[00:16:00] hello [14:47:43] Anyone want to merge https://gerrit.wikimedia.org/r/#/c/207450/ in time for SWAT? [14:55:22] anomie: so that patch is just going to cause the log entry to throw a different exception? [14:56:12] legoktm: No, it'll replace the truncated UTF-8 character with � [14:56:56] ok +2'd [14:59:49] <^d> legoktm: I got a patch up starting the process for adding mail and mail_mime from packagist [15:00:42] woot [15:01:34] <^d> If we ship it in the vendor repo we could drop the debian packages from the puppet role (which are likely ancient :p) [15:06:13] will there be any issues if we do both? [15:06:46] <^d> Nah, it'll just pick the first one it finds in the include path [15:07:04] <^d> I assume vendor is higher up than /usr/share/php/ or w/e debian's using these days [15:08:45] <^d> s/nah/it shouldn't/ [15:08:51] <^d> Gotta hedge my bets :p [15:10:57] do autoloaders take precedence over the include path? [15:16:40] <^d> legoktm: Should? [15:32:32] "do autoloaders take precedence over the include path?" -- no. Autoloader only works based on a class being referenced without having been defined in a file that is explicitly sourced. [15:33:10] so if we do `require PEAR/something.php` that bypasses autoloaders and uses the search path [15:34:34] ...right. [15:34:40] and we do require_once 'Mail/mime.php'; [15:35:30] so the fix would be to remove that after we have Mail_Mime in compsoer.json and vendor [15:35:32] so we'd probably want to see if the class_exists() first, triggering the autoloader, before trying to require it directly? [15:35:53] yeah, that would work [15:36:11] belt and suspenders style :) [15:37:10] A block that did that could even whine politely to switch to the composer loading method [16:03:37] I'm looking for orphaned revisions [16:03:40] > SELECT rev_page, COUNT(*) FROM revision LEFT JOIN page ON rev_page = page_id WHERE page_title IS NULL GROUP BY rev_page; [16:03:46] but that's probably going to take a while? [16:15:56] AaronSchulz: OK to merge https://gerrit.wikimedia.org/r/#/c/207384/ ? [16:44:35] reviews on https://gerrit.wikimedia.org/r/#/c/207486/ would be appreciated [17:14:28] <^d> bd808: I have no idea. https://phabricator.wikimedia.org/T95282#1243408 [17:20:42] <^d> bd808, legoktm: Might want to double check that. iirc, class_exists() returns false, triggering autoloader. [17:20:49] <^d> So you have to check twice before requiring? [17:20:59] <^d> (or maybe I did it wrong awhile back...) [17:21:38] ^d: That's a known HHVM bug [17:21:44] there is a dup ticket somewhere [17:21:58] <^d> nvm on class_exists() [17:24:20] <^d> Yep, definitely same bug [17:24:36] <^d> I guess it probably doesn't block the 1.25 release. [17:24:48] <^d> It's "just" testing and "just" hhvm. [17:24:57] <^d> Plus, who knows when upstream will land + release a patch [17:37:22] anomie: https://gerrit.wikimedia.org/r/207501 does it properly (I think!) in MovePage [17:47:36] Hrm. I hit the wrong button, and Gerrit let me. [17:51:09] :| [17:52:14] Krinkle, sure [17:59:35] Ooh, now WMF has a "Byron", to go with "Bryan", "Brion", and "Brian". [18:04:43] csteipp: are you at R66? [18:04:49] i'm remote but there's no one on the call [18:04:57] yes [18:08:15] http://imgur.com/8uDY1WM [18:08:46] well, ping me if anyone wants my input or anything [18:11:43] and linked to in #wikimedia-kawaii ;) [18:21:04] Anyone ideas for how to better disambiguate Templates and Templating? [18:21:04] https://phabricator.wikimedia.org/T93546 [18:21:19] wiki-templates vs html/mustache templates [18:21:31] but without the latter being specific to html or mustache [18:24:05] <^d> templates and templatesv2 [18:36:06] ^d: Oh I like it :P [18:36:15] No, templatesv5 [18:36:28] other-templates [18:45:39] legoktm: https://gerrit.wikimedia.org/r/#/c/207512/ comments [19:02:31] Krinkle: ping, yt? [19:02:41] Yep [19:02:45] pm [19:02:55] Currently taking a shot at replacing timestamps with hashes in RL. [19:14:49] anomie: Do you know anything about ContactPage? [19:15:23] bd808: If I say yes, how much extra work does it mean? [19:15:23] Just looking for a consultation [19:15:48] on an patch that K.renair made for an affcom request [19:16:03] Sure, let me look [19:16:09] https://gerrit.wikimedia.org/r/#/c/207332/5/wmf-config/AffComContactPages.php,unified [19:16:43] He's trying to determine which form is being passed to the EmailUserForm hook [19:16:57] and I'm trying to follow the bouncing ball [19:21:08] So it looks like his trick should work [19:21:46] HTMLForm::getLegend( $key ) returns $this->msg( "{$this->mMessagePrefix}-$key" )->text(); [19:22:37] so assuming nobody ever makes the message "contactpage-affcomusergroup-do-not-create-this-message" then that guard should work [19:36:53] anomie: if you think there is anything horrible in that config patch let me know and I'll try to fix it up. I think I convinced myself that the guard condition will actually work. [19:38:59] bd808: Really that hook shouldn't be necessary in the first place. Writing lots of comments. [19:39:11] stellar [19:39:33] Just know that I'm going to get stuck fixing them ;) [19:42:51] bd808: Summary: two hopefully-simple changes to ContactPage, and use a different hook to override the subject. [19:43:54] :~ [19:44:11] slaporte wants the form by Friday [19:44:35] bd808: Want me to write the ContactPage changes? [19:45:34] anomie: if you have the time and energy, yes please [19:46:16] and we need to rename/add some i18n in WikimediaMessages? [19:46:58] Probably [19:51:21] ^d: did you have to do anything weird to get elasticsearch to start when you were playing with jessie? [19:51:43] On logstash1004 /etc/init.d/elasticsearch start gives me -- Starting elasticsearch (via systemctl): elasticsearch.service. [19:52:02] but then no java process and no logs in /var/log/elasticsearch [19:52:18] <^d> Hmm, second [19:54:23] journalctl -u elasticsearch is empty too :( [19:58:08] <^d> systemd thinks it started [19:58:11] <^d> pasting, sec [19:58:41] <^d> https://phabricator.wikimedia.org/P575 [19:59:32] hmmm [20:00:21] and yet "ps ax|grep java" is blank [20:01:04] <^d> I don't remember having this problem. But granted I also dropped jessie real quick when we decided for 1:1 parity with prod [20:01:44] k. I'll put this on the back burner and focus on the ContactPage thing first [20:01:52] I can dork with it later [20:02:56] bd808: https://gerrit.wikimedia.org/r/#/c/207548/ [20:06:32] Hm.. from now on I will judge people's age by the first digits they remember seeing in their first epoch timestamps [20:06:46] I just realised we entered the 14xxxx epoch digits [20:06:55] It used to be 12xx and 13xx [20:07:56] Krinkle: http://www.epochconverter.com/epoch/timestamp-list.php [20:08:15] for me something around 725846400 [20:08:53] That would have required me to have ever paid that much attention to the digits in epoch timestamps... [20:09:38] yeah, I've seen them since 1998 probably, but only really worked with them since 2010 when debugging ResourceLoader [20:09:41] I do remember when it rolled past 1,000,000,000 [20:10:06] and 1234567890 [20:10:43] <^d> legoktm: 10 error: Call to undefined method CentralAuthUser::getSlaveDB() in /srv/mediawiki/php-1.26wmf3/extensions/SecurePoll/includes/user/Auth.php on line 273 [20:11:01] <^d> Oh secure poll, I read that as centralauth [20:11:08] <^d> Hmm, Tim was working on it [20:11:25] <^d> Wait, SecurePoll broke using CentralAuth [20:11:29] <^d> I did read that right [20:12:34] it should be getCentralSlaveDB [20:12:35] i have a fix [20:12:50] (jamesofur just pinged me about it) [20:16:29] Krinkle: I think I remember you having a script to convert wmf branches into tags? [20:17:16] legoktm: yes [20:17:34] I'ts on the phab ticket that asks for conversin [20:18:18] https://phabricator.wikimedia.org/T1288 [20:18:20] https://phabricator.wikimedia.org/T1288#836914 [20:18:32] "Note that this requires Gerrit administrator permissions to run." :( [20:19:08] because it involves deleting branches and pushing tags with authors other than you [20:19:55] <^d> I should just give everyone admin rights to gerrit [20:20:21] anomie: thanks. setting up to test and figure out what else I need to change [20:20:21] sounds good to me! [20:21:22] legoktm: Do document when you've ran it on a project :) [20:21:32] May as well use that task [20:21:39] Krinkle: well I'm not an admin... [20:21:58] I figured Chad made you one by now [20:22:02] :P [20:22:21] You may have the permissions needed already on the relevant project [20:22:23] where are you looking to run it [20:22:31] <^d> whoops, there's a new admin! [20:22:43] :o [20:22:44] thanks [20:23:11] legoktm: you keep falling in holes full of new rights on the cluster ;) [20:23:17] <^d> bd808: Who owns RequestContext? [20:23:20] <^d> :p [20:23:47] MasterBlaster runs BarterTown! [20:23:49] Krinkle: MassMessage, but I have all the extensions already checked out so I might as well do it for all of them [20:25:41] ^d: I nominate Daniel Friesen [20:25:51] <^d> :D [20:26:24] legoktm: https://github.com/Krinkle/dotfiles/blob/master/hosts/KrinkleMac/modules/functions.sh#L19-L27 [20:26:39] I ran domwextforeach ~/tmp last time I ran it [20:26:43] where tmp is the pasted script [20:26:54] and set the target version accordingly [20:26:58] inside the script [20:27:48] Try it one first though, I don't know how well it tolerates error. You dont' wanna end up deleting brarnches without pushing tags and losing the referenes [20:28:07] ah, I could just do "git submodule foreach" ... [20:28:26] I stopped using the metarepo [20:28:30] But yeah, that works :) [20:30:18] TimStarling: fyi, https://gerrit.wikimedia.org/r/#/c/207552/ [20:30:22] I just synced that [20:34:12] thanks [21:18:17] legoktm: If an extension (ContactPage) is loading in mw-vagarnt via extension.json, do I have to do something other than populate globals to feed it extra config? [21:18:47] bd808: nope [21:19:39] it *should* just work, but there might be some bugs in edge cases [21:19:52] I'm going to test again... [21:22:39] legoktm: seems to work now. I must have messed something else up earlier [21:30:52] ori: Hm.. wanna bounce off some ideas about hashing when you have time. Got it mostly figured but blocked on the client-side part. We need a way to combine sha1 hashes into a new one. Ideally without shipping a sha1-js, though we could just do that. [21:36:31] ori, I hope the multi-dc stuff doesn't break labswiki with all it's weird extensions :) [21:42:51] labswiki has, what... OSM and SMW extensions? [21:43:31] and presumably labswiki stuff would only ever be in eqiad [21:45:10] ldap stuff too [21:45:25] as long as it's just eqiad it's fine [21:54:05] anomie: ping [21:54:15] Or anyone else involved in the ApiResult changes [21:54:49] Krinkle: I have an interview scheduled for this coming hour, but am free afterwards, if you're still awake. [21:57:09] ori: Yeah, I've got another 3-4 hours here. [23:30:36] * bd808 grumbles about mw-vagrant not sending emails out of the box [23:30:52] guess I have to fix that to keep testing these contactpage changes