[00:06:46] Do we still need PhpXmlBugTester? "Known fixed with PHP 5.2.9 + libxml2-2.7.3" [00:06:54] Or do we reckon that can go away with PHP7 bumps? [00:09:26] 18.04... with php 7.2 [00:09:31] libxml2 2.9.4+dfsg1-6.1ubuntu1 [16:11:22] Krinkle: After I547a7899e, "npm install" tells me "No compatible version found: wdio-mediawiki@'file:tests/selenium/wdio-mediawiki'". Any idea what's wrong? [16:30:07] anomie: Which repo? [16:30:24] I mean, in mediawiki-core itself? [16:31:25] Node/npm version? It should be fine in any non-EOL version of npm. The file: prefix was added back in npm 1.5 or 2.x I think. [16:31:33] Assuming that is the issue, not sure. [16:31:51] Did you try rm -rf node_modules and then 'npm install' again? To rule out local corruption. [16:33:25] Krinkle: mediawiki/core. node -v v8.11.1, npm -v 1.4.21 (i.e. https://packages.debian.org/sid/npm). Did the rm -rf, and `npm cache clear` too. [16:33:46] npm 1.4 and node 8.x contradict each other. How did that happen? [16:33:54] I assume node v8 did not come from sid. [16:34:10] well, I'll be damned, it does. [16:34:11] they doin't include npm in nodejs on the debian repo [16:34:20] they have a very old npm version though [16:34:24] through the npm deb [16:34:31] Right. Debian has excluded npm and then abandoned npm [16:34:34] npm install -g npm [16:34:52] yeh, you could use nodejs offical apt mirror though [16:34:59] which includes the latest npm version with it [16:35:26] anomie: Yeah, 1.4 is.. 2013? Definitely not supported, and various things "should" already not be working correctly with that. I'm surprised this is what tripped it up. [16:35:50] Krinkle: I did recently have weird errors about various packages not existing. [16:36:02] And before that I haven't had to try to check bananas recently. [16:36:15] 1.4 is also before they did ssl, I don't know if Debian made their own secure version, I'd certainly hope so. [16:37:50] but yeah, peerDependencies, file:, de-duplication, a $HOME/.npm cache that doesn't break when running npm from different directories at te same time, and others, a lot has changed in 5 years. [16:38:27] npm install -g npm might not actually work, but it's worth a short. [16:38:43] I'd recommend `npm install -g npm@5.6` in that case, which matches what Node bundled with Node 8 [16:38:46] I'll just rely on Jenkins to check the bananas for me, I suppose. [16:38:50] https://nodejs.org/en/download/releases/ [18:50:31] CindyCicaleseWMF: the charts on the mcr wiki seem broken [19:21:37] Reedy, James_F, legoktm: If I said something like "I'd like to put AFT V5 on wikitech as a feedback mechanism for documentation quality" would you yell at me, sigh deeply, or think that seemed like a workable idea? [19:22:28] bd808: I'd nominate Wikitech wiki for immediate closure. ;-) [19:22:57] bd808: It's… not a great tool. And it probably has some code standard issues by now. [19:23:04] that's one way to solve my doc quality problems ;) [19:23:23] Indeed. :-) [19:23:40] *nod* I haven't looked into the code at all, but the design page seemed magically what Sarah R and I have been talking about this week [19:24:03] You could do what Wikinews does (!) and auto-transclude the talk page at the bottom of every page with open threads from LQT (loathe as I am to encourage more use of that). [19:24:06] we went to a documentation conference and came out with grand ideas [19:24:53] I think I could hack up a gadget that gets us a basic "rate this page" metric [19:25:07] Yeah, something custom like that might be less bad, honestly. [19:26:24] all I need to do now is invent a day with more hours [19:31:03] anomie: the npm in sid is so old that it's unusable, which is why it got removed from stable/testing [19:32:17] bd808: AFTv5 conceptually is reasonable, but the implementation was severely flawed. A "leave feedback" button that had a textbox and put the comment on the talk page would be a lot more effective IMO [19:43:51] Krinkle: so, how do you want to move forward on https://gerrit.wikimedia.org/r/#/c/430644/ ? is just disabling them ok? [20:43:26] tgr: I moved them off the main page, since it was taking too long to load with all of the tabs of tables and charts. You can now find them linked at the top of the sidebar. They do need a bit of cleanup with the addition of the new, unlinked tasks, though. [20:53:54] ahh, sidebar [20:54:02] haven't thought of that [20:54:15] I only checked the links on the page [21:17:22] tgr: sorry - I meant to announce that at the meeting yesterday [22:02:12] legoktm: I don't know MMV code well enough to know whether or not the interaction being tested there is actually broken or not. [22:02:25] It triggering twice certainly seems less worse than not triggering [22:02:29] but for all I know could still cause bugs [23:01:20] Krinkle: I don't think it's broken, because I bisected it to your qunit upgrade. And no one has reported any bug against MMV being broken either... [23:02:06] I'm not saying it's visibly broken, but I don't doubt that assertion failure is a QUnit bug. [23:02:09] the test is broken. [23:02:42] Maybe it wasn't triggered before due to a lucky coincidence with internal QUnit behaviour [23:02:50] ah [23:03:08] I doubt that the failing assertion is a bug in QUnit. [23:03:33] It's a simple sinon mock. It them performs a click. result: The click handler got fired twice instead of once. [23:03:55] I'm sure that's real. Now maybe MMV just silently tears down and re-creates the model twice on each click, I don't know. I guess that works.