[16:18:53] bd808: Do we have a doc anywhere to tell someone "Here's how you set up MediaWiki to log stuff"? Re T138072#2433523. [16:18:54] T138072: persistent login session issue on some wikis - https://phabricator.wikimedia.org/T138072 [16:21:23] anomie: well... https://www.mediawiki.org/wiki/Manual:How_to_debug#Logging and https://www.mediawiki.org/wiki/Manual:Structured_logging [16:21:31] * bd808 reads bug [16:22:17] what do those channels look like? Do they need context data to be useful? [16:22:32] * bd808 assumes they probably do [16:23:12] What some IP posted at https://www.mediawiki.org/w/index.php?title=Topic:T75cloz7981b8i92&topic_showPostId=t75tzy17866kk4qu#flow-post-t75tzy17866kk4qu for a different third-party wiki seems to have sufficient detail. [16:24:35] ok. so if the context isn't needed then it's just $wgDebugLogGroups stuff that is needed [16:24:58] the stock logger doesn't output context information :/ [16:25:17] (because I couldn't figure out a nice way to do it) [16:26:20] The easiest thing to tell them would be just to set $wgDebugLogFile [16:27:00] that's the catch-all log destination if you haven't messed with $wgDebugLogGroups [16:30:42] Thanks [19:07:54] https://phabricator.wikimedia.org/T589 / https://phabricator.wikimedia.org/T28741 [19:08:14] Refactoring oldimage tables - starting with some agreement on how we're going to do it (implementation isn't resourced yet, worry about that later). [20:06:19] not sure who to ask... is it just me or the (-37) / (+25) change sizes on https://www.wikidata.org/wiki/Special:Contributions/Aude are no longer colored? [20:06:24] is that intended? [20:06:44] they are still colored on recent changes [20:08:27] not colored for me either aude but there are "mw-plusminus-pos" and "mw-plusminus-neg" css classes. Maybe the styles got removed? [20:08:33] oh [20:08:37] still colored on enwiki [20:08:42] * aude files a task [20:09:14] https://phabricator.wikimedia.org/T139522 [20:09:21] styles are defined in resources/src/mediawiki.special/mediawiki.special.changeslist.css [20:09:34] that's only recent changes / watchlist maybe [20:13:54] aude: https://gerrit.wikimedia.org/r/#/c/297644/ [20:14:35] oh [20:15:00] that seems to fix it for me locally [20:17:08] looking to see how it broke [20:17:47] aude: i broke it [20:17:51] commented on the task [20:18:07] thanks for spotting :) i'll have it backported [20:18:19] MatmaRex: ok [20:19:05] we fixed something! [20:19:10] my work here is done [20:19:19] :) [20:19:53] thanks :) [20:20:15] if that's the worst thing / only thing broken on wikidata then i'm happy :) [20:22:05] anomie: I fixed https://gerrit.wikimedia.org/r/#/c/297537/, any chance you could test that? [20:22:14] * anomie looks [20:22:25] my vagrant seems completely broken and fixing that is not a fun vacation project [20:22:36] (I did test the previous revision) [20:22:38] thx [20:38:27] tgr|away, anomie: thanks for picking up the globalrename stuff. Does that patch need to be deployed to both branches simultaneously? [20:39:28] legoktm: Hmm. It probably does, otherwise when the job runs on an old-version wiki the next wiki's job won't be scheduled/. [20:39:46] (good catch) [20:43:17] legoktm: Do you want to do the backporting? I could, except I don't have any way to test it in prod. [20:44:01] yeah, I can put it in tonight's SWAT window [20:44:05] Thanks [20:57:25] thcipriani: Re T139526, Icbc987fa4 looks suspicious. Specifically, the bit where DifferenceEngine::getDiffBody() no longer exits early when there's no old revision unless the new 'DifferenceEngineShowEmptyOldContent' returns false. [20:57:25] T139526: Catchable fatal error: Argument 1 passed to DifferenceEngine::generateContentDiffBody() must implement interface Content, null given - https://phabricator.wikimedia.org/T139526 [21:00:21] RFC meeting starting now in #wikimedia-office: Improve the per-programming-language listings for our tools [21:07:11] anomie: that is also the only difference in DifferenceEngine between wmf.8 and wmf.9 aside from the patch from this morning. [21:08:00] note that we have 2 different errors here: null and string being passed [21:08:29] null can come from SpecialUndelete, for example [21:08:50] also, we have mostly string in prod and null in labs [21:09:49] legoktm: Landed your $wgExternalDiffEngine cleanup [21:10:56] ostriches: thank you! do you think I could get that backported and included in 1.27.1? mostly because it'll let me kill more complexity in the debian package that has to add that one line of php config to localsettings [21:11:18] Yeah we can do that [21:11:29] * ostriches cherry picks [21:14:21] Ok +2'd too, it'll land in a bit [21:14:44] \o/ [21:21:30] ostriches: Speaking of 1.27.1, https://gerrit.wikimedia.org/r/297699 would be nice when dealing with people trying to figure out why their sessions aren't working. [21:22:13] Go ahead and put in on REL1_27 [21:22:20] (I don't have a timeline for a 1.27.1, but can def throw it in) [21:23:00] * anomie will once someone merges it to master [21:23:09] k [21:29:15] AaronSchulz: so… could you take a look at https://gerrit.wikimedia.org/r/#/c/295252/ today/tomorrow? it's a pretty simple patch :( [21:29:27] anomie, I'm reverting the diff thingie, agree with your conclusion [22:01:49] anomie: are you able to reproduce https://integration.wikimedia.org/ci/job/luasandbox-compile-php55/8/console ? I'm not able to locally... [22:02:36] (though I do have 3 failures when running `make test`) [22:05:52] oh, stupid me. I wasn't on the right branch [23:36:14] hi, casey from fr-tech here. we are trying to upgrade to .26 from .25 and have a regression due to some crufty js that depends on mw before document ready ("mw is not defined"). was this behavior understandably broken in that revision? just want to understand the nature of the breakage to decide the best fix. [23:37:16] cwd: yes. there's some function to wrap your code to make it wait for the JS globals to be set up [23:38:01] cwd: you should probably be using ResourceLoader::makeInlineScript( $js ) [23:38:48] MatmaRex_: thanks! i will mess with that.