[00:08:40] legoktm: is EchoNotificationFormatter deprecated now? [00:09:18] most of its functionality seems to have taken over by EchoEventPresentationModel but I can't find any documentation that would clarify the relationship [00:14:18] and is most of the stuff in https://www.mediawiki.org/wiki/Notifications/Developer_guide#Notification_parameters now deprecated as well? [00:15:42] tgr: yes, https://www.mediawiki.org/wiki/Notifications/New_formatter_system [00:24:40] yeah, I saw that one but it does not tell which parts of the old documentation have become invalid [01:03:42] You don't have to return true in hooks now, do you? [01:03:46] well, for a while [01:33:54] Reedy: since 1.22 or 23 I think [02:41:51] it was a bad design [02:44:54] ori: https://gerrit.wikimedia.org/r/#/c/300217/ [02:45:29] do you speak exclusively in Gerrit URLs? :P [02:45:37] * ori looks [02:46:35] well, I could brush up on my French [02:47:39] :) [02:52:06] AaronSchulz: why is unlockAndRollback needed in the other two places it gets called? [02:52:30] there is a patch for those too [02:53:40] in one case, it checks for races by trying to write and see how much it updates, so I changed that to do two phases, with locking reads to verify first before doing anything [03:02:50] * ori trying to test it [03:04:23] AutoCommitUpdate is nice [03:53:05] Did stackoverflow just "invent" a wiki? -- https://stackoverflow.com/tour/documentation [05:05:40] bd808: if you're wondering: "isn't this conceptually very similar to wikis?", the answer is: yes. But they're not re-inventing the wheel. There are several important features we don't have. [05:06:30] there's a system for proposing changes and commenting on proposed changes, and for adding annotations to the text [05:07:11] there are https://www.mediawiki.org/wiki/Category:Annotation_extensions but nothing (IMO) that doesn't look glued-on [05:10:07] I also think subtle changes to how content is presented and authored can amount to innovation even if the underlying functionality is essentially the same. (e.g., yes, Twitter is just Blogger with a character limit, but it gave rise to a genre) [05:10:40] here the crucial product design decision seems to have been making examples first-class citizens and demoting prose to a supporting role [05:11:19] that works very well for SO itself, and I think they're right that it could work here [05:14:01] it's a good solution to deletionism (well, question-locking), too [05:14:36] a lot of very good SO Q&A threads get killed because they don't strictly fit the parameters of what is allowed [11:39:47] Can anyone think of a reason that someone would be changing the path of $wgTorProjectCA or $wgTorOnionooCA ? [11:40:05] I'm struggling to see a reason why someone would need to. And, as such, why it needs to stay as a config global [14:31:25] Reedy: maybe some OS has a package which puts that file in another directory? [14:31:56] or something like that [14:38:04] bleugh [14:48:00] Nemo_bis: Does anyone package it? :P [15:17:31] ori: lots of good points. I certainly agree that having workflows (ask, answer, approve), annotations, and a robust tagging system are things that SO does that we could learn something from. [15:26:45] StackExchange vs. wikis is addressed at http://meta.stackexchange.com/a/261059/248268 and more https://www.mediawiki.org/wiki/Requests_for_comment/Exposure_of_user_IP_addresses#See_also [15:55:38] Can't decide if https://github.com/wikimedia/mediawiki-extensions-TrustedXFF/blob/master/update is worth updating, or just delete it [15:55:42] We don't seem to use it [16:29:27] TimStarling: Numerous XFF entries that don't resolve in DNS anymore https://gerrit.wikimedia.org/r/300571 [16:31:30] ori: Should TrustedXFF be switched away from CDB to using a json file? [16:31:46] yes [16:36:24] * Reedy files a task [16:36:31] * Reedy files a task to create a TrustedXFF project too [20:10:24] bd808: is the global rename bug considered fixed now? [20:55:13] legoktm: I think so. see https://phabricator.wikimedia.org/T141020#2488774 for my latest info on the topic [20:56:41] legoktm: is the EchoEventPresentationModel always called on the same wiki where the event got scheduled? and only once per request? [20:56:57] (trying to decide what should be cached) [20:57:17] tgr: yes, and it should only be once per request [20:57:55] well, per web request, just once. the email maint scripts are a different matter, but caching shouldn't matter much there [21:50:44] Reedy: packagers are the worst. [21:50:52] ikr? [21:50:59] Can we use manifest version 2 now? [21:51:02] yes [21:51:12] zomg [21:51:20] I guess we should fix TrustedXFF similarly [21:51:30] (not merged yet) [21:51:48] oh wait, convertExtensionToRegistration won't work with the new version [21:51:58] not sure that matters too much [21:52:05] if we can define to the spec, and use it.. [21:54:41] legoktm: Do we have an example of the path stuff? [21:54:56] no, it hasn't been written yet :P [21:55:04] OH [21:55:08] GETTING MY HOPES UP [21:55:11] legoktm: you're the worst [21:55:15] writing it now! [22:06:33] legoktm: EchoEventPresentationModel::getPrimaryLinkWithMarkAsRead gives me the impression marking as read won't work out of the box when the primary link goes to a special page [22:06:48] is that something I need to implement manually? [22:10:09] tgr: uh, why wouldn't it work? that query parameter is read in the PersonalUrls hook [22:11:02] legoktm: Merged your Pingback show-the-json change [22:11:09] :D thanks [22:13:38] ostriches: also, is there a way to force override " ! [remote rejected] HEAD -> refs/for/master (no changes made)" when I know for a fact gerrit is wrong? [22:13:52] Nope. And are you sure? :p [22:14:18] * anomie tends to get that when he forgets to `git commit` [22:14:36] legoktm: ah, ok, thanks [22:14:38] yes, pretty sure :P [22:14:46] seems a bit scary to use an unprefixed parameter for that [22:15:21] maybe... [22:45:56] legoktm: I closed T141020 as resolved. I don't see any evidence that unattached local accounts will continue for that particular cause [22:45:56] T141020: Users with detached local accounts after rename during train deployment of 1.27.0-wmf.11 - https://phabricator.wikimedia.org/T141020 [22:46:26] greg-g: I wrote it up on wikitech and sent to ops-l too -- https://wikitech.wikimedia.org/wiki/Incident_documentation/20160721-LocalRenameUserJob [22:47:25] * bd808 wonders how to make the index of incidents at https://wikitech.wikimedia.org/wiki/Incident_documentation more useful [22:49:56] anomie: how should the job be delayed in T141160? onTransactionIdle? [22:49:57] T141160: Unattached accounts created on registration - https://phabricator.wikimedia.org/T141160 [22:55:52] tgr: I think that'd do it.