[01:02:21] I don't run the whole suite of phpunit tests very often because it takes a long time, and there are always a lot of spurious failures [01:02:40] PHP Warning: mkdir(): Read-only file system in /srv/mw/core/tests/phpunit/includes/GitInfoTest.php on line 8 [01:02:57] It's almost whack a mole running them locally [01:03:11] You do wonder how we seem to have relatively few problems with them on jenkins [01:03:24] obviously tests cannot write to my source tree [01:03:40] cuz we have several people babysitting our CI [01:03:56] Maybe we need a $this->skipIfReadOnlyFS(); [01:04:04] heh [01:04:28] but generally if you run an OS similar to CI (aka Debian) it should be pretty smooth [01:04:59] I find the same as Tim... Run it locally, and you get a load of random stuff possibly broken [01:05:11] in a few minutes I'll give you the summary, there's going to be a lot of failures [01:05:34] I quite often get failures caused by extensions, I have to disable the extension to get the test to work [01:09:04] that bodes well for support of other OSes... [01:09:10] my current test server is ubuntu 16.04 running under systemd-nspawn with the source tree bind-mounted into the container [01:11:44] Tests: 20617, Assertions: 113516, Errors: 1554, Failures: 90, Warnings: 1, Skipped: 137, Risky: 1. [01:13:35] Tests: 19673, Assertions: 129913, Errors: 29, Failures: 15, Skipped: 59, Risky: 13. [01:14:08] most of the errors are due to LabeledSectionTransclusion, not sure if it just needs a git pull [01:14:34] oh heh, I just disabled that extension because the xdebug output+deprecation warning was making my wiki unusable [01:16:21] I think about a quarter of these failures are legit, others are permission issues with my log directory, or "count(): Parameter must be an array or an object that implements Countable" [01:17:01] https://phabricator.wikimedia.org/T198173 bleh [01:20:41] the reason I'm looking at this is because of the test failure in https://gerrit.wikimedia.org/r/c/mediawiki/core/+/444803/ , wondering if I should have tried to detect that before I uploaded it [01:21:09] legoktm: jfding it? [01:22:23] so it's production logspam? probably worth doing [01:22:25] yeah, it looks pretty simple... [01:22:43] I'm only 4 things removed from the thing I wanted to do when I turned on my computer anyways [01:25:45] Yeah, we made deprecation warnings trigger php errors in CI as of late 1016 (if memory serves), and a month or so later, made Jenkins automatically fail the build if mw-error.log is non-empty, so as to avoid regressions with regards to any PHP Notice, or deprecation warning triggered during install/update.php and qunit/selenium browser tests, and PHPUnit also turns them into exceptions. [01:26:14] We had computers in 1016? [01:26:54] Have to poke at updating the rest of the extension that use the hook.... So we can finally kill it, 17 versions later [01:27:31] ok, now I see why no one else has bothered to try and fix this yet [01:28:07] 2016 :P [01:29:24] > [{exception_id}] {exception_url} ErrorException from line 19 of /srv/mediawiki/php-1.32.0-wmf.10/includes/tidy/RaggettInternalHHVM.php: PHP Notice: Undefined index: tidyConfigFile [01:29:28] Noticed this in the logs just now [01:29:48] 1-2 hours ago, various hits from https://*.wikipedia.org/wiki/Special:LintErrors/tidy-font-bug?* urls [01:30:25] Also from /w/index.php?title=Wikipedia:Village_pump_(technical)&action=parsermigration-edit ErrorException from line 19 of /srv/mediawiki/php-1.32.0-wmf.10/includes/tidy/RaggettInternalHHVM.php: PHP Notice: Undefined index: tidyConfigFile [01:30:40] Seems very rare, so must be an uncommon code path, but it does trigger. [01:40:47] hmm, there was a question on the change to remove tidy as to whether it would break ParserMigration, and someone said it wouldn't [01:41:38] we could just disable ParserMigration now, since we're fully migrated, but Linter probably needs to be fixed [01:42:41] [17:54:26] We should keep tidy around on servers for a month or two while wikis continue fixing their pages and templates .. and in case they need to see how it looked with Tidy .. ParserMigration needs to support it. [01:42:56] (from yesterday) [01:43:27] Linter should automatically detect that ParserMigration isn't installed and use normal edit links instead [02:17:15] LST patch: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/LabeledSectionTransclusion/+/445059 [03:26:06] TimStarling: Should I file a task for the tidyConfigFile error? [03:39:07] yes please [03:54:32] https://phabricator.wikimedia.org/T199293 [05:05:08] TimStarling, there are 25 instances of that notice since July 10. https://logstash.wikimedia.org/app/kibana#/dashboard/default?_g=h@8525ade&_a=h@d8f2b95 .. logstash opens up that for me within 5 secs or so. [05:09:29] I wrote a comment on https://phabricator.wikimedia.org/T189333 [05:10:28] I was mostly fiddling with firefox profiling since our meeting finished [05:12:38] this is a much better interface: https://logstash.wikimedia.org/elasticsearch/_search?q=tidyConfigFile+_type:mediawiki [05:12:53] especially now that FF has JSON pretty printing [05:43:03] anyone want to give https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/445070 a +1 before I merge it? [05:43:47] I've got a reproduction procedure, the error definitely happens every time action=parsermigration-edit is requested [05:44:25] the effect would be that the tidy config file is not used, which should produce quite badly broken output [05:49:38] I guess not [05:59:36] deployed and fix confirmed [16:15:56] DanielK_WMDE_: I've put up https://gerrit.wikimedia.org/r/#/c/mediawiki/core/+/445200 - needs some more debugging but you can see the shape of things [16:31:17] DanielK_WMDE_: are there any MCR SoS updates for this week? write-both is coming next week, right? [16:35:01] tgr: yes. but it's probably a good idea to point people to the deployment plan [16:35:06] the spreadsheet is public [16:35:19] the mroe people are aware of this, the better [16:53:27] tgr: i'm reviewing. lots of comments. but i think the direction is right! cool! [16:54:45] yeah, it's a lot saner than the old version. Thanks for the guidance :) [17:13:23] tgr: there is something wrong with the first patch in the chain, did you see the test failures? [17:13:44] yeah, I have a fix, not pushed yet [17:13:51] k [17:55:37] AaronSchulz: do we use 'storageLayout' => 'sha1', for images anywhere? I just tried to test it locally for https://phabricator.wikimedia.org/T199140#4416179 and it doesn't work [17:58:30] No [19:35:21] anomie: damn it. took quick with the rebasing that the other schema change patch landed in the meantime [19:38:02] <3 [21:57:04] mw.loader.using(['jquery.client']).then(function () { [21:57:04] if ( $.client.profile().name === 'msie' ) { [21:57:04] importScript( 'MediaWiki:Common.js/IEFixes.js' ); [21:57:04] } [21:57:04] }); [21:57:08] Why does ckbwiki have this? [21:58:09] i would hazard a guess that someone copy-pasted it seven years ago [21:59:07] https://ckb.wikipedia.org/wiki/%D9%85%DB%8C%D8%AF%DB%8C%D8%A7%D9%88%DB%8C%DA%A9%DB%8C:Common.js/IEFixes.js [21:59:10] Is it still needed? [21:59:23] Nemo removed a load of it a year ago... [21:59:46] enwiki deleted it 2014 [22:01:04] probably not, but you never know [22:01:27] if($.browser.msie) [22:06:54] ^ anyone any idea why that'd give a js "cannot read property 'mise' of undefined" [22:07:01] aka why is $.browser undefined? [22:07:04] https://ckb.wikipedia.org/wiki/%D9%85%DB%8C%D8%AF%DB%8C%D8%A7%D9%88%DB%8C%DA%A9%DB%8C:Gadget-ArabicLetterConvertor.js [22:07:18] Reedy: pretty sure we removed $.browser a couple years ago [22:07:29] i mean, one of the jquery updates removed it [22:07:32] lol [22:07:39] and instead there is $.client [22:07:44] That gadget hasn't been touched since 2013 [22:07:53] https://ckb.wikipedia.org/wiki/%D8%AA%D8%A7%DB%8C%D8%A8%DB%95%D8%AA:%DA%BE%DB%95%DA%B5%D8%A8%DA%98%D8%A7%D8%B1%D8%AF%DB%95%DA%A9%D8%A7%D9%86#mw-prefsection-gadgets [22:07:58] Unfortunately, I can't work out which it is [22:08:10] But it is a default gadget [22:08:16] try with ?uselang=qqx [22:09:26] https://ckb.wikipedia.org/wiki/%D8%AA%D8%A7%DB%8C%D8%A8%DB%95%D8%AA:%DA%BE%DB%95%DA%B5%D8%A8%DA%98%D8%A7%D8%B1%D8%AF%DB%95%DA%A9%D8%A7%D9%86?uselang=qqx#mw-prefsection-gadgets [22:09:29] It doesn't show on there :( [22:09:56] Is that because it's default? [22:10:12] somebody just removed it: https://ckb.wikipedia.org/w/index.php?title=میدیاویکی:Gadgets-definition&diff=578311&oldid=574302&diffmode=source [22:10:26] Yeah [22:10:29] I just told them to disable it [22:14:30] sweet [22:14:34] wikilove causes a db error [22:16:16] AaronSchulz: "Cannot execute query from Block::newLoad while transaction status is ERROR." [22:16:18] Any idea? :P [22:19:11] MatmaRex: https://api.jquery.com/jquery.browser/ jQuery 1.9 removed it [22:23:44] Reedy: So, removed from MW in 63c3650665313eede9d5f68c473bbc9c28e234d1 in May 2014. [22:23:49] heh [22:25:42] Interesting [22:25:48] ckbwiki got wikilove 1 year 1 week ago [22:25:54] They've only just apparently realised it's broken [22:25:56] (no db tables) [22:26:09] Clearly it's well-used. [22:26:31] Mmmm [22:31:02] Apparently it worked before [22:31:03] Meh [22:50:10] Maybe we dropped the DB as it had no writes for months and so was "cleary" unused? [22:53:50] Not unless T127219 deleted it by accident [22:53:51] T127219: Drop wikilove_image_log table from Wikimedia wikis - https://phabricator.wikimedia.org/T127219 [22:54:10] Not that ckbwiki is in the list [22:56:35] Reedy: any trace on that error? [22:56:51] It seems it was a referred error from wikilove_log missing [22:57:25] Reedy: Low priority, but if we're going to release a final branch, I made https://gerrit.wikimedia.org/r/c/mediawiki/core/+/445312