[13:03:59] tgr: Because I forgot to add @deprecated to its doc block. [13:35:05] anomie: any thoughts on https://phabricator.wikimedia.org/T135884 ? that seems to be our main blokcer for now [13:36:02] could we make a User::newSystemUser-like API for selenium? (except it would set a normal password) [13:38:59] tgr: The shared-secret SessionProvider isn't a bad idea, assuming it doesn't test actually logging in. And if it does test actually logging in, the test probably needs to be rewritten for AuthManager anyway. [13:40:00] tgr: If it controls $wgAuthManagerConfig, it could easily enough just change its createaccount API calls to match whatever the configured AuthenticationProviders are expecting. [13:41:02] breaking a few specific tests is acceptable IMO, breaking all tests requiring a user not so much [13:41:33] (that may be as simple as changing "name" to "username" and "token" to "createtoken" and adding a dummy "createreturnurl") [13:42:09] (oh, and adding "retype") [13:42:27] we have some tests running on beta and testwiki, I think, tests don't control $wgAuthManagerConfig there [13:42:41] then again it won't contain anything unexpected either [13:43:18] updating the usercreation logic seems the most viable short-term solution, I'll look into that [13:43:51] I don't know much of anything about selenium, but if you need help converting the actual API queries it's making let me know. [15:18:40] bd808: around? [15:18:53] hey aude. what's up? [15:19:16] i am updating the wikidata build (with Composer version 1.1.1 2016-05-17 12:25:44) [15:19:29] Don't use 1.1.x [15:19:29] and it's adding autoload_static.php which concerns me some [15:19:32] ok :) [15:19:35] what should i use [15:19:43] you have to stick with the 1.0.x compser for now [15:19:46] ok [15:20:05] some details at T135161 [15:20:05] T135161: Composer v1.1.0 generated vendor dirs will fail lint by PHP <5.6 - https://phabricator.wikimedia.org/T135161 [15:20:13] i saw that [15:20:37] simply introducing the file also concerns me, espeecially in swat [15:21:05] The file will break sync-dir if nothing else [15:21:34] And upstream is mostly right that this is our problem and not theirs [15:22:13] how do i get composer 1.1.0 ? [15:22:20] trying to figure otu how to get specific version [15:23:16] oh, it's an argument that i can provie [15:23:18] provide* [15:23:26] I'm not sure short of manually downloading the older phar or doing a rollback if you upgraded recently [15:24:05] oh, you're right `composer self-update 1.0.0` looks like it should work [15:24:05] composer self-update 1.0.3 [15:24:13] yeah [15:28:26] bd808: thanks :) [15:46:06] anomie: Were you involved in the api log grepping for people making requests to HTTP? [15:46:22] Or anyone else in here? [15:47:59] Reedy: I don't know about any grepping, but https://logstash.wikimedia.org/#/dashboard/temp/AVSbGv_JDxp7yus2LXCd [15:48:54] aha, thanks [15:49:51] It looks like the AWB ones were mostly "rogue" users, so not a big deal [15:53:39] anomie: what's the most robust way of checking which account creation API is in use? [15:53:55] calling paraminfo and checking classname / params seems fragile [15:55:20] tgr: Probably using action=paraminfo and checking the parameters. Or possibly seeing if action=query&meta=authmanagerinfo&amirequestsfor=create actually returns stuff (versus an error or the response from ApiDisabled). [21:37:54] ostriches: https://gerrit.wikimedia.org/r/#/c/290269/ fixes the test breakage that enabling AuthManager would cause, but apparently there is no one still up who could review it [21:54:22] tgr: I have no idea what this stuff does [22:17:02] ostriches: what it says in the commit summary, hopefully :) [22:17:23] anyway just reporting the status because you wanted to reenable AuthManager today