[00:03:30] I'll merge https://gerrit.wikimedia.org/r/#/c/6894/ right now if there are no objections [00:04:31] <^demon> TimStarling: Reviewing. [00:05:53] I tested it locally [00:06:12] <^demon> Looks good, merged. [00:06:17] seems ok [00:07:01] I have to now merge it to two deployment branches, right? [00:07:42] and both require going through gerrit rather than a simple push? [00:07:42] <^demon> Cherry pick to 1.20wmf1 and wmf2, if we want it deployed right now. [00:07:57] <^demon> Yes, we disabled direct pushing for the deployment branches. [00:10:51] tricky [00:11:12] what's the easiest way to merge everything except the RELEASE-NOTES change? [00:11:27] <^demon> Reedy? [00:11:32] <^demon> How've you been doing it? [00:13:14] I think I just fixed the conflict [00:14:02] though, you could abandon the changes to that file [00:14:17] Well that's a valid way of "resolving" the conflict [00:14:41] I never bothered merging RELEASE-NOTES to wmf branches with subversion [00:15:24] makes sense [00:17:24] http://stackoverflow.com/questions/1526044/partly-cherry-picking-a-commit-with-git [00:17:34] cherry pick, reset, add, commit [00:17:54] right [00:17:56] thanks [00:20:47] !log aaron synchronized wmf-config/CommonSettings.php 'Switched enwiki to new backend config.' [00:20:50] Logged the message, Master [00:21:53] TimStarling: don't forget to check https://gerrit.wikimedia.org/r/#/c/6783/ when you get the chance [00:22:07] ok [00:31:03] the php- prefixes on our MediaWiki directories annoy me [00:31:20] you know that naming convention predates the name "MediaWiki", otherwise it would have been mw- or something [00:32:00] TimStarling: you can change it if you like, as long as nothing explodes :p [00:32:14] * AaronSchulz stirs the symlink soup [00:32:21] * AaronSchulz adds some salt [00:32:29] it's hard coded in multiversion now [00:32:40] not to mention scap etc. [00:32:48] and the apache configuration files [00:32:54] probably a lot of places [00:32:56] why the apache config? [00:32:57] You know there's a label on that soup pot saying "nitroglycerin 73% vol." right? [00:33:07] * AaronSchulz sighs [00:33:09] oh well [00:34:05] I thought that we would fix the prefixes when we did hetdeploy, it was in some notes I gave priyanka [00:34:13] but that idea was lost when the project transitioned to you [00:34:22] I thought I'd mention it in case another opportunity comes by [00:34:39] * AaronSchulz is a notebook now [00:36:47] !log tstarling synchronized php-1.20wmf1/includes/DefaultSettings.php [00:36:50] Logged the message, Master [00:37:12] !log tstarling synchronized php-1.20wmf1/includes/Exception.php [00:37:15] Logged the message, Master [00:42:10] !log tstarling synchronized php-1.20wmf2/includes/DefaultSettings.php [00:42:13] Logged the message, Master [00:42:32] !log tstarling synchronized php-1.20wmf2/includes/Exception.php [00:42:35] Logged the message, Master [00:43:56] exception.log is full of api.php spam [00:47:06] ah right, reedy broke it [00:47:10] typical [00:47:20] What's up with it? [00:48:10] // Log it [00:48:10] if ( $e instanceof MWException ) { [00:48:10] wfDebugLog( 'exception', $e->getLogMessage() ); [00:48:10] } [00:48:22] see, the instanceof was meant to make it not log UsageExceptions [00:49:37] Ah [00:50:39] I guess you're gonna revert it (no issue with that), would you add a comment to make it explicit that it's done for a reaosn? [00:50:43] it can probably be !($e instanceof UsageException) instead [00:50:51] that would make more sense, right? [00:51:17] That would, yeah [01:10:43] !log tstarling synchronized php-1.20wmf1/includes/api/ApiMain.php [01:11:07] !log tstarling synchronized php-1.20wmf2/includes/api/ApiMain.php [01:12:56] Logged the message, Master [01:12:59] Logged the message, Master [01:16:08] 2012-05-08 01:14:57 srv229 enwiki: [667e119d] /w/wiki.phtml?title=User:%C6var_Arnfj%F6r%F0_Bjarmason/monobook.css&action=raw&ctype=text/css Exception from line 914 of /usr/local/apache/common-local/php-1.20wmf2/includes/WebRequest.php: Invalid file extension found in the path info or query string. [01:17:12] I guess he used the .phtml extension just to be funny [01:17:36] it was .php long before action=raw was introduced [01:18:37] I wonder where that's being called.. [01:23:27] TimStarling: well how long has it been since any ^/w/wiki\. ? (should be /w/index.php , right?) [01:24:35] " using index.php and redirect.php instead of wiki.phtml and redirect.phtml [01:24:36] keeping phtml stubs around for compatibility and CVS history [01:24:36] " [01:24:53] Erik Moeller, Feb 2004 [01:25:28] lol [01:25:43] * RoanKattouw doesn't see Erik at his desk [01:26:28] he used to do a lot of development work back then [01:27:04] he was the one who suggested adding defaults to template arguments [01:27:10] i.e. {{{arg|default}}} [01:27:16] and we all know how that turned out, right? [01:27:37] haha [01:27:43] {{qif}} [01:27:46] yes [01:31:47] and action=raw was introduced in April 2004 by Gabriel [01:32:00] hah [01:32:02] so not much of a gap at all really [01:32:11] but I had the order right [02:00:16] !log LocalisationUpdate failed: git pull of extensions failed [02:00:20] Logged the message, Master [02:01:23] Server problems? [02:01:29] eww [02:01:31] * RoanKattouw looks [02:01:40] RoanKattouw: it did that like 4-5 days in a row now [02:01:51] error: Your local changes to '.gitmodules' would be overwritten by merge. Aborting. [02:01:53] Please, commit your changes or stash them before you can merge. [02:01:53] vssun1: like what? [02:01:54] * RoanKattouw rages [02:02:02] Local changes? On MY cluster? [02:02:25] sites are not loading / take long time to load [02:02:44] you could be a little more specific [02:03:08] getting better now.. I was looking for ml.wiktionary and ml.wikisource [02:03:36] git sucks [02:04:26] Reedy: be careful what you say, the british don't have seem to have the same freedoms of speech as US ;P [02:04:42] say what? [02:05:15] Reedy: "git sucks" [02:05:15] Chad isn't here [02:05:17] I'm safe ;) [02:06:22] vssun1: ya'll need to get a local logo made for mlwikt [02:07:03] otherwise i see no issues. (well if you discount me not having the right fonts) [02:07:36] might be some routing problems [02:07:54] now working fine. [02:08:11] sure, could be lots of things. could also be that i just didn't check until after it was half recovered [02:10:38] Thinking about the logo [02:10:44] will upload it soon [02:14:10] OK, l10nupdate should be fixed now [02:14:33] My workaround for the lack of magic submodules broke because we actually have magic submodules now :) [02:14:44] So I deleted and recreated the extensions checkout, should be all good now [02:17:37] does l10nupdate flush APC? [02:17:42] no [02:18:11] !log Removed and recloned /var/lib/l10nupdate/mediawiki/extensions , it was in a weird state because magic extension submodules work now but my hacky workaround for them not working was still in place [02:18:14] Logged the message, Mr. Obvious [02:18:15] jeremyb: Why would it need to? [02:18:52] idk, i figured it would have cached strings [02:19:11] No, APC doesn't do that [02:19:18] Not the l10n data at least [02:19:22] That's all cached in CDB & memcached [02:19:42] CDB? [02:20:01] .cdb files [02:20:11] It's a flat file key-value format [02:20:50] The l10n for each language is precomputed (because of fallbacks and because of l10nupdate overriding .i18n.php files in the MW source), then put in a .cdb file, then synced to the Apaches [02:21:02] So they just load a .cdb file into memory and they've got the entire l10n for a given language [02:21:55] oh, huh [02:22:02] Additionally, individual messages are cached in memc, that's populated from either .cdb (=disk) or the MediaWiki: namespace (=DB + external storage) [14:15:59] is there a list of wikimedia wikis that disabled local uploads? [14:16:51] http://noc.wikimedia.org/conf/highlight.php?file=InitialiseSettings.php Ctrl+F wgEnableUploads [14:27:35] heh [14:28:18] ToAruShiroiNeko, it's linked from the page I linked on your RfC: https://commons.wikimedia.org/wiki/Commons:Turning_off_local_uploads [14:33:35] anyone with access to http://dumps.wikimedia.org/android/ and http://dumps.wikimedia.org/iOS/ [14:33:38] ? [14:34:47] why? [14:35:02] It loads nicely for me [14:35:18] Hydriz: no, I am pushing out new betas, and want someone to upload files there for me :) [14:35:29] oh I see [14:35:38] ping apergos [14:35:48] * YuviPanda pings apergos [14:35:51] he manages that server (I think) [14:36:13] I do [14:36:31] apergos: can you copy over two files there for me? [14:36:39] where are they now? [14:36:58] apergos: http://dl.dropbox.com/u/8768784/Wikipedia-v3.2-beta2.ipa (into /iOS) and http://dl.dropbox.com/u/8768784/Wikipedia-v1.2-beta2.apk (into /android) [14:37:33] apergos: At the same time update dvd.html to link to mlwiki/mlwikisource ISOs which I told you a few days ago [14:38:00] the dvd link is still on my todo [14:38:06] it take more than an edit in place [14:38:12] lol [14:38:45] ToAruShiroiNeko, I spammed your RfC in a few places [14:38:55] your files are there; please check that everything looks correct [14:39:24] apergos: thanks! checking [14:40:03] apergos: looks good. Thanks :) [14:40:07] sure [14:40:30] oh yes, does dumps.wikimedia.org cap bandwidth at 10MB/s? [14:41:57] for downloaders yes it does [14:42:13] I see [14:42:21] cos Meta was reporting 20MB/s [14:42:29] you can have two connections [14:42:43] zzz [14:42:52] anyways yeah we cap, sadly [14:43:08] which is why mirrors are a good thing [14:44:05] sigh [14:44:14] is your.org rsync working? [14:53:47] Nemo_bis thats ok [14:53:54] spam away :p [14:54:09] I want to avoid a flame war with the entire it.wikipedia if possible [14:54:15] I hope we don't have a flood of Commons users saying "noo we don't want more uploads to check please" [14:54:29] we might [14:54:36] the problem is mitigated to commons [14:54:38] but thats one place [14:55:07] if you promise to do the cleanup on Commons there will be no problems :) [14:55:22] I am not a commons admin so I doubt I can [14:55:47] the problem only lasts for one or two months then people give up [14:56:24] hm [14:58:45] I served as an admin when spanish wikipedia turned off their uploads [14:58:53] commons didnt die [14:59:11] btw how can I check if a value is a number or not with parse functions [15:53:41] if you revdel an edit summary on someone's userpage, that shouldn't trigger a notification email, right? [16:02:57] MBisanz, are you going to oversight Jimbo? :P [16:03:57] huh? [16:04:00] no [16:04:02] haha [16:36:42] [A [17:00:20] Aaron|away: May 8 16:20:30 10.0.11.35 apache2[13592]: PHP Fatal error: require_once() [function.require]: Failed opening required '/home/wikipedia/common/multiversion/MWVersion.php' (include_path='.:/usr/share/php:/usr/local/apache/common/php') in /usr/local/apache/common-local/live-1.5/MWVersion.php on line 12 [17:00:36] Aaron|away: That's very weird cause 10.0.11.35 is not the testwiki host so it shouldn't be trying to access /home [17:00:46] [17:02:01] I pasted that one yesterday [17:16:56] !log reedy synchronized php-1.20wmf1/extensions/MobileFrontend/ 'Pushing out head' [17:17:00] Logged the message, Master [17:18:04] !log reedy synchronized php-1.20wmf2/extensions/MobileFrontend/ 'Pushing out head' [17:18:06] awjr: ^ [17:18:07] Logged the message, Master [17:18:10] Reedy: :O [17:18:21] Reedy: thanks! [17:18:34] Nikerabbit: yours is next [17:18:41] yay [17:19:26] mw35 code looks fine [17:19:59] I'd better fix the fixme on cr [17:20:02] * Reedy glares at AaronSchulz [17:20:23] ah, right [17:23:55] RoanKattouw, wrt https://bugzilla.wikimedia.org/show_bug.cgi?id=28026#c56 , how hard it is to change the current preference of all existing users from unset (NULL?) to false? so that we can then change the default to true [17:24:47] !log reedy synchronizing Wikimedia installation... : Rebuilding localisation cache files before TranslationNotification deploy [17:24:50] Logged the message, Master [17:25:14] Nemo_bis: I don't follow. There are two possible default values (true and false), not three (unset is not an option) [17:25:22] Oooh wait [17:25:25] I see what you're saying [17:25:33] sorry for bad terminology [17:25:36] Set all users' prefs to false explicitly [17:25:38] Right [17:25:39] yep [17:25:40] That could be done [17:25:55] You would need to set the default to true /first/ , then start setting stuff back to false [17:26:02] Otherwise it'll just convert false back to unset [17:26:50] RoanKattouw, and how long does it take to convert all users to false after setting the default to true? [17:26:53] It could be done fairly easily, someone would have to write a (fairly trivial) maintenance script to set all prefs to false [17:26:58] but it would take a bit of time to run [17:27:09] I can't tell you offhand how long it would take [17:27:31] hmm so one should kill the email relay to avoid spamming accounts? :) [17:27:31] But given that there are 20 million users on enwiki, I think it could easily take multiple hours [17:27:36] yeah [17:27:43] Hmm yeah you're right this is a problem [17:27:46] Well -- [17:27:58] We could hack the code so that the preference is offered but not actually functional [17:28:08] Reedy: [17:28:12] ops [17:28:16] Then default it to true, then start setting it to false for each user [17:28:19] Nikerabbit: ? [17:28:19] RoanKattouw: can't you modify database directly? [17:28:30] Then only when that is done, remove the hack such that the pref does things [17:28:33] RoanKattouw, could you comment on the bug explaining how to proceed for best? [17:28:38] Reedy: scap stil running, right? [17:28:41] Yup [17:28:59] Nikerabbit: Yes, but 1) you need to do it in 500-row batches so as to not make replication explode and 2) you need to invalidate the User obj cache in memc for every user you touch [17:29:11] Nemo_bis: Sure [17:29:18] thanks [17:37:37] Reedy: still running? [17:38:02] yup [17:38:15] sync done. [17:38:21] Nikerabbit: ^ [17:38:23] ;D [17:38:41] ~13 minutes [17:38:55] Reedy: yay [17:39:01] Reedy: so now I can go and break the site? [17:39:05] Sure [17:39:08] Where are you enabling it? [17:39:42] Reedy: mworg and meta [17:39:49] and maybe incubator and wikimaniawiki too [17:39:53] Cool [17:40:47] is it technically possible to setup a local wiki that uploads to commons instead of locally? [17:40:57] provided it is a sul account [17:41:15] but doing so using the local form and not commons forms [17:41:32] Yes, it would be [17:42:10] was that a reply to me? [17:42:56] ToAruShiroiNeko, that's a feature under development AFAIK [17:43:27] AaronSchulz: has someone made you aware that InstantCommons is broken? https://translatewiki.net/wiki/File:Yes_check.svg / https://translatewiki.net/wiki/Thread:Support/Template:Rights_using_non-existing_image [17:43:54] <^demon> siebrand: I know we got some fixes in for IC a couple of days ago. [17:44:07] !log nikerabbit synchronized wmf-config/InitialiseSettings.php 'Enable TranslationNotifications on mediawikiwiki' [17:44:10] Logged the message, Master [17:44:55] Nemo_bis how much "under"? [17:44:56] I'd expect that tests would catch these kinds of breakages. How could this get broken in master HEAD without something firing? [17:45:36] ToAruShiroiNeko, currently only planned IIRC [17:45:47] <^demon> siebrand: There aren't any tests for InstantCommons. [17:45:54] ToAruShiroiNeko, just don't mention it anywhere or bikeshedding will kill your RfC ;) [17:46:05] simple links to Commons from the sidebar are perferctly ok [17:46:23] ^demon: okay. Will open an an issue in bugzilla. [17:47:00] <^demon> The problem with instant commons tests is they'd require external requests (which was part of the flaw with the HttpRequest tests). [17:47:17] <^demon> We need to make a new group like @externalhttp or something so we can exclude them by default. [17:47:45] Nemo_bis somepeople find the default forms confusing [17:48:00] ToAruShiroiNeko, who cares about default forms? [17:48:04] !log nikerabbit synchronized wmf-config/InitialiseSettings.php 'Enable TranslationNotifications on meta, incubator and wikimania2012' [17:48:07] Logged the message, Master [17:50:31] thanks for the comment RoanKattouw :D [17:50:42] now I only need to know who could do that... :[ [17:53:53] ^demon: done: 36653/36655. [17:54:03] <^demon> gerrit's down. [17:54:11] <^demon> Investigating why. [17:54:20] <^demon> Networking issue [17:56:20] <^demon> siebrand: Back [18:00:36] RoanKattouw can i bug you with some git questions real quick? [18:00:59] Suer [18:01:29] RoanKattouw: i need to roll back MobileFrontend to a previous revision, and then cherry pick a change on top of it in deployment [18:01:40] but im not entirely sure how to do that [18:01:43] Hmm [18:01:58] Do you currently have a deployment branch for MF? [18:02:00] basically, i want to roll wmf1/wmf2 back to https://gerrit.wikimedia.org/r/#/c/6869/ [18:02:00] deploy from SVN XD [18:02:16] and then apply https://gerrit.wikimedia.org/r/#/c/6915/ [18:02:26] RoanKattouw: yes we currently have a deployment branch [18:02:38] Hmm OK [18:02:58] How many commits are you seeking to revert? [18:03:31] RoanKattouw: 6, i think, including the merges [18:04:27] OK [18:04:39] RoanKattouw: would it be easiest to blow away our wmf/1.20wmf1 branch, start over again by creating the branch from point in the past, then cherry pick the change on top of it? [18:04:45] So really what you should do is just revert those individually in the deployment branch [18:04:55] Because blowing away a branch is so rude that Gerrit prohibits it [18:04:59] lol [18:05:00] ok [18:05:44] 'git revert ab56f7....' will create a revert commit, and you'll need to run 'git commit --amend' after that to make sure it gets a Change-Id [18:06:01] You'll want to revert all commits that need undoing, skipping the merges [18:06:32] <^demon> RoanKattouw: It's only blocked because you can't force push and the syntax is awkward ;-) [18:07:07] ok in that case i'm just reverting three changes [18:07:18] OK cool, that's managable [18:07:24] manageable ? [18:07:31] should i amend after i revert all three? [18:07:37] or amend after each revert? [18:08:10] amend after each revert [18:08:14] Alternatively [18:08:27] No, wait, never mind, that won't work [18:08:28] ignore me :) [18:08:29] oops [18:08:41] Yeah just amend after every revert [18:08:51] I was going to suggest something else and then I realized that that wouldn't work [18:08:53] ok i jumped the gun and reverted them all first [18:08:57] with no amend [18:09:00] Oh OK [18:09:06] uh [18:09:06] So none of the revert summaries have change-IDs? [18:09:08] That can still be fixed [18:09:10] currently no [18:09:19] OK [18:09:29] There's 3 of these right? [18:09:35] correct [18:09:39] Then run git rebase --interactive HEAD~3 [18:09:45] oh yeah [18:09:46] makes sense [18:09:57] That'll throw you into an editor listing all your revert summaries and nothing else (hopefully) [18:10:04] correct [18:10:08] pick 'em all? [18:10:08] For each revert summary, change 'pick' to 'edit', then save and exit [18:10:10] k [18:10:12] Ah sorry [18:10:14] 'reword' [18:10:17] s/pick/reword [18:10:28] ok [18:10:42] and then save each one? [18:10:44] Then that'll allow you to edit the commit summary of each revert in order [18:10:48] And just save those without changes, yes [18:10:55] cool [18:10:55] now [18:10:57] git review? [18:11:01] Then look at 'git log' to check they all have Change-IDs [18:11:11] yes they do [18:11:14] Sweet [18:11:17] Then yeah git review [18:11:28] It'll warn you about submitting multiple changes but that's fine [18:11:35] Because, you know, you are [18:11:42] hehhehe ok cool thanks [18:11:49] then i'll do the same thing on the other branch [18:12:43] Nemo_bis should I mark supports and opposes with template? [18:12:56] oh woops [18:13:03] i should have cd'd out of my extension before git review i think [18:13:20] ToAruShiroiNeko, dunno, only if it advantages you of course :D [18:13:50] adding templates might be considered altering comments, but moving them to sections is acceptable refactoring (if it doesn't alter threaded replies) [18:15:04] awjr: No, you should git review from inside the extension too [18:15:15] o [18:15:25] It should have submitted 3 commits for MF and one for core[wmf/1.20wmfN] [18:15:36] shit ok i think i screwed this up [18:15:41] Nemo_bis I suppose so [18:15:47] I shall leave them as is then [18:15:56] RoanKattouw: it submitted three changes from inside the extension to master; i abandoned those. [18:16:00] Oh [18:16:06] You should un-abandon them [18:16:10] lol ok [18:16:12] Oooh it submitted them to *master* [18:16:13] OK [18:16:15] In *that* case [18:16:20] Keep them abandoned [18:16:36] But resubmit them (from inside the extension dir) with 'git review wmf/1.20wmf2' [18:16:39] ok yeah that's what i thought [18:18:16] aw crap [18:18:19] i think i need to rebase again? [18:20:42] RoanKattouw woot [18:20:58] I thought I was the master [18:21:25] awjr: ? [18:21:41] awjrichards@virtualbot:~/Dev/wikimedia_git/deployment/core/extensions/MobileFrontend$ $ git review wmf/1.20wmf2 [18:21:41] Errors running git rebase -i remotes/gerrit/wmf/1.20wmf2 [18:21:42] fatal: Needed a single revision [18:21:42] Invalid base [18:21:45] RoanKattouw ^ [18:21:46] awjr: Your commits were based off the wmf/1.20wmf2 branch I hope? [18:22:00] i believe so yes [18:22:03] OK [18:22:08] Then rerun it with -R [18:22:20] RoanKattouw if you have the time I'd like to show you a few bugs [18:22:56] ToAruShiroiNeko: Today's my 20% day anyway, so hit me [18:23:28] ok [18:23:42] https://bugzilla.wikimedia.org/show_bug.cgi?id=21338 [18:23:49] https://bugzilla.wikimedia.org/show_bug.cgi?id=22521 [18:23:49] gah RoanKattouw: [18:23:50] awjrichards@virtualbot:~/Dev/wikimedia_git/deployment/core/extensions/MobileFrontend$ $ git review wmf/1.20wmf2 -R [18:23:50] Had trouble running git log --decorate --oneline wmf/1.20wmf2 --not remotes/gerrit/wmf/1.20wmf2 [18:23:50] fatal: ambiguous argument 'wmf/1.20wmf2': unknown revision or path not in the working tree. [18:23:50] Use '--' to separate paths from revisions [18:23:56] https://bugzilla.wikimedia.org/show_bug.cgi?id=32077 [18:24:02] https://bugzilla.wikimedia.org/show_bug.cgi?id=34041 [18:25:04] RoanKattouw i think i really screwed this up i think i should just start over [18:26:30] awjr: wtf... do you even have a local wmf/1.20wmf2 branch? [18:26:40] RoanKattouw: i do, at least for core [18:26:45] And for the ext? [18:26:53] wtf [18:26:55] i thought i did [18:27:01] but i appear to be in a detached state [18:30:21] ok i see, MobileFrontend did not have a wmf/1.20wmf2 branch [18:36:14] RoanKattouw any remarks? :0 [18:36:44] ToAruShiroiNeko: Oh sorry lemme look at those [18:38:26] brion: So what's the state of chunked uploading now, do you know? Wasn't it like enabled experimentally last week? [18:40:51] RoanKattouw, i believe it should work; last i saw erik was requesting it be turned on but i forget if we did yet [18:40:54] lemme check the live config [18:41:09] brion: Where "it should work" means UW supports it as well? [18:41:23] $wgUploadWizardConfig['enableChunked'] = 'opt-in'; in CommonSettings.php [18:41:29] so yeah, UploadWizard should be using it where enabled [18:41:32] brion would ordinary users be able to use it? [18:41:44] ToAruShiroiNeko, if you write an API client to upload in chunks, then yes :) [18:41:50] or else use UPloadWizard [18:41:52]