[02:32:36] <Krinkle>	 Krenair: No, the login failure is text coming from MW API, so presumably genuine.
[02:32:51] <Krinkle>	 But then that async stack doesn't catch it, and then it continues somehow.
[02:33:04] <Krinkle>	 that Promise.reject without Promise.catch is reported by Node.js
[02:33:28] <Krinkle>	 Given it's not a real throw/catch, it's not logically required to catch in order to continue.
[02:33:46] <Krinkle>	 E.g. $.Deferred.reject(new Error('foo')); console.log('reached'); works fine
[02:34:23] <Krinkle>	 Krenair: have you used the 'a' option yet to do a whole bunch of '$j' replacements automatically? Feels quite magical :D 
[03:32:27] <Krenair>	 Krinkle, how can that error be genuine though?
[03:32:52] <Krenair>	 no
[03:33:00] <paladox>	 Is that a promise? Promise.reject
[03:33:14] <Krinkle>	 Krenair: login session expired, bot password changed, typo in the auth.json file, account not existing on that wiki. 
[03:33:37] <Krenair>	 okay well we can rule out bot password changing and auth.json file typos
[03:34:48] <Krinkle>	 https://el.wikipedia.org/wiki/Special:Log/Krenair
[03:34:50] <Krinkle>	 2012 auto create
[03:34:50] <Krenair>	 account not existing on dtywiki, that might be it
[03:35:06] <Krinkle>	 oh async
[03:35:11] <Krinkle>	 the one before it
[03:35:16] <Krenair>	 the first error occurred under dtywiki
[03:35:24] <Krinkle>	 Yeah
[03:35:28] <Krinkle>	 https://dty.wikipedia.org/wiki/Special:Log/Krenair
[03:35:30] <Krenair>	 where my account only just got created
[03:35:30] <Krinkle>	 joined today :)
[03:36:00] <Krinkle>	 confusing async the when error is
[03:36:43] <Krenair>	 interesting that it fails to log in
[03:36:51] <Krenair>	 instead of logging in and autocreating the account
[03:39:19] <Krinkle>	 Krenair: yeah, API login is different.
[03:39:34] <Krinkle>	 Afaik we neve autocreate in those cases
[03:39:51] <Krinkle>	 there's no central auth cookie to trigger it
[03:40:51] <Krinkle>	 it's something that mainly happens on page views. maybe, if you somehow submit the login form without viewing it first, maybe it does it there as well (based on cookie).
[03:40:55] <Krenair>	 why does it need a cookie?
[03:41:08] <Krinkle>	 but afaik there's no logic to autocreate as part of login itself. if the username doesnt' exist logically, login stops.
[03:41:10] <Krenair>	 CA should see that there is a global account with those credentials but no local account right?
[03:42:04] <Krinkle>	 if before the login (on a page view,or in the login submission request before login logic is reached), if it sees the CA session cookie, it'll auto create if local is absent and token is valid.
[03:42:33] <Krinkle>	 and then probably also forego the login part, given you're already logged in.
[03:42:54] <Krinkle>	 so in practice, it's impossible to reach the login form let alone submit it, without already having been autocreated.
[03:43:02] <Krinkle>	 and thus auto-logged in
[03:43:19] <Krenair>	 so the whole autocreation system did not really take the API into account
[03:44:43] <Krinkle>	 Yeah, it's based on session authentication.
[03:45:03] <Krinkle>	 Which kinda makes sense historically, given that initially credentials were not centralised, and local clashes were allowed.
[03:45:52] <Krinkle>	 you could even attach a local account under a different name back then.
[03:46:05] <Krinkle>	 and still be auto-logged in, with the global user/luname mapping
[03:46:13] <Krinkle>	 (which still exists, although mostly pointless now)
[21:08:28] <tgr>	 mediawiki/libs/Assert has a v0.3.0 in its release notes but no corresponding git tag
[21:08:33] <tgr>	 what's up with that?
[21:26:53] <legoktm>	 tgr: https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/libs/Assert/+/v0.3.0
[21:26:59] <legoktm>	 oh, you just created that tag
[21:27:39] <legoktm>	 maybe it was supposed to be like v0.3.0-pre or something?
[21:28:05] <tgr>	 no, there's also a v0.3.1-dev section
[21:29:20] <tgr>	 anyway, sofixit territory, it just took me a while to realize I can
[21:31:26] <legoktm>	 :)
[21:32:22] <Hauskatze>	 legoktm: is the whole TrackingCategories deprecated and to be removed?
[21:32:38] <Hauskatze>	 mw docs ain't clear to me
[21:32:53] <legoktm>	 Hauskatze: the $wg global is deprecated, but the feature itself is fine, and as long as the extension is using extension.json it's fine
[21:33:09] <legoktm>	 https://codesearch.wmflabs.org/search/?q=wgTrackingCategories&i=nope&files=&repos=
[21:33:31] <Hauskatze>	 legoktm: alright. It's because I've proposed in Phab that MassMessage actually uses that for the opt-out category
[21:34:23] <legoktm>	 seems reasonable to me
[21:34:48] <Hauskatze>	 mind if I /try/ to code a patch and add you as reviewer?
[21:37:40] <legoktm>	 Hauskatze: sure
[21:37:45] <legoktm>	 should just be adding it to extension.json
[21:38:19] <Hauskatze>	 that's why I volunteered to do it, because it sounds easy :)
[21:52:48] <tgr>	 legoktm: does https://gerrit.wikimedia.org/r/c/mediawiki/libs/Assert/+/485526/1/RELEASE-NOTES.md#6 mean that codesniffer needs php 5.6.99+ to run or that it checks the code assuming it is 5.6.99+?
[21:55:24] <tgr>	 https://github.com/wikimedia/mediawiki-tools-codesniffer/blob/master/HISTORY.md#2000--2018-05-24 says "require PHP7 or HHVM to run" but it's not very clear what is required for the tested code