[01:57:12] <TimStarling>	 anomie: there's definitely no way to configure tidy to leave <style> elements alone
[01:57:36] <TimStarling>	 you'll just have to armour it somehow and reconstruct the element afterwards
[02:05:18] <TimStarling>	 the way <link> and <meta> is handled is probably identical to what you need
[02:05:25] <TimStarling>	 		$wrappedtext = preg_replace( '!<(link|meta)([^>]*?)(/{0,1}>)!', '<html-$1$2$3', $wrappedtext );
[02:07:02] <TimStarling>	 writing it on the bug
[18:35:36] <Reedy>	 anomie: Need/want help getting those merged?
[18:36:44] <anomie>	 Reedy: Please!
[18:37:39] * RainbowSprinkles mutters something about pushing with a +2 attached next time :p
[18:37:54] <Reedy>	 RainbowSprinkles: Is that well documented?
[18:38:02] <Reedy>	 If not, we should probably make sure people are awaer
[18:38:09] <Reedy>	 Cause, it's pretty useful
[18:38:13] <Reedy>	 *aware
[18:39:14] <RainbowSprinkles>	 It's in the gerrit docs!
[18:39:17] <RainbowSprinkles>	 Nobody reads docs :p
[18:39:50] <RainbowSprinkles>	 https://gerrit.wikimedia.org/r/Documentation/user-upload.html#review_labels
[18:39:56] <anomie>	 RainbowSprinkles: Self-merges are allowed?
[18:40:04] <Reedy>	 For stuff like this, yeah
[18:40:43] * anomie will try to remember that for next time then
[18:41:41] <legoktm>	 I totally missed that https://gerrit.wikimedia.org/r/#/c/321604/ happened, how should I do a DELETE FROM ........ query without creating tons of slave lag?
[18:41:55] <Reedy>	 It also stops some work for jerkins
[18:42:32] <anomie>	 legoktm: SELECT a batch of primary keys then DELETE them?
[18:42:36] <RainbowSprinkles>	 That ^
[18:42:53] <RainbowSprinkles>	 LIMIT isn't really safe in DELETE clauses if you're using transaction-based replication (AIUI)
[18:42:59] <Reedy>	 Shouldn't need a seperate select maybe...
[18:43:01] <RainbowSprinkles>	 (hence why we want primary keys everywhere!)
[18:43:06] <Reedy>	 If you've got a reasonable field to WHERE on
[18:43:08] <Reedy>	 Such as dates
[18:43:26] <legoktm>	 sorry to clarify, I'm trying to empty a db table with a million rows
[18:43:33] <Reedy>	 legoktm: behave
[18:43:35] <legoktm>	 T167758
[18:43:35] <stashbot>	 T167758: Need to empty "linter" database table on large wikis - https://phabricator.wikimedia.org/T167758
[18:43:43] <RainbowSprinkles>	 Gotta do some kind of limiting without using LIMIT
[18:44:06] <Niharika>	 Wait, there's documentation *on* gerrit?! That's news to me. 
[18:44:06] <RainbowSprinkles>	 so either select some primary key, or do ranges of data like Reedy suggested
[18:44:07] <legoktm>	 normally I would have done mwscript runBatchedQuery.php --wiki=arwiki "DELETE FROM linter LIMIT 500"
[18:44:13] <Reedy>	 use linter_id < 1000; wait for slaves; linter_id < 2000; wait for slaves
[18:44:23] <RainbowSprinkles>	 Niharika: It's from that lovely "Documentation" link at the top of every page ;-)
[18:44:33] <legoktm>	 is there already a script that does this? or do I need to write my own?
[18:44:46] <Niharika>	 Never noticed it. :P
[18:45:00] <Reedy>	 legoktm: Is it really gonna be a one off task?
[18:45:05] <legoktm>	 yes
[18:45:15] <Reedy>	 Really really? :P
[18:45:20] <legoktm>	 unless I've totally fucked up again
[18:46:09] <Reedy>	 doBatchedQuery looks like it should still do what you want
[18:46:26] <Reedy>	 Oh
[18:46:29] <Reedy>	 It's an update, not a delete
[18:46:44] <Reedy>	 (in the script)
[18:46:45] <legoktm>	 I think I'll just copy into my homedir and fork it?
[18:47:03] <Reedy>	 Maybe, yeah
[19:13:34] <RainbowSprinkles>	 PHP Fatal Error: Access level to CategoryTreeCategoryPage::$mCategoryViewerClass() must be public (as in class CategoryPage) or weaker
[19:14:48] <RainbowSprinkles>	 Also: ErrorException from line 257 of /srv/mediawiki/php-1.30.0-wmf.5/includes/libs/objectcache/MemcachedPeclBagOStuff.php: PHP Warning: Memcached::touch(): touch is only supported with binary protocol
[19:14:54] <RainbowSprinkles>	 (These are both new with wmf.5)
[19:15:47] <RainbowSprinkles>	 Former is easy
[19:18:14] <Reedy>	 public or weaker? lol
[19:19:37] <Reedy>	 https://github.com/wikimedia/mediawiki-extensions-CategoryTree/blob/1e45467be343afbf7e48502f30f7dcafab7b03c2/CategoryPageSubclass.php#L4
[19:19:45] <Reedy>	 I guess it must have to be public hten
[19:19:47] <RainbowSprinkles>	 Heh yeah
[19:20:00] <RainbowSprinkles>	 I've got a patch I'm pushing through to have it public
[19:20:05] <RainbowSprinkles>	 Needs to match core, which swapped to public
[19:20:09] <Reedy>	 aha
[19:20:21] <RainbowSprinkles>	 https://gerrit.wikimedia.org/r/#/c/358781/
[19:27:46] <RainbowSprinkles>	 That patch fixed it
[19:27:56] <RainbowSprinkles>	 Still getting the binary protocol shit again though :\
[19:33:41] <anomie>	 RainbowSprinkles: T158613 exists, if that helps
[19:33:44] <stashbot>	 T158613: PHP Warning: Memcached::touch(): touch is only supported with binary protocol - https://phabricator.wikimedia.org/T158613
[19:34:24] <anomie>	 which dups to T143464
[19:34:25] <stashbot>	 T143464: "Memcached::touch(): touch is only supported with binary protocol" from wikitech as it's not running HHVM - https://phabricator.wikimedia.org/T143464
[19:35:42] <RainbowSprinkles>	 This is labstestwiki. I thought we disabled the binary protocol on those 2 wikis....
[19:38:01] <RainbowSprinkles>	 anomie: I just filed T167826 for you btw, unrelated to ^
[19:38:01] <stashbot>	 T167826: ApiParse: Call to member function getModel() on non-object - https://phabricator.wikimedia.org/T167826
[19:43:52] <anomie>	 RainbowSprinkles: https://gerrit.wikimedia.org/r/358784
[19:44:49] <RainbowSprinkles>	 Thx, I'll cherry-pick that
[19:48:14] <bd808>	 RainbowSprinkles: we did try disabling binary and found out that blows everything up
[19:48:31] <bd808>	 err.. using binary
[19:48:31] <RainbowSprinkles>	 We could get rid of memcached on wikitech
[19:48:48] <bd808>	 we just need to get it on hhvm really
[19:49:02] <RainbowSprinkles>	 I think I could disable memcached faster than we could swap to hhvm ;-)
[19:49:38] <bd808>	 we have new boxes being racked this week that will replace sliver & californium. We could try hhvm in the new build and just fix whatever breaks
[19:49:54] <RainbowSprinkles>	 heh, considering it runs on a local machine anyway, could get away with just APCU instead of memcached.
[19:50:03] <RainbowSprinkles>	 :p
[19:50:05] <bd808>	 actually... we could really
[19:50:55] <bd808>	 the new setup was sort of planned to balance across 2 hosts, but that isn't 100% necessary
[19:51:25] <bd808>	 I really want to get wikitech into the normal mw host pool but that is still a bit off
[20:03:04] <RainbowSprinkles>	 anomie|meeting: Patch live, error's all gone. Thx
[20:16:31] <RainbowSprinkles>	 PHP Fatal Error: Call to undefined method CentralAuthUser::newMasterInstanceById()
[20:16:33] <RainbowSprinkles>	 ew
[20:19:54] <Reedy>	 o_0
[20:20:23] <Reedy>	 it's fromId
[20:20:33] <Reedy>	 newMasterInstanceFromId
[20:49:46] <RainbowSprinkles>	 Reedy: https://gerrit.wikimedia.org/r/#/c/358865/
[20:50:08] <Reedy>	 Ta
[20:51:36] <RainbowSprinkles>	 I'm backporting and will sync
[21:12:42] <RainbowSprinkles>	 bd808: Anyway back to binary mode. I thought we explicitly disabled it on wikitech & friends (because of the reasons you mentioned)
[21:28:11] <bd808>	 RainbowSprinkles: yeah it is disabled. the bug is that the memcached driver we are using can't do the touch() call in text mode. We need a newer pecl module
[21:28:36] <RainbowSprinkles>	 Then why am I getting warnings from labswebtest2001 or w/e it is?
[21:28:38] <RainbowSprinkles>	 :)
[21:28:40] <bd808>	 funny enough they "fixed" this by removing the warning
[21:29:08] <bd808>	 you are getting the "Memcached::touch(): touch is only supported with binary protocol" error correct?
[21:29:38] <bd808>	 we tried to enable binary protocol to stop that and then found that twemproxy blows up
[21:29:53] <bd808>	 so we put things back like they were (text protocol)
[21:29:57] <RainbowSprinkles>	 https://logstash.wikimedia.org/goto/efc3a97154bd2416c3f181e9bbe2f417
[21:30:18] <RainbowSprinkles>	 Yep, that's the warning we're getting in wmf.5
[21:30:52] <bd808>	 I didn
[21:31:01] <bd808>	 didn't know it ever stopped
[21:31:08] <bd808>	 that's T143464
[21:31:08] <stashbot>	 T143464: "Memcached::touch(): touch is only supported with binary protocol" from wikitech as it's not running HHVM - https://phabricator.wikimedia.org/T143464
[21:31:33] <bd808>	 we need php-memcached >= v2.2.0RC1
[21:31:43] <bd808>	 we have php-memcached v2.1.0
[21:31:54] <bd808>	 (5 years old)
[21:32:17] <bd808>	 jsut upgrading to jessie should actually fix this
[21:32:43] <bd808>	 which I think we will be doing in the next month or so
[21:33:06] <bd808>	 unless we have to sticky with trusty for OpenStack stuff...
[21:33:14] <bd808>	 which actually may be the case :/
[21:33:54] <bd808>	 if we stay on trusty we need to either backport a newer pecl module or switch to hhvm
[21:34:10] <bd808>	 Andrew and I need to talk out the options
[21:36:02] <RainbowSprinkles>	 I'm pretty sure I didn't see it last week as a warning, but ok
[21:39:54] <bd808>	 its pretty constant -- https://logstash.wikimedia.org/goto/8263849c137dd12163abd0e24f78dab6
[21:45:08] <RainbowSprinkles>	 Could've sworn we shut that damn thing up.
[21:45:09] <RainbowSprinkles>	 Oh well
[21:45:57] <RainbowSprinkles>	 I'm curious why we would even try touch when binary is disabled :)
[21:46:01] * RainbowSprinkles shrugs