[06:38:00] Well, she said she knew someone, past-tense, so I'm assuming so. [07:36:15] nah lol [07:36:18] we just aren't as close anymore [08:36:15] Re image imports from fandom on https://issue-tracker.miraheze.org/T14081, what are we supposed to say/not say? [08:38:16] can always tell them that if they need help with those they can join here [08:38:28] [1/3] Also, is someone available to retrieve backtraces for https://issue-tracker.miraheze.org/T14100 ? [08:38:29] [2/3] The browsers used are not terribly outdated (about 3 years ago), so the UA different might have caused a server error. [08:38:29] [3/3] Using Windows 7 in 2025 is on a different level though. [08:40:47] w7 is still supported apparently lol [09:14:42] I'd be very surprised if a db connection error is caused by a browser UA [09:24:13] My suspicion is that different UAs trigger different code paths. Otherwise it would be very difficult to explain why this error only happens to certain mobile devices but not others. [09:32:02] It should only be on mostly front end stuff tbh [09:32:09] I can take a look at some point today [09:35:14] A DBQueryError shouldn't happen for invalid paths either [09:35:17] It should never happen [09:38:06] I meant to say that the UA triggers a different code path, which executes code not normally done in a request. That code could be malformed and trigger a DB error. [09:38:31] Well, at this point very little can be known about it anyway... [09:40:28] I think we're probably randomly guessing and putting 2 and 2 together and getting about 9 here [09:40:40] We have a request ID so we can look at the logs [10:32:29] Sounds familiar, I think this is what happens in the daily runninng of the govt [10:32:32] But alas I digress [10:52:33] Generous [11:17:32] @posix_memalign that error is utterly irrevelant [11:17:59] it's 8 days old and all it tells you is on that single period the db was down for a small while [11:18:13] The date on it is 8 days old, are we sure its from today? [11:18:23] if it is, it's a caching problem [11:19:52] damn, 8 days old cache [11:21:27] 5XX errors should be excluded from the edge cache [11:21:48] Probably a caching problem then. So now the question is why it only happens to some users on mobile. [11:23:04] ok that's dumb [11:23:11] somehow this is a 200 code [11:23:23] I would like to thank mediawiki for being awesome [11:24:09] [1/10] Request URL [11:24:09] [2/10] https://pgrfm.miraheze.org/w/load.php?lang=zh-cn&modules=ext.gadget.comtable%2Chantsect%2Chlist%2Cinfobox%2Cmbox%2Cnavbox%2Csite-styles&only=styles&skin=timeless [11:24:09] [3/10] Request Method [11:24:10] [4/10] GET [11:24:10] [5/10] Status Code [11:24:10] [6/10] 200 OK [11:24:11] [7/10] Remote Address [11:24:11] [8/10] 104.18.6.190:443 [11:24:11] [9/10] Referrer Policy [11:24:11] [10/10] strict-origin-when-cross-origin [11:24:23] @pskyechology can you please file an upstream bug that they are dumb [11:24:32] I vaguely remember reading about some people always using 200 in the return code and putting the actual error code in the response. [11:25:57] https://issue-tracker.miraheze.org/P558 [11:26:12] yes and they are dumb fuckers [11:28:24] i triggered a manual purge [11:28:45] and would like to express my disstatisfaction at whoever decided this was a good idea [11:30:03] i think this issue stems from the fact that they expected individual modules to possibly fail but didn't want everything else to die with it [11:31:24] they could at least set a cache header to say don't cache me [11:31:46] or add something to vary on [11:33:46] will figure out how to file an upstream issue when home (big girl words hard) [11:36:10] @blankeclair how many uwus and owos can I get away with before wmf phabricator tells me to piss off? [11:36:31] 40 [11:37:07] mmm remote address [11:38:01] What [11:38:18] you copied the remote address :p [11:38:27] oh wait, the port [11:38:30] nvm i was paranoid [11:38:50] do you wanna talk about it [11:39:04] i thought that the remote addr was the ip address of the requester :p [11:39:10] No [11:39:13] then i looked closer, then realised that the port is 443, not a > 10k [11:39:16] It's cf's ip [11:39:24] yeah, just did a whois to confirm ^^; [11:54:57] ResourceLoader the single worst thing about mediawiki [11:55:07] (Actually scratch that, Parsoid comes close) [12:36:55] I mean as a backend and desktop apps engineer ye [12:37:18] Rewrite it in C#! [12:37:20] I talk system administration, application backends and desktop app development [12:37:26] I don't talk front end web [12:37:37] Front end web is the one area of development that I don't do [12:37:47] ResourceLoader be backend dev tho no? [12:38:12] Some of it is but not all [12:38:28] It's all about delivering css + js in a way that works [12:39:02] I could deliver you a pipeline for delivering text but I'm not gonna care about css + js output [12:39:08] I wonder what came before ResourceLoader [12:39:16] You need front end skills for resource loader [12:39:17] I know Fandom had AssetManager but [12:39:41] Yeah I hate frontend too [12:39:49] Its for babies [12:39:59] (JOKE for any front enders reading) [12:42:00] I can't see any obvious what it replaced [12:42:07] It came out in 1.17 though [12:42:53] I wonder if you just had to link the scripts etc in the head manually [12:43:14] No idea [12:43:45] [[mw:JS2 Overview]] [12:43:46]

[12:44:05] @originalauthority ^ is the start of resource loader [12:45:03] Oh interesting [13:39:34] i think this behaviour is intentional and they dont intend on changing it https://phabricator.wikimedia.org/T87175 [13:40:33] i'll see if someone's complained about it still caching as well tho [14:18:29] Would be nice if they'd not set cache headers or set an X-Exception header or something we can choose to not cache on [19:56:32] [1/3] can we throw shellbox more memory so larger thumbnails could be generated? [19:56:32] [2/3] bunch of errors in graylog about it too [19:56:32] [3/3] https://issue-tracker.miraheze.org/T13500' [19:57:37] If you want, make a task about increasing memory for it and I'll review it later. We may end up needing to increase some memory on servers to so I'll have to see. [19:58:22] sounds like I'm assigning it to you [19:58:30] Feel free. [21:10:39] https://github.com/miraheze/MirahezeMagic/pull/520 is finally up-to-date lol. Should apply that task to it also perhaps or perhaps its not needed if we make certain things available anyway... [21:11:25] que [21:11:34] CI still failing though lol [21:12:03] quick, blame it on quibble [21:12:06] That adds the ability to configure certain things from RequestWiki [21:12:31] what about gerrit [21:12:53] I think I figured out a way to hack a temporary fix together for VE PortableInfoboxes [21:12:59] idk how this involves gerrit but i have no qualms with it [21:13:09] oh parsoid support? [21:13:23] if it works in VE it must support parsoid [21:13:37] Veg out [21:14:24] you're just mad because you can't follow simple instructions [21:14:43] (implication that you had issues with gerrit, thus you hate) [21:14:52] I’ve made a Gerrit patch before [21:15:50] [1/3] Not really, kind of a semi-support dependent on the legacy parser, which is how I think Fandom does it. [21:15:50] [2/3] The issue is that with Parsoid, you get the content of the template when the parser tag runs, but you don't get access to the parent frame so the parameters aren't passed. BUT... you can get the full text (the template) that was being transcluded at the time. My theory is that if I take this template, run it through the legacy parser through a FauxA [21:15:50] [3/3] pi request internally, and then Parsoid can be instructed to return that, which would make it display. [21:16:42] But its a really big hack and obviously dependent on the legacy parser. So I'm thinking to just try it, and if it works, put it behind a feature flag. it would fix the issue of PI's not displaying in VE until WMF fix Parsoid. [21:17:20] replace until with uf [21:17:29] [1/2] > it would fix the issue of PI's not displaying in VE until WMF fix Parsoid. [21:17:29] [2/2] good joke man [21:17:54] This is all I can think that Fandom could be doing so idk [21:18:26] interesting idea and could work I suppose but yeah it is a very big hack but not fully against it in the meantime. But it can't be used forever that is... [21:18:56] vendor a copy of mediawiki before the legacy parser is removed /j [21:19:05] Honestly there was a version of parsoid that PI worked in I think with MW 1.41 or 1.42 or something PI worked in VE [21:19:37] Reapply whatever they broke it and it works. Which Fandom could have done it. [21:19:39] I agree. my thinking is that it could work in the meantime, because in reality, either WMF fix Parsoid, or PortableInfoboxes go away anyway. [21:20:08] so portable you can put them away and forget about them forever! [21:20:12] they did promise to fix it, but that ticket is like 2 years old and no progress has been made on it [21:20:29] [[w:dysfunctional family]] [21:20:34]

[21:20:35] no bot? [21:20:38] yay bot [21:20:41] eepy bot [21:21:10] I have a way to permanently fix but its a hack 100x worse then what you are proposing. [21:21:25] oh? [21:22:18] Basically to copy a bunch of core files rebuilding a parser that PI can use and pass that to html using a bunch of raw stuff to build the PI in parsoid without going through the parsoid frame stack. [21:22:26] oh no [21:22:26] oh [21:22:38] so my idea, but more primitive [21:22:45] can you just rewrite parsoid instead [21:23:47] I did have a working version a while ago which I think is still on my old macbook. But that consisted of using the parseWikitextFragment api which is basically the same as what I'm proposing now only probably a lot quicker being on the php side rather than waiting for a http request [21:24:52] oh yeah here it is [21:25:37] very interesting [21:25:38] [1/2] https://github.com/Universal-Omega/PortableInfobox/commit/d16bbee7da0e8c10be22b3d065736501c486f358 [21:25:38] [2/2] was the code for it [21:26:27] it still work? [21:26:52] I shall try now [21:27:25] if it does feel free to PR it. At least we can have it working on VE for now [21:28:04] Yay passes now lol [21:30:03] [1/2] OH it does [21:30:03] [2/2] https://cdn.discordapp.com/attachments/1006789349498699827/1403128079668350996/Screenshot_2025-08-07_at_22.29.58.png?ex=68966c5b&is=68951adb&hm=06965e4d31127e88b97f4f66a8023068d715db7deac0a48bfefa1499451dc16f& [21:30:40] Although I wonder if I could do a bit more processing so that it displays in the right place rather than where it is now [21:30:52] becuase at the moment its knocking all the content dow [21:32:39] oh If i somehow hit it with the .portable-infobox class it fixes itself [21:33:00] [1/2] and editing works too [21:33:00] [2/2] https://cdn.discordapp.com/attachments/1006789349498699827/1403128821607305447/Screenshot_2025-08-07_at_22.32.54.png?ex=68966d0b&is=68951b8b&hm=77e7f0ccfab9ee0c188c262f1d291fc2a386d395a94ff6b9a4788ed1a951264d& [21:33:15] Nice! [21:35:40] Should I put this behind a feature flag you think, or just put it for everyone? [21:35:59] i'd vote everyone [21:36:41] If it reliably works I think we can just enable for everyone but if you think it should be a flag feel free just enabled by default maybe? [21:47:31] my core patch with 1k lines changed was just +2ed first try (it's on PS5 because I uploaded it while still working on it), gg https://gerrit.wikimedia.org/r/c/mediawiki/core/+/1176231 [21:48:51] huge W!!! nice! [21:50:44] you wrote it on a PS5? 😉 [21:52:19] absolutely [21:52:22] using a controller [21:53:20] i swear if one of you makes a mw core patch with a smart fridge... [21:53:34] I don't have a smart fridge, but my smart watch has a keyboard too [21:53:38] PLEASE NO [21:54:31] I have actually done a small emergency GH patch on a smart watch before. [21:54:49] Like 2 lines. [21:55:05] Which took about 20 minutes for those 2 lines. [21:55:06] 😭 [21:55:07] why [21:55:20] Honestly no idea. [21:55:53] I must have had my phone somehow otherwise it wouldn't have worked anyway. No idea why I did it like that... [21:55:55] still more respectable than having your emails end in "Sent from my iPhone" [21:56:19] (completely unnecessary shade at iphone users) [22:59:20] gpt5 just dropped [23:00:37] can't wait to use it for like 2 messages before having it fall back to gpt4 or whatever model they've used before [23:02:56] Yeah and in their live stream apparently 50 is greater than 47.4 [23:03:06] I mean less then lol [23:03:11] I reversed it [23:03:40] https://cdn.discordapp.com/attachments/1006789349498699827/1403151640302977155/gpt-5-announcement-megathread-v0-4x62vxisumhf1.png?ex=6896824c&is=689530cc&hm=4dbb34ccd61b14c5c7f4f5789032104f1126b05c1ed31874706d2f795e05124a& [23:04:09] Only GPT 5 exists for me now. [23:04:20] All other models are gone [23:05:22] [1/2] It doesnt even know its GPT 5 its apparently GPT 4.5 also [23:05:22] [2/2] https://cdn.discordapp.com/attachments/1006789349498699827/1403152066637070486/Screenshot_20250807_154455_ChatGPT.png?ex=689682b1&is=68953131&hm=78e7359b750ae93c16c06985e1fcb4132a6b36fb8a30b85839edff6e166eb1eb& [23:06:01] Not the best start to it [23:07:02] @originalauthority thanks for PR btw. Ping if you want a review/merge when finsihed/out of draft.