[01:08:03] wtf https://3v4l.org/5UESC [01:08:57] legoktm: loose comparision [01:09:43] https://3v4l.org/0FKio [01:09:45] legoktm: Aka "PHP hates you". [01:10:03] sigh. [01:10:05] bd808: thanks [01:11:58] > Passed 1337 of 1337 tests (100%)... ALL TESTS PASSED! [01:12:00] yay [01:12:07] l33t [01:13:11] the loose check passes because PHP sees a string and an int and decides to cast the string to an int to compare [01:14:21] (int) 'noclasses` === 0 -- https://3v4l.org/2HOLj [15:14:04] anomie: the session patches should be backported regardless of what happens to AuthManager, right? [15:16:30] tgr: The "force-use" patches fix a bug in OAuth, so yes. The invalidation patches wouldn't hurt, but I can't think of anything that breaks without them. [15:18:07] nothing breaks but extensions don't have to support both invalidation syntaxes for three years that way [15:37:39] anomie: hi, will you have a few minutes today to look at the API parts of https://gerrit.wikimedia.org/r/287103 ("Refactor upload dialog to make it configurable")? [15:37:52] Krinkle: hi, will you have a few minutes today to look at the ResourceLoader parts of https://gerrit.wikimedia.org/r/287103 ("Refactor upload dialog to make it configurable")? [15:38:10] (i also wouldn't mind if you reviewed the whole thing ;) ) [15:39:40] ostriches: by the way, https://gerrit.wikimedia.org/r/287103 is another thing i'm trying to wedge into 1.27, i hope that's okay. [15:40:59] MatmaRex: The API code itself looks ok, although the new output data structure seems crufty to me as we already discussed via comments in gerrit. The META_BC_BOOLS setting probably isn't needed, but doesn't really hurt anything besides making the formatversion=1 output not follow the usual conventions. [15:42:37] anomie: it's needed so that i have the same structure for the mw.config data and the siteinfo data. (otherwise i'd have false/true in one, and undefined/'' in the other.) although i guess i could pass formatversion or something to avoid it in the query, if you prefer that? [15:43:05] MatmaRex: Using formatversion in your query would be preferable, IMO. [15:43:10] legoktm: Remind me, does CentralAuth care about backwards compatibility with old versions MediaWiki, or is it "CA master is compatible with MW master only"? [15:43:28] alright. can do that :) [15:43:32] * anomie can never remember which extensions have that policy [15:53:10] anomie: master/master. The other people who use CA use the release branches and we do backport stuff ocassionally when they request it [15:54:10] legoktm: Good. Should we also get rid of the pre-CentralIdLookup hooks for OAuth? [15:54:21] sure [16:30:54] MatmaRex: I'll check it out, but might not make it until tomorrow. [16:31:54] legoktm: Is there any particularly sane way to do CentralAuthHooks::onRunExtensionFunctions() in extension registration, or is that about it? e.g. will it work right to make those sub-extensions and call wfLoadExtension() from the callback or from $wgExtensionFunctions? [17:25:48] legoktm: It looks like wfLoadExtension() will fail from either place. ExtensionRegistry->load() would work, more or less, from $wgExtensionFunctions but will screw up badly if called from a callback. Fixing the ->load() from a callback and allowing (and loading) ->queue() calls during callbacks inside ->loadFromQueue() both seem easy enough to fix, if we want to. Do we want to? [17:28:38] anomie: I think for now it's best to just leave those like they are in an extension function. In https://gerrit.wikimedia.org/r/#/c/250060 (which is supposed to make it into 1.27) we are going to require all extensions to be loaded at once for dependency resolution, and not allow anything to be loaded after that [17:30:22] ok [17:49:02] * James_F legoktm: Off-topic, but do you have a plan for soft-dependency conditional RL loads, BTW? [17:49:47] James_F: I have vague ideas right now, but nothing concrete [17:49:53] * James_F nods. [17:50:05] I loathe the hacks we do related to VE for those. [17:50:16] No rush though. [22:04:28] tgr: FYI, I went through all extensions in wmf/1.27.0-wmf.23 looking for stuff that might need changing for AuthManager, and put some details as to what I found on most of the existing tasks. [22:04:50] yeah, saw that [22:04:51] Most of the additions were "This uses the AddNewAccount hook, which has been deprecated since 1.26wmf24" [22:04:54] great job, thanks [22:05:31] should we reach out to the owners or is it less effort to just do it? [22:07:20] if you have the hook emit deprecation warnings, most people should start fixing it quickly [22:07:57] Some of both, probably. For example, Translate and TranslationNotifications would be nice if the Language people would look at that. T111483 (MobileFrontend) is ugh, the recent retitle seems to be trying to put it on AuthManager to keep their garbage working rather than them reevaluating their garbage. [22:07:57] T111483: AuthManager should retain the MobileFrontend watermark at the top of the login screen - https://phabricator.wikimedia.org/T111483 [22:08:42] watermark? [22:09:06] the "W" on https://en.m.wikipedia.org/w/index.php?title=Special:UserLogin&returnto=Main+Page&returntoquery=welcome%3Dyes ? [22:09:24] MF hides the skin and then modifies the auth form to put the logo back on the login page [22:09:41] -.- [22:09:49] probably just need to add a HTMLInfoField to the top of the form [22:09:51] Yes, the "W"... [22:29:05] tgr, legoktm, anyone: Are https://gerrit.wikimedia.org/r/#/c/287432/ and https://gerrit.wikimedia.org/r/#/c/287500/ not merging because Jenkins isn't set up to merge in those repos, or is/was something broken? [22:29:41] because of the Depends-On [22:29:48] Oh, duh. [22:30:03] maybe I should remove that, technically it's not a dependency [22:30:13] +1 [22:34:22] anomie: done [22:34:43] hm, although the tests might break without that [22:34:59] since they use AuthenticationRequestTestCase