[05:01:29] bd808: sorry if I asked this before, but have you considered reserving a gerrit project namespace for PHP libraries? [05:01:37] it's a bit awkward for them to all be at the top level [05:02:03] Chad wants the gerrit project namespaces to die [05:02:16] and phab won't have them [05:02:46] but yes having everything in one big pile does look messy [05:02:50] Not much of a point, then. Ok. [05:03:57] we debated it a bit during the librarization project and Chad voted it down pretty strongly [05:04:22] * ori nods. [07:06:40] Users are reporting they're unable to log into Wikidata until they clear their cookies (aude had the same issue in here in scrollback) and I'm not sure what's going on and I need to sleep: https://phabricator.wikimedia.org/T109038. Reverting my config patch might help, but I'm not sure if that would just break it for people who did login?? [07:26:17] legoktm: i'll give it a go, but no promises [07:29:09] i don't think reverting would break it for people who did login [09:01:58] ori: legoktm thanks for looking into it [10:25:57] hi [10:25:59] how to import mediawiki page dump to new namespace...any help appreciated [10:27:04] i have 2 gb of media wiki dump from server 1 and I want to import the dump in to server2 with new namespace [15:18:06] bd808: around? [16:15:20] aude: I am now. was on a conf call [16:18:24] bd808: i would like to profile one of our maintenace scripts on terbium or somewhere [16:18:45] sometimes it is possible with --profiler=text [16:19:05] https://noc.wikimedia.org/conf/highlight.php?file=StartProfiler.php indicates it is possible if hhvm.stats.enable_hot_profiler is set [16:19:21] so, where/how? mw1017? [16:19:38] hmm.... maybe [16:19:53] ori and/or AaronSchulz may be able to help with details on that [16:19:53] but the script doesn't work on mw1017 forme [16:20:29] no maintenance scripts work for me there :/ [16:20:30] mw1017 wouldn't have mwscript right? so you'd have to make a cli that did the stuff it does [16:20:32] terbium is zend [16:20:44] seems it does [16:20:56] ori: i see [16:21:04] so osmium? [16:21:19] in the absence of mwscript, you can preface your script with: require_once '/srv/mediawiki/multiversion/MWVersion.php'; require_once getMediaWiki( 'maintenance/commandLine.inc', 'wikidatawiki' ); [16:21:28] ok [16:21:29] no, osmium has been repurposed [16:21:38] oh [16:22:29] i get something about " Failed to initialize central HHBC repository" on mw1017 [16:22:59] needs to run as www-data [16:23:05] ok [16:23:23] ok, works :) [16:23:49] that's one of the bits mwscript does (sudo -u www-data) [16:24:17] ok [16:24:30] i don't get any profiling though with profiler=text [16:24:49] ah. the mwscript on mw1017 is an old hand copy [16:24:59] it's trying to sudo -u apache [16:25:19] i see [16:25:22] or maybe we haven't changed that for prod [16:25:34] since tin/terbioum are still 12.04 [16:25:45] * aude nods [16:26:43] so now we are back to "how do you get profiling data"? [16:27:39] i could probably wrap our script in something that adds $wgProfiler (but sounds evil) [16:27:48] hhvm.stats.enable_hot_profiler is not set in /etc/hhvm/php.ini there [16:27:58] i see it's not and wonder where it is set? [16:28:11] possibly nowhere actually [16:28:18] * aude thought forceprofile goes there [16:28:20] we had problems with it [16:28:27] oh [16:28:49] the xenon sampling mostly replaced it [16:28:49] then i see the part about xenon in StartProfiler.php [16:29:15] xenon would also be useful [16:31:45] actually -- var_dump( ini_get( 'hhvm.stats.enable_hot_profiler' ) ); == 1 for me [16:32:10] oh [16:32:35] ~bd808/mwscript eval.php --wiki=aawiki --profiler=text [16:32:39] seems to work [16:33:10] :) [16:33:46] does that get you what you need? [16:35:31] yeah :) [16:35:54] thanks! [17:19:37] anyone up for an easy CR? https://gerrit.wikimedia.org/r/#/c/231489/ [17:35:59] no takers boo [17:42:11] ori: behave [17:42:52] ori: -1'd [17:43:00] haha [17:49:12] Reeeeeeeeeeeeeeeeeeeeedy [17:50:01] Reedy: I'm matching the whitespace conventions of the file [17:50:16] ori: there's a space before the ' [17:50:27] Reedy: hah! right [17:50:30] if( $token === 'true ' ) { [17:50:46] you're right [17:50:51] amended [17:51:48] the whitespace convention is in that file are inconsistent anyway :) [17:53:14] so i was consistent with it! [17:53:27] :D [18:18:53] fix for UBN! in MMV: https://gerrit.wikimedia.org/r/#/c/231604/ [18:19:08] i'll ask brion [18:20:42] or maybe Reedy is up for it :P [18:25:44] James_F: VE tests run for >17 mins? (https://integration.wikimedia.org/zuul/) [18:26:08] ori: CI issues – see https://integration.wikimedia.org/ci/job/npm/19321/console [18:26:36] we don't enforce a timeout? [18:26:50] thanks for the link tho, yeah, looks stuck [18:27:18] ori: It'll die after 30s of no activity. [18:27:32] ori: But if the node's process is slow rather than dead it'll crawl along. [18:27:38] ori: Normally it's 3–4 minutes. [18:27:44] ah, nod. [18:29:26] ok composer, why the feck can't you see pear/mail [18:30:48] James_F: is it the composer/github issue I saw fly by? [18:30:51] Reedy: https://packagist.org/packages/pear/mail try "dev-trunk" as the version? [18:31:00] I did [18:31:34] + "pear/mail": "dev-trunk", [18:31:44] James_F: https://phabricator.wikimedia.org/T106519 is what I was thinking of [18:32:00] Problem 1 [18:32:00] - Installation request for pear/mail dev-trunk -> satisfiable by pear/mail[dev-trunk]. [18:32:00] - pear/mail dev-trunk requires pear/pear * -> no matching package found. [18:32:31] or is it complaining about pear/pear? [18:32:52] James_F: (probably not, is the answer) [18:33:47] greg-g: No. Well, it might be; if that's running on the same CI slave and slowing it down that might do it? [18:34:14] greg-g: Also, is config dirty? [18:34:31] greg-g: 20+ hours of queued beta-mediawiki-config-update-eqiad generally means that, doesn't it? [18:35:10] eek [18:36:00] Reedy: do you have a prefer-stable in your composer.json? [18:36:19] legoktm: it's the wmf vendor repo... so yeah [18:36:28] :( [18:36:42] the only available version for pear/pear is dev-master, which isn't "stable" so composer pretends there are no versions [18:36:51] silly composer [18:37:03] hackhackhack [18:37:40] not enough [18:37:40] hmm [18:41:40] - If you require explicitly a version like dev-master, a rule is added [18:41:40] for that package that dev packages are ok. [18:42:42] it's that pear/pear [18:43:23] yup [18:43:29] done manually... [18:44:02] see what a mess this makes :D [18:44:21] tl;dr: don't use pear [18:44:42] heh [18:47:47] legoktm: ostriches: bd808: https://phabricator.wikimedia.org/T97454 [18:47:50] ffs [18:47:58] wrong copy [18:48:00] https://gerrit.wikimedia.org/r/231611 [18:48:34] is that all of pear? [18:48:50] probably not al [18:48:52] l [18:48:57] I'm sure I could bundle a lot more shit [18:49:11] "pear-core-minimal" [18:49:19] orly? [18:49:47] will mail work with that? [18:49:57] that's what you installed :P [18:50:01] https://gerrit.wikimedia.org/r/#/c/231611/1/composer.lock,cm [18:50:26] woo, minimal [18:52:31] bd808: just created https://phabricator.wikimedia.org/T109121 for pear/mail_mime-decode [18:55:12] +396830, -3 that's more than 9000! [18:55:37] I'll take a look this afternoon. another conf call awaits me now [18:56:08] I dunno if it's more or less scary then apt-get install php5-pear [18:56:08] :) [18:56:54] Reedy: speaking of email... https://gerrit.wikimedia.org/r/#/c/223369/ ! [18:59:11] ori: this.localStorage = false; looks funky. why not var localStorage = false; like the other file? [18:59:59] reasoning about js variable scope always makes my head hurt; don't do it often enough [19:00:08] bd808: because that means having to repeat the trick in line 104 [19:00:35] ah. ok [19:02:45] (bd808: you won't be able to reproduce the bug in prod any more because i cherry-picked the fix.) [19:07:11] network.cookie.cookieBehavior=2 breaks all kinds of stuff [19:07:17] like hangouts :) [19:08:02] it causes an exception to be thrown for ext.centralauth.centralautologin too, but you can't login with cookies disabled anyway [19:08:16] but for clicking on images to do nothing is a bit more severe [19:08:29] yeah. +2'd [19:09:13] https://twitter.com/baptx/status/589502685260472322 "#censorship #tracking" [19:09:16] jeez [19:09:47] #paranoid-but-maybe-for-a-good-reason [19:10:39] not really. if the non-mmv fallback worked (i.e., simple link navigation to the image page), we'd have server logs [19:10:58] this produces no logs at all [19:11:12] people like to forget about server logs when invoking paranoid delusions [19:12:13] ipv6 addrs are a good example. Pretty damn unique the last time I measured (~2.5 years ago) [20:04:30] dapatrick: whenever you're ready... google finally let me in. [20:35:19] 10MediaWiki-Core-Team, 6Discovery, 10Wikidata, 10Wikidata-Query-Service, 7Software-Licensing: Wikidata Query - add license - https://phabricator.wikimedia.org/T86833#1541167 (10Ricordisamoa) [21:09:48] legoktm: https://gerrit.wikimedia.org/r/#/c/227274/4 [21:42:58] legoktm: when you are bored -- https://github.com/wikimedia/composer-merge-plugin/pull/54 -- the guy who reported the problem says it works for him -- https://github.com/wikimedia/composer-merge-plugin/issues/50#issuecomment-131000860 [21:43:24] I've got an epic refactor follow on patch just waiting too [21:43:40] * legoktm puts in queue [21:48:54] AaronSchulz: commented [22:53:36] ori: heh. the outage that check was wanted for was in 2014-05 [22:53:44] better late than never I guess [22:54:39] bd808: much better