08:32] I created it [08:08:33] https://labsconsole.wikimedia.org/wiki/Nova_Resource:I-0000017b#Config [08:08:46] cd /var/lib; mkdir testswarm; chmod 0755 testswarm; chgrp www-data testswarm [08:09:19] I guess chown is still me or root [08:09:47] Weird do, I didn't get warnings about missing users back then [08:09:54] verify who it belong to though [08:09:59] stat /var/lib/testswarm [08:10:14] drwxr-xr-x 4 root www-data 4096 Mar 5 19:55 testswarm/ [08:10:19] ( ll ) [08:10:37] Access: (0755/drwxr-xr-x) Uid: ( 0/ root) Gid: ( 33/www-data) [08:10:53] so just: chown -R testswarm:www-data /var/lib/testswarm [08:11:00] ( -R for recursive ) [08:11:15] ok [08:11:28] just curious though, why didn't the other script do that? [08:11:29] then I will award you the [ I hack using shell ] badge :-)) [08:11:48] the postinst script does it [08:11:51] when creating the user [08:11:58] chown: changing ownership of `/var/lib/testswarm/scripts/.git/refs': Operation not permitted [08:12:00] hehe [08:12:16] it probably belong to root :-D [08:12:30] I changed that to be a git clone of the new gerrit repo instead of puppet://files/testswarm [08:12:41] anyway rest looks good [08:12:56] http://testswarm.wmflabs.org/testswarm/ is still off though [08:13:05] chmod on /etc/testswarm as well? [08:13:10] or chown? [08:13:15] all files or just config.ini [08:13:23] (there's 4 files) [08:13:47] that part is a bit messy [08:13:53] hm.. none of the chown's succeeded [08:14:02] chown: changing ownership of `/var/lib/testswarm': Operation not permitted [08:14:05] I didn't scroll down yet [08:14:12] were you root? [08:14:27] I was but not when I pasted that [08:14:34] re-doing [08:14:57] * hashar awards the [ only root change ownership ] badge to Krinkle [08:15:07] yay :) [08:15:08] drwxr-xr-x 4 testswarm www-data 4096 Mar 5 19:55 testswarm/ [08:15:24] ls /etc/testswarm/ apache.conf config.Debian.ini config.ini fetcher.ini [08:15:50] all root/root [08:16:09] the site is installed in /usr/share/testswarm/site/ [08:16:12] yes [08:16:17] the config file is loaded from /usr/share/testswarm/site/index.php [08:16:21] yep [08:16:21] a [08:16:22] nd [08:16:22] de [08:16:22] b [08:16:27] and debian patched index.php [08:16:34] exactly [08:16:47] so chown on /etc/testswarm/config.ini, but also on fetcher.ini ? [08:16:52] not sure from where that one is used [08:16:55] so the file /etc/testswarm/config.ini is read by Apache [08:17:05] on gallium it is : -rw-r----- [08:17:10] testswarm www-data [08:17:19] this way only testswarm is allowed to alter the configuration [08:17:24] but apache is allowed to read it [08:17:34] so that's 640 ? [08:18:01] obviously :-] [08:18:13] it already is 640 on labs [08:18:26] make sure it belong to testswarm:www-data [08:18:29] k [08:20:07] congratulations Krinkle :-))))))) [08:20:19] brb, charger [08:20:23] * hashar awards the [ I fix mad stuff ] to Krinkle [08:21:23] thx [08:21:46] well I should have written a bit more documentation about it [08:22:33] * Krinkle updated https://labsconsole.wikimedia.org/wiki/Nova_Resource:I-0000017b#Config [08:22:54] great [08:23:07] I am still denied from connecting to the labs instance :/ [08:23:57] hashar: If I create a new tag on github (v0.2) how to update? [08:24:11] I want the new user agents in there on labs "the right way: [08:24:11] we [08:24:57] debian package management system comes with a way to detect upstream changes [08:25:02] I never manage to make it work though [08:25:22] supposedly when you tag a new version, some magic command will make the package to auto update :-] [08:25:50] because testswarm only has 2 main sql files, no patch [08:25:54] so how does it figure out the database? [08:26:20] so far the testswarm instructions are: if you want to update to new database, just wipe them and re-populate from scratch [08:26:38] that is not supported by the package [08:26:39] losing existing data, but nobody cares [08:26:52] so debian just assumes database never changes? [08:27:01] indeed [08:27:06] but that is not debian fault [08:27:10] ok [08:27:38] on each upstream release we will have to figure out what changed [08:27:44] then add some SQL script to alter the db [08:28:07] those scrips can be given to dbcommon (a debian helper to manage database) so it can runs the alter [08:28:50] ok, so in the debian package you created would be an sql file somewhere tagged like "run this sql if previous version is < 0.2" ? [08:29:22] exactly [08:29:24] doc: http://people.debian.org/~seanius/policy/dbconfig-common-using.html/ch-develguide.html#s-updates [08:29:45] you just put the SQL script in somepath/testswarm/upgrade/mysql/0.2 [08:30:26] k [08:30:41] I'll create a bug for that [08:30:53] just gonna use v0.1 for testing [08:30:57] not important :) [08:31:35] I just insert the user agents manually from time to time [08:32:46] https://bugzilla.wikimedia.org/show_bug.cgi?id=35027 [08:32:58] sure, but there've been a few other minor changes [08:33:04] nothing important for swarm-specialpage [08:33:24] but we will need those new changes in order to start browserstack VMs for needed browsers [08:33:32] (there's a new getneeded api) [08:33:42] but pretty much untested y et [08:33:46] not in v0.2 [08:33:50] https://bugzilla.wikimedia.org/show_bug.cgi?id=35027 [08:33:53] oops, twice [08:34:24] hashar: do you know if it's hard to monitor testswarm db in some way? perhaps figure out the slowest queries? [08:34:25] https://github.com/jquery/testswarm/blob/master/config/testswarm.sql#L127 [08:34:34] probably insufficient indices [08:34:37] we will have to log them somewhere [08:35:22] there is a mysql command to log them [08:35:39] wmf uses that? [08:35:49] I have no idea [08:35:53] probably not [08:36:08] I guess domas just launch gdb :-) [08:36:33] here are the relevant parts: [08:36:34] #log_slow_queries = /var/log/mysql/mysql-slow.log [08:36:34] #long_query_time = 2 [08:36:34] #log-queries-not-using-indexes [08:36:58] asher would be able to told us [08:39:33] 2 seconds I suppose ? [08:39:39] yup [08:39:46] that logs any queries that takes long than 2 seconds [08:39:49] wow that looks magical, those 3 lines doing it all [08:39:57] the last setting additionally logs anything not using index [08:40:11] that is also a good way to fill your disk [08:41:35] hehe [08:41:40] yeah, we better not keep it in there [08:41:52] just for 2 minutes will give a good amount of data to start with [08:42:38] basically the only queries that matter are the ones triggered by getjob and submitresults [08:42:52] any other actions are rarely viewed [08:43:22] I'm even considering to implement a mode in testswarm where all html serving pages are removed (except /run/) [08:44:07] although right now it's more of a conceptual goal than a plan, there's no point to remove those for those that want to look at it [08:44:46] is there anyone using testswarm beside mozilla and us ? [08:44:58] Is mozilla using it? [08:45:18] (confusingly, it's no longer a Mozilla Labs project, the logo needs updating :D ) [08:45:26] it's a jQuery foundation project now [08:45:30] but that aside [08:45:50] I don't think any other major players are using it [08:46:01] other than us and all regions of jQuery [08:46:41] q [08:47:24] https://www.google.com/search?q=%22Welcome%20to%20the%20TestSwarm%20Alpha!%22 [08:47:30] guess who's on top :P [08:47:32] it's not fair [08:47:37] integration.mediawiki.org/testswarm/ [08:59:17] hashar: thanks :) Gonna be back later [09:49:08] how can i let users of my extension implement their own custom css? [10:22:57] how can i add tag arguments as javascript parameters using ResourceLoaderGetConfigVars? [10:37:48] Nikerabbit: :) [10:37:56] I think scap clear the cache too [10:39:09] and I am afraid of scap nowadays [10:40:07] hashar: did you change channel? [10:40:11] ah I see [10:40:32] as far as I know scap is the only thing rebuilding also l10n cache [10:40:42] I've asked a way to do that manually [10:40:44] but meh [10:41:57] we could extract that part out of the scap script [10:42:22] it is in operations/puppet ./files/misc/scripts/scap [10:44:14] March 6 20:32 logmsgbot: reedy synchronizing Wikimedia installation... : Rebuild message cache stuffs for r113129 [10:44:22] Nikerabbit: looks like reedy did it yesterday [10:46:00] so there might be an other issue :-/ [10:46:56] hmph [10:47:03] check if the message exists locally? [10:47:17] if not it might be a bug in l10ncache and that is mostly roan's area [10:48:23] no message http://sa.wikipedia.org/wiki/MediaWiki:pagetitle?uselang=en [10:49:40] $ mwscript eval.php sawiki [10:49:46] > var_dump( wfMessage( 'pagetitle' )->text() ); [10:49:46] string(0) "" [10:49:48] \o/ [10:55:36] argh.. Im scratching my head like crazy. Anyone know how i can output javascript variables from within the tag hook function? [10:55:55] Nikerabbit: I have just reopened bug 34938 [10:56:18] i've tried $wgOut->addJsConfigVars() with no luck.. [10:56:26] I have no idea sorry :/ [10:56:36] maybe ask on wikitech-l / mediawiki-l lists ? [10:58:20] hashar: poke roan? [10:58:59] have you been able to reproduce the issue ? [10:59:26] because the issue only appear on certain pages IIRC [10:59:41] and your fix in https://www.mediawiki.org/wiki/Special:Code/MediaWiki/113039 would have applied on any page [11:03:27] hashar: it's cached of course [11:03:43] sure [11:03:54] but I can't find out how to check the cache [11:06:18] Nikerabbit: how can we get the raw content of a message key ? [11:06:36] hashar: what do you mean? [11:06:48] wfMessage( 'foo' )->plain() ? [11:06:53] will try [11:07:17] yeah!!!!!!!! [11:10:51] hashar: yeah what? [11:11:05] ->plain() was what I was looking for [11:12:58] Nikerabbit: maybe the 'pagetitle' key is still set to empty in translatewiki.net ? [11:13:14] thus l10nupdate would still apply the empty string [11:14:38] hashar: l10nupdate takes it directly from svn [11:15:05] hashar: and I think I removed the key from trunk... has it been readded? [11:16:00] repeating what i said in #mediawiki: i suspect that it may have something to do with the fact that this message existed there in the past and was deleted. [11:16:10] i can't see its history, because i'm not an admin. [11:16:26] it happened to me a few times that deleted local messages were still seen, although i couldn't reproduce this consistently. [11:16:39] might be a bug in l10nupdate so [11:16:54] the script ran this night and it seems page title is still set to "" in the cache file [11:19:21] yeah 'pagetitle' => '', [11:19:23] great [11:25:17] hashar: in trunk? [11:25:34] that is from the output of l10nupdate [11:25:41] oh it has output too? [11:25:43] which ran at 2am [11:25:55] well not really, but it generates a serialized cache file [11:26:10] so you can download the file, unserialize the content and see the cache :-) [11:26:11] I would just rm -rf everything and regenerate it, but that might unstabilize the site ;) [11:26:53] Updating LocalisationCache for 1.19... [11:26:56] waiting now [11:27:12] preparing lunch [11:27:21] spam some beep if there is any trouble. I am not far [12:28:08] ahh we have a new drama thread on wikitech-l ! [12:28:11] about git and gerrit [12:28:13] * hashar reads [12:34:42] * hashar wonders if he should start trolling [12:52:35] * hashar thinks is going to fly around europe to train people :-]]] [12:55:54] Nikerabbit: GiGeGa for Git/Gerrit/GatedTrunk is a great find! [12:56:40] hashar: siebrand wanted GiGeGat [12:57:37] I am sure that git / gerrit thread is going to be a very long one [13:13:25] hashar, the most hardcore trolling is to switch to Gerrit ;] [13:14:12] it's kind of collective insamity [13:14:34] I do like Gerrit [13:14:53] epic, epic [13:16:48] You've gotta appreciate this community - the people on that thread are still not calling each other idiots. (Unless i missed it.) [13:17:24] I am seriously considering a road trip across europe to evangelize people about git/Gerrit ;) [13:24:01] hashar, you're concentrating on "oh, Gerrit allows us to have a gated trunk and merge with one click", while Nikerabbit concentrates on "its UI sucks, and it actually makes some stuff slower". just two ends of the same stick [13:24:23] I actually don't like either of those "features" [13:24:47] git is a great tool that the use of gerrit totally undermines, afaict [13:24:55] yep [13:25:14] naaaa [13:25:15] we might as well move to github for a while [13:25:23] NO FOR GOD SAKE!!!!!!!!!! [13:25:24] so we can be hacked? :-P :-D [13:25:25] pleasE? :) [13:25:40] seriously [13:25:58] we could have used github. that would be a great way to get patches from ton of people [13:28:02] MaxSem: yeah nike moved off topic :-] [13:28:11] MaxSem: I guess that disqualify him as a good troll [13:28:59] hashar keeps slowly gaining troll points meanwhile XD [13:30:37] ah bite me gdb + fedora [13:30:57] it's almost enough to make me switch to gentoo. [13:30:59] almost. [13:32:02] smewhere...there is a bad interaction with python, open2.poen4, php readline and gnu readline and pipes. [13:32:45] guess I'll get dessert, this download of alll the symbols is gonna take awhile [13:38:27] MaxSem: I think i will summon Jidanni to assists us in the git migration [13:39:26] nope, his lynx doesn't support tabs so he's screwed too [13:39:32] ohhh [13:39:34] <^demon|zzz> gerrit won't work in a text browser. [13:39:48] good morning Chad [13:39:54] * hashar sends Chad a donut [13:39:58] that's a potential problem [13:40:06] chad or the donut? [13:40:16] no [13:40:19] gerrit [13:40:31] I'll take an extra donut if you have one and they aren't stale [13:40:37] <^demon> You can use the command line interface, but that's totally unintuitive. [13:40:47] ugh [13:40:55] <^demon> (yes, gerrit has a non-interactive shell you can use via the 29418 port) [13:41:28] ok 117 mb of the libc symbols downloaded so far [13:41:29] grumble [13:41:55] https://www.mediawiki.org/wiki/File:Gerrit_under_lynx.png [13:42:14] <^demon> Heh, at least it does that rather than not returning anything. [13:42:15] please note it actually get a cookie from code.google.com !! [13:42:18] +1 for privacy [13:42:22] <^demon> Or returning something that doesn't work. [13:42:36] I like the IFRAME: javascript:'' [13:42:40] what cookie? [13:43:04] when browsing gerrit.wikimedia.org , we are asking to allow a cookie for code.google.com [13:43:11] <^demon> Not for me. [13:43:19] <^demon> I don't see a code.google.com cookie [13:43:30] <^demon> And if so, that's Totally Evil and we need to stop that now. [13:43:40] https://www.mediawiki.org/wiki/File:Gerrit_under_lynx.png <-- -bottom of lynx window [13:43:49] <^demon> Yeah I see it. [13:43:55] <^demon> Chrome's not showing any cookies. [13:44:02] ahahaha [13:44:18] Chrome does not know cookie. Chrome already knows everything it needs to know about you. [13:44:32] * hashar opens a bug report [13:44:34] <^demon> I'm inspecting the cookies, kthnx. [13:44:49] <^demon> I'm installing lynx. [13:45:17] I can't, rats [13:45:27] cause I'm busy installing debugging symbols for a bunch of crap :-( [13:45:51] <^demon> lynx isn't prompting me for a cookie. [13:45:55] <^demon> You sure that was gerrit.wm.o? [13:46:00] yeah because you have Chrome [13:46:10] the loading splash screen has Loading Gerrit Code Review ... [13:46:11] he says lynx doesn't prompt [13:46:12] it silently sniff all your ethernet traffic to auto accept cookies [13:46:19] so if your browser has prefetching enabled [13:46:26] code.google.com would send a cookie. #fyi [13:46:27] the loading splash screen? mmm [13:46:45] au: awesome! thanks :-)))))))) [13:46:53] <^demon> Seriously? That's totally evil. [13:47:42]