[01:05:41] anyone available for a quick review? https://gerrit.wikimedia.org/r/339334 [01:11:12] MaxSem: is "unsupported bt" a typo or some acronym? [01:11:37] "by" [01:12:11] tgr, fixed [01:13:08] MaxSem: what does unsupported by RL mean here, anyway? does it break the CSS minifier? [01:13:43] it fails to strip #fontname from filename, and doesn't find it as a result [01:14:03] thus fails structure tests [01:15:01] +2'd [01:15:04] thanks! [04:32:57] bd808: https://gerrit.wikimedia.org/r/#/c/339345/ if you can :) [04:35:19] Krinkle: neat! [04:41:24] Thanks [04:59:22] bd808: Yay, first green in over a year [04:59:23] https://travis-ci.org/wikimedia/mediawiki/builds/204469354 [04:59:31] not all of the matrix yet, but it's a start :) [20:12:28] Krinkle: And now it's green for PHP7 too; just PHP55/postgres that's failing. [20:29:08] lol [20:32:26] It'd be tempting to stop testing with postgres. ;-) [20:34:57] Certainly at least till someone said the tests are working locally with it [20:45:53] James_F: yeah, I broke that actually with my php5/7/hhvm fix [20:46:08] had to change mysql user from travis to root but forgot to update the psql code [20:46:14] Trying a fix now at https://travis-ci.org/Krinkle/mediawiki/builds/204736240 [21:03:49] The command "if [ "$dbtype" = postgres ]; then psql -c "CREATE DATABASE traviswiki WITH OWNER root;" -U postgres; fi" failed and exited with 1 during . [21:03:58] ERROR: role "root" does not exist [21:17:29] Once we get these fixed, how will we keep them fixed? Are we going to suddenly add postgres to Wikimedia CI? [21:25:43] James_F: lolno [21:26:20] Indeed. [21:26:28] Ops would be… unhappy. [21:28:07] I think releng would be unhappy having to support it too :) [21:41:04] Here I go, breaking users' hearts: https://www.mediawiki.org/wiki/Topic:Tlje635i30vqbod4 [23:01:34] James_F: there's a bug for adding postgres to CI, I don't see any reason why not to have it as a postmerge job [23:06:02] Because we'd have to install and support postgres in our VMs? [23:06:03] :) [23:06:26] (i'd rather us *stop* pretending to support so many DBs we half-ass) [23:07:16] probably as hard as adding require_package('postgres') to some puppet file [23:07:21] last time I suggested we drop MSSQL I was called a dick [23:07:57] but it would be confusing to have a postgres CI job without actually committing to fix postgres problems [23:08:09] MaxSem: Yeah, I was there. For both attempts [23:08:12] tgr: Indeed [23:08:22] If a tree falls in the forest and nobody cares.... [23:08:33] this is one of those problems where we are being bitten by our utter lack of user research [23:08:57] if 50% of MediaWiki reusers run on postgres, we can probably find the motivation to care [23:09:14] See, I think Phab has the right idea. Pick a DB backend, support that *well* and others be damned [23:09:22] if it's 1% then meh [23:09:29] I've seen other software that requires Postgres and no mysql [23:09:41] Why is picking a single db backend a bad idea? [23:09:58] Phab is an application [23:10:20] ....so is MediaWiki? [23:10:27] MediaWiki is a software with identity crisis which does not know whether it's an application or a framework [23:11:04] It's an application, if massively over-extensible [23:11:05] :) [23:11:58] writing plugins for Phabricator requires a serious amount of masochism, they change the internal APIs all the time, the external APIs suck [23:12:15] I wish writing extensions for MW was as hard ;-) [23:12:15] mysql:research@analytics-store.eqiad.wmnet [log]> select count(*) as num, event_database from MediaWikiPingback_15781718 group by event_database order by num desc; [23:12:24] | 8566 | mysql | [23:12:24] | 265 | sqlite | [23:12:24] | 149 | postgres | [23:12:24] | 3 | mssql | [23:12:28] they don't really try to serve uses cases which are different from what the Phab developers envision [23:12:59] MaxSem: that's only 1.28+ though [23:13:22] that's a slice of usage per some period [23:13:35] not usage, upgrades [23:13:39] 0.03% mssql [23:13:58] 0 oracle ;-) [23:14:15] essarily need to care about what was in the past - if someone's still running 1.5 because they like PHP4, let them run it [23:14:27] *no necessary [23:14:29] IMO we shoult get alternative DB backends out of core and set up webhooks so if you want to run Travis with your DB plugin for every MW patch you can [23:14:30] *not [23:14:34] fuck you typing [23:15:00] tgr, OMGTHINKOFTHECHILDREN [23:15:36] but having better stats would certainly help to make that decision [23:15:44] pingback is great but a bit young [23:16:58] wikiapiary does not seem to track DBs [23:21:34] what do you know, it does track them [23:21:45] god those semantic interfaces are horrible [23:22:23] pingback was always designed so we could get db stats and make a reasoned (data-driven) decision to kill the ones we don't like :p [23:22:34] Also so we can decide when old php versions don't matter [23:22:46] Have we made that very visible yet? [23:22:50] James_F: your heart breaking post actually seems correct to me. I have advocated recently though to guarantee security fixes for 'extended core' including VE, Parsoid, and RESTbase [23:23:25] extended core sounds scary and like I'm going to be on the hook for more releases [23:23:26] but certainly LTE does not mean feature backports [23:24:05] RainbowSprinkles: magical container fairies will handle it according to some ;) [23:24:17] Pfft. Yeah, and I'm winning the powerball tomorrow [23:24:33] RainbowSprinkles: buy me a pony? [23:25:15] related stuff for extension LTS: T156500 T156640 [23:25:16] T156500: Document extensions' MediaWiki version compatibility better - https://phabricator.wikimedia.org/T156500 [23:25:16] T156640: Improve LTS support of extensions - https://phabricator.wikimedia.org/T156640 [23:25:33] bd808: https://www.amazon.com/Aurora-World-Little-Princess-Celestia/dp/B010GL554S/ [23:26:13] so fluffy I could die. :) [23:26:33] I'll take one, if you win the lottery [23:27:37] I would have expected you to pick Derpy for me though [23:29:59] You're assuming I picked or put any thought into it [23:30:07] I just typed my little pony into the amazon search box, first result [23:30:22] Er, second overall result, but first toy [23:31:40] hmmm https://github.com/wikimedia/mediawiki/commits/master/maintenance/oracle [23:32:03] anomie, did you really test your patches on Oracle? :P [23:46:40] MaxSem: got the data out of wikiapiary after an epic battle [23:47:17] mysql: 21214, postgres: 283, sqlite: 144, oracle: 3, mssql: 1 [23:47:27] ffffuuuu [23:47:42] Aren't "we" spending more time on mssql than postgres atm? [23:47:44] so 1%-ish [23:48:28] time for an RfC? [23:48:34] "we" should just kill SQL server and Oracle support with fire and stop thinking about it [23:49:51] should we have some migration path? [23:50:13] dump / import [23:50:14] :) [23:50:34] As long as dumping to xml works on those :P [23:51:23] make the DB layer a service so that it can be swapped out if someone really wants that [23:51:24] https://www.youtube.com/watch?v=Z_ucAP9tRB4 [23:51:44] just don't keep the non-mysql ones in core [23:54:53] killing PG support will also have the benefit of killing Citizendium (in 10 years, when they finally decide to upgrade from 1.24) [23:55:47] I have no problem with a DB extension existing [23:56:45] Just get them out of core where we pretend to care about them [23:59:56] MaxSem: Nope.