[00:00:26] We have code in core for adding stuff to the session so it seems like yeah that's a global right [00:00:34] bd808: this is re: -- Echo making multiple round-trips to memcached per logged-in page view for data like notification counts [00:01:04] I think they have been consolidated into a single pipelined get, but I suggested that storing it in the session would be even better [00:01:29] Roan wasn't sure what I was talking about, and Roan knows his stuff, so I wanted to check that I'm not saying something dumb. [00:01:43] it's maintained outside of sessions though right? [00:01:54] I mean those valuse are precomputed by jobs or something [00:02:09] or are they not? [00:02:17] I have no idea [00:03:26] that would make sense, in which case it's not a smart suggestion [00:03:35] thanks [00:03:39] sure [01:27:10] Argh! PHPMocks are so annoying. $logger = $this->getMock( 'Psr\Log\NullLogger' ); $logger->expects( $this->once() )->method( 'warning' )->with( 'x' ); $logger->warning('x'); [01:27:12] doesn't match [01:27:16] because warning is final [01:27:27] making it check method('log')->with(*'warning', 'x') instead works [01:27:34] *face palm* [01:27:40] no run-time error or anything [01:28:13] meanwhile.. 20 echo debug statements in various places later.. [01:34:00] wait no it's still not working [01:34:01] wtf [01:34:53] ah, expectation has to be set beforehand [07:52:57] Krinkle: https://www.mediawiki.org/wiki/User:Tgr_(WMF)/Mocking has some comparisions with an IMO saner mocking framework; I want to turn it into an RfC eventually [07:53:33] mocking finals is impossible with any framework though, unless you do something drastic like using runkit [08:11:18] bd808: could you set up a web proxy from authmanager-disabled to am-02? [08:11:33] * tgr still can't log in to horizon [14:34:55] Nice heads up [14:35:01] Upgrade to 16.04, and you get PHP 7 [14:35:04] MW seemingly breaks [14:55:44] php5-memcache(d)? packages dropped the 5 [15:05:44] tgr: authmanager-disabled.wmflabs.org -> http://am-02.authmanager.eqiad.wmflabs:8080 [15:16:44] thanks! [15:32:07] tgr: When you have some time and andrew is online you should work with him to figure out what's wrong with your horizon access [15:34:34] I thought I would be polite and wait until someone gets around to https://phabricator.wikimedia.org/T131630 [15:54:12] tgr: yeah... nobody is ever going to get around to that. Best bet is you attempting to login repeatedly while andrew watches logs on the horizon server. [15:55:06] I'll try [16:19:01] bd808: Oh, you wrote I86168c64. Why does it not try globalizing the account if it checked (and failed) the login with username~wiki? [16:25:48] anomie: Globalize the username~wiki account? [16:26:33] If I'm understanding the question that isn't done because username~wiki would always be a CA account. The ~ is banned in local usernames [16:26:51] bd808: Try to login as Username. It fails, so it tries Username~wiki which also fails. Why not go on to try globalizing Username like you would have if Username~wiki doesn't exist? [16:28:29] * bd808 tries to remember [16:31:16] bd808: Well, the second "globalize" case is "unattached local with a matching central password" which can't be the case because the central password check failed. What about the "no central account, but all remaining unattached accounts have the same password" case though? [16:32:19] anomie: I honestly don't remember. The hard finish was added in PS2 so I'm sure that I talked it through with legoktm and csteipp at the time. Maybe it's related to legoktm's comment on PS1 -- "We probably don't want to start doing this if they're a renamed user..." re automigrate [16:36:40] bd808: I think the comment on PS1 might have been more aimed at $passwordMatch being set to true thanks to $renamed but $central is still the old user... [16:38:55] is there a tool to work out why nix is throwing permissions errors? [16:38:58] I wonder if that return was meant to be one level deeper; it's at a weird place [16:39:03] anomie: looking at it all again --- I think that there is an assumption that if wgCentralAuthCheckSULMigration = true then we are past the point of auto-globalization that will otherwise happen in the "if ( !$central->exists() ) {" block [16:39:47] tgr: yeah things are a bitt different in later patches [16:40:34] today there is another check to ensure that username~wiki actually exists locally [16:42:05] anomie: by "past the point" I mean that the flag is meant to be enabled at the point in a SUL migration where everyone who can be trivially migrated to a global user has already been handled [16:42:29] and CA should be the strict auth method across the farm [16:42:49] All that renamed user stuff is scheduled to get ripped out now [16:43:17] we are past the 1 year for people to deal with it date [16:44:11] Maybe it should get ripped out before you try to replicate it in the new framework? [16:44:29] T127887 [16:44:30] T127887: Disable ability to log in with old user credentials (pre-SUL finalization) - https://phabricator.wikimedia.org/T127887 [16:45:12] that codepath is getting used ~100 times per day [17:14:15] oops, I was going to do that [17:32:45] bd808: That renamed-user stuff in CentralAuth wasn't bad at all. Now, if someone wants to rip out OpenStackManager before we have to update LdapAuthentication... [17:33:29] we're working on killing the need for OSM but it will take a few more months [17:34:08] T87279 is the list of things that need doing [17:34:08] T87279: Make OpenStack Horizon useful for production labs - https://phabricator.wikimedia.org/T87279 [17:34:32] the puppet role management is happening next [20:38:40] https://gerrit.wikimedia.org/r/#/c/284976/1/composer.lock,unified [20:38:52] Ooh, it's not reorganising itself [21:09:51] SMalyshev: Do you know any details on the php openssl issue that was recently patched in ubuntu (http://www.ubuntu.com/usn/usn-2952-1/)? No cve, so I'm trying to find the orignal issue somewhere.. [21:18:58] csteipp: I left some replies for you on https://gerrit.wikimedia.org/r/#/c/240052/ [21:20:18] csteipp: hmm... let me look it up [21:31:35] csteipp: please see https://bugs.php.net/bug.php?id=70014 [21:33:21] SMalyshev: Awesome. Thanks!