[00:18:04] legoktm: https://gerrit.wikimedia.org/r/#/c/222874/ ? :) [00:18:55] ori: oh I had one question, why do php libraries need a .jshintrc? [00:20:32] uhhhh [17:24:46] bd808: so...user merge is not going to go out soon. We forgot to check our update queries had indexes to support them, and didn't notice on beta because the tables aren't that big [17:25:12] que sad trombone [17:25:38] so we need to add indices to big ass prod tables? [17:26:24] for some yes. Some others we need to change to use the *_user_text instead of *_user. [17:27:14] sounds like a plan. Have you talked to Sean about the index stuff yet? [17:27:24] no, still looking to see what we need [17:27:25] https://phabricator.wikimedia.org/T105437 [17:28:49] You might be able to convince anomie to give you a hand in looking at db optimizations. He's pretty good at that [17:48:59] * anomie gets sick of sql.php being stupidly broken after a caught exception, and does https://gerrit.wikimedia.org/r/#/c/224116/ [22:20:36] TIL that HHVM has a bug that was fixed in PHP 5.2.5 -- https://github.com/facebook/hhvm/issues/5657 [23:18:40] bd808: how does Php::Composer::Require in vagrant extend the MediaWiki autoloader? [23:20:45] tgr: badly. It dirties $IP/composer.json [23:20:50] I should fix that [23:21:06] it predates composer-merge-plugin [23:22:23] that does not seem to happen [23:22:24] I guess more correctly it dirties composer.json in whatever directory it is run in [23:22:51] there's really no nice universal way to handle this [23:22:57] other than what it does [23:23:04] https://gerrit.wikimedia.org/r/#/c/218291/ [23:23:11] am I missing something? [23:25:21] I'm not sure that the things you are doing in https://gerrit.wikimedia.org/r/#/c/218291/4/puppet/modules/role/templates/elk/monolog.php.erb,unified actually work [23:25:42] I tired something similar for wikitech and it didn't take effect [23:26:02] haven't tested yet [23:26:03] which is one of the use-cases for the merge stuff [23:26:37] I though of using that but don't see a non-awkward way [23:26:45] It all depends on if $wgMWLoggerDefaultSpi has been read yet or not basically [23:27:00] once it is read for the first time it's frozen [23:27:13] from a user point of view it would be nice to have the merge command on LoggerFactory instead of (maybe) on a logger instance [23:27:48] it's not on the logger, it's on the SPI that the factory uses [23:28:02] there's an accessor on the factory to get the backing SPI [23:28:05] right, on a logger SPI instance [23:28:31] We could put it into the factory I suppose yeah [23:28:40] and make it part of the spi interface [23:28:48] anyway, ATM I am stuck at trying to install Monolog [23:29:25] *nod* we could do the dirty thing and create a stub extension for it [23:29:29] I think Php::Composer::Require should trigger a dump-autoloader run [23:29:51] it probably should [23:30:24] all the composer puppet stuff was from getting the wikidata role working [23:30:38] and hasn't been touched much since last wikimania [23:31:08] a few little fixes but nothing looking seriously at it [23:31:13] nevermind [23:31:28] the unless condition fails [23:31:46] since mediawiki includes monolog in the 'suggests' field [23:31:56] ugh [23:31:59] that's a bug [23:32:13] heh. it's pretty caveman using grep for that [23:32:24] just parse the json? [23:32:31] like I said it was good enough to get wikidata running and not much else [23:32:39] I think I added jq to vagrant at some point [23:32:54] ok, I'll look into that [23:34:10] it should probably use `composer show --installed` and grep that [23:34:53] composer show --installed|grep ^monolog/monolog [23:36:57] will require change anything that's already installed? [23:37:44] if not, then the unless is not really needed [23:38:28] and we cover the edge case of two roles requiring the same library with different version numbers [23:42:36] the unless is to keep puppet from logging that it did things that aren't really changing the system state [23:43:15] that edge case is bad too [23:43:35] having things fight over composer resources is not a good thing [23:43:45] it should puke if that happens [23:46:45] I guess that's handled by puppet's resource conflict detection [23:47:10] since both puppet rules would have to be called by the same name [23:47:29] I'll just do the composer show thing then