Policy Technical Proposals Idea lab WMF Miscellaneous 
The technical section of the village pump is used to discuss technical issues about Wikipedia. Bug reports and feature requests should be made in Phabricator (see how to report a bug). Bugs with security implications should be reported differently (see how to report security bugs).

If you want to report a JavaScript error, please follow this guideline. Questions about MediaWiki in general should be posted at the MediaWiki support desk. Discussions are automatically archived after remaining inactive for five days.

« Archives, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198


System problem?

Hey, tech folks and tech lovers,

It's Thursday and I think there might be a system problem but I don't see any messages here about it. It goes from minor things, like my edit count page not updating, to curious things, toolforge says that an AnomieBOT III report was issued at 21:56 UTC when no report has been issued since this morning, to Wikipedia:Database reports/Empty categories, which was just updated a few minutes ago, including a category that was deleted 15 hours ago.

Since AnomieBOT III wasn't working, I used a Quarry query to find broken redirects in article space and, yes, it did find one...one that existed early this morning and was later fixed a few hours later so it wasn't broken any longer! It's like the system is working with a version of Wikipedia that existed over 12 hours ago, not its current state with all of the changes that have happened since early this morning. Any ideas on what's up? Thanks for any insight your collective minds can offer. Liz Read! Talk! 01:17, 10 June 2022 (UTC)[reply]

Toolforge bots and tools like Quarry rely on copies of the Wikipedia database, occasionally the process that keeps them up to date breaks and they fall behind. According to the replag tool, it's currently 16 hours behind. I filed phab:T310325 to let the wiki replicas maintainers know about it. Legoktm (talk) 01:43, 10 June 2022 (UTC)[reply]
I love when people who know how this MediaWiki system works have immediate answers to my questions! Thank you, Legoktm. I guess the only part that surprises me is that it's just now being noticed. Thank goodness we have regularly scheduled bot reports, sometimes the only time I know something has gone wrong is when they don't operate as they always do so dependably. Liz Read! Talk! 02:16, 10 June 2022 (UTC)[reply]
Wow, there is now a 34 hour lag. How does the system ever catch up? Liz Read! Talk! 20:19, 10 June 2022 (UTC)[reply]
When it is working it can replicate faster than new data comes in on average. — xaosflux Talk 22:21, 10 June 2022 (UTC)[reply]
The latest message, linked from the phab ticket, says that it should be recovered by Monday. Enjoy the weekend, everyone! – Jonesey95 (talk) 05:01, 11 June 2022 (UTC)[reply]
Monday, huh? Those bot reports are going to be loooong when they finally get updated. That's 4 days worth of data on things that need fixing. Monday will be a busy day! Liz Read! Talk! 05:44, 11 June 2022 (UTC)[reply]
The latest info on that phab ticket is that Monday is the latest it would take to recover from this system lag, they expect some parts to be back to normal tomorrow, on Sunday. Liz Read! Talk! 01:12, 12 June 2022 (UTC)[reply]
According to WP:REPLAG - we are caught up. Happy Editing--IAmChaos 20:25, 12 June 2022 (UTC)[reply]
The lag maxed out at about 75 hours and then took multiple hours to recover, so services that depend on an up-to-date database were down for more than three days. Ideally, we will receive some advance warning when this will happen again, which it apparently will, since this outage was caused by normal (but infrequent) updates, not a system failure. – Jonesey95 (talk) 20:47, 12 June 2022 (UTC)[reply]

Important reports

@Liz and Jonesey95:: could you share which reports you find the most important to have updated on a regular basis? We should investigate moving those into MediaWiki special pages so they are supported at a production level. Legoktm (talk) 19:41, 11 June 2022 (UTC)[reply]

Here are a few that I look at daily, or more often:
Also see this thread. It occurs to me that converting reports to Special pages may not be desirable, since Special pages are not typically watchable or customizable and have no talk pages or history to generate diffs from. For example, looking at a daily diff of Wikipedia:Database reports/Unused templates/1 is very useful, since it shows any new entries in a straightforward way. – Jonesey95 (talk) 20:35, 11 June 2022 (UTC)[reply]
Oh, there are so many. It is easier to name the bots whose work I rely on. They include AnomieBOT III, SDZeroBot, BernsteinBot, DumbBot, JJMC89 bot, these are the main ones, they all issue reports daily or throughout the day, that include pages that need tending to. I can track down the exact pages if you need those. So far, SDZeroBot's and DumbBot's reports don't seem to be as affected maybe because they don't rely as much on up-to-date information database information since PRODs were tagged 7 days ago and expiring drafts can be predicted far ahead. Other bots that report broken redirects or empty categories rely on edits that have happened recently, over the past 24 hours. I also rely on Special Page reports but those, so far, have been unaffected by this system lag.
The phab ticket has some updated information on when this problem will be resolved but I think I saw you post there, Legoktm, so you've probably seen it. I think the most frustrating aspect of this is seeing the tech folks say that this lag was "unavoidable" which implied that it was expected. There used to be notices I'd see when Wikipedia was going to be worked on and would go into read-only mode, so they could have post some kind of notification, if not to the general pool of registered editors than here at WP:VPT or on the bot noticeboard. I think it's an issue of managing expectations and when some aspect of Wikipedia doesn't work as it reliably has for days, weeks or months, it can cause confusion and alarm. Knowing that the bots will be down for 3 or 4 days would eliminate much of this frustration. As Jonesey95 and you state, they also implied that some bot reports should be moved over to Special Pages which I guess is on MediaWiki but I don't know how realistic or feasible that suggestion is. The special page talk pages are unwatched and the ones I rely upon about categories can only be updated every 3 days which makes them less useful than bot reports which are typically more frequent. You can also come to the bot operator when there are problems and, in my experience, they have been very responsive. Liz Read! Talk! 00:57, 12 June 2022 (UTC)[reply]
If you don't mind a dumb question, what is this distinction that the developers make on the phab ticket between a task or job being "production" or it not being production? Does this define what problems they work on and ones that are someone else's responsibility? Liz Read! Talk! 01:05, 12 June 2022 (UTC)[reply]
Good question, it's really jargon these days. Broadly speaking, Wikimedia infrastructure are divided into two main categories, "production", which consists of the servers, databases, and associated services that run all the wikis (anything on the primary wikipedia/wikimedia/etc. domains), and "cloud", which consists of Toolforge, Quarry, Beta Cluster, and more (usually on wmflabs/wmcloud/toolforge domains). production is ultimately maintained by the Wikimedia SRE team (~45 people), who are on-call 24/7 in case of outages/emergencies. In contrast, cloud is maintained by the Cloud Services team (5 people), which doesn't have amount of staffing for on-call. (I haven't been on the Cloud team since late 2020, so I don't have a good sense of that the current threshold for outages that would page people on a weekend/late night).
Complicating matters is that the wiki replica databases straddle the production/cloud boundary (the sanitization step occurs in production, while the views that allow tools to make queries are maintained on the cloud side), leading to shared responsibility, which both is helpful but can also make it more difficult in diagnosing what is wrong and who is responsible for that part (when I was trying to escalate the issue on IRC I pinged the wrong group of people initially). I hope that answers your question on what "production" is.
In this case, the lag is caused by a schema change on the production databases that is taking quite a while to apply to the wiki replica database. Up until recently, schema changes were a really manual and rare process. I think we're seeing some of the growing pains of being able to do schema changes faster - and I expect that we'll get better at notifying editors or avoiding the lag entirely over time. Legoktm (talk) 04:19, 12 June 2022 (UTC)[reply]
Wow, that is really helpful, Legoktm. You are clearly very familiar with the backend of the system we all use. I appreciate the explanation and how some tasks might straddle the line between work groups. Do you expect that schema changes will happen more frequently and we will run into these lags more often? Again, it's about expectations, in reality, I think we all appreciate improvements to the system even if they take some time to do. Liz Read! Talk! 04:37, 12 June 2022 (UTC)[reply]
I'll defer to what Ladsgroup said at phab:T310325#7996931. The impression I get is that we have a large backlog of database cleanup, so in the short term we might have some more lag incidents but my hope is that this isn't really an issue long term.
P.S. back in the Toolserver era we had to deal with replag on the scale of months. I think it's really great that we are now complaining about a few days of replag, it shows how far we've come (and that we still have room to improve!). Legoktm (talk) 05:39, 13 June 2022 (UTC)[reply]
Thanks both for the input. In my opinion, things that are essential to the operation of the wiki should be part of MediaWiki so they get that "production" level quality of support, both from SREs as explained above, but also from the shared maintenance MediaWiki core gets from all the various devs that work on it versus the one or two botops that operate current reports. MW has a pretty robust system of special pages that can effectively serve as database reports, but as always need more love. I've started working on existing issues like T307314 and T310456 before looking into implementing more reports as special pages.
I am intrigued by the idea of highlighting new entries somehow, I agree that makes the special pages somewhat inferior to wiki pages. Legoktm (talk) 05:56, 13 June 2022 (UTC)[reply]
Moving to special pages might have advantages but could cause problems. As one small use case, I follow WP:Database reports/Linked misspellings (many are typos even when not wikilinked) but deal only with its new entries. I want a diff with the previous day, and I want to get that in arrears after a wikibreak. With a report, that comes free with the page history; with a special page it would be difficult. Enhancing or fixing a report can be done locally; modifying a special page would have to join a long queue of tasks awaiting scarce developer time. Certes (talk) 11:56, 13 June 2022 (UTC)[reply]
I started jotting down some notes on wiki pages vs query special pages at mw:User:Legoktm/wiki page or query page - feel free to add. (I'm intentionally skipping over the "who has access to modify/fix things" problem). Legoktm (talk) 23:30, 13 June 2022 (UTC)[reply]

PetScan missing results

I'm trying to list pages in Category:Music memes that are not within the Category:Internet memes hierarchy, e.g. its sub-cats by country/year. https://petscan.wmflabs.org/?psid=22260983 says there are 202 not within 1st-level sub-cats, starting with "We Didn't Start the Fire". But https://petscan.wmflabs.org/?psid=22260984 says there are none that are not within 2nd-level sub-cats. Well, that song is not in any other meme categories, so PetScan seems to be malfunctioning. – Fayenatic London 08:53, 11 June 2022 (UTC)[reply]

Are contributions from the last two days going to be eventually counted on XTools?

Not sure if this is the right place to ask, but currently, if you look at an edit counter like mine, there is a notice at the top that says "Caution: Replication lag is high, changes newer than 2 days may not be shown." What does this mean, and will the changes from the last 2 days be eventually counted when the lag ends? weeklyd3 (message me | my contributions) 02:15, 12 June 2022 (UTC)[reply]

I can confirm the lag. Before posting this section, I had 1145 edits according to XTools, and that number still hasn't changed. So yes, there is replication lag. weeklyd3 (message me | my contributions) 02:16, 12 June 2022 (UTC)[reply]

User not found in Xtools

Bluefuego (talk · contribs)

I've noticed that this user has contributions and a live talk page and does not generate the nonexistent user message, but the Xtools tools fail to find this user. They have 3 edits to date, all of which regarded a BLP draft that has been speedy deleted per G10. –LaundryPizza03 (d) 11:51, 12 June 2022 (UTC)[reply]

All of that happened yesterday, more recently than the lagged database is up to. Anomie 12:18, 12 June 2022 (UTC)[reply]

wikipedia down

It was completely down 5 minutes ago, I confirmed with isitdown. Now it is running very slow as of 08:31, 12 June 2022 (UTC) Anybody else experiencing this? —usernamekiran (talk) 08:31, 12 June 2022 (UTC)[reply]

wikimediastatus.net reports an outtage at 8:14 which was resolved at 8:44 today. That page is an user readable website of the status of WMF wikis. Can not find an phab ticket behind it, might be hidden.--Snævar (talk) 09:07, 12 June 2022 (UTC)[reply]
T310431 is one. There might be cross-links to others later — GhostInTheMachine talk to me 11:24, 12 June 2022 (UTC)[reply]

upstream connect error, 12 June 2022

I've had a recurrence of phab:T301505, it lasted for 15 mins or more. Sometimes "upstream connect error", sometimes a HTTP 503 Service Unavailable. This time, the problem also affected Phabricator itself. --Redrose64 🌹 (talk) 08:33, 12 June 2022 (UTC)[reply]

I had this but it seems ok now.--♦IanMacM♦ (talk to me) 08:42, 12 June 2022 (UTC)[reply]
Well, yes, it must be OK now since we can post here. The thing that I posted is an after-the-event report, but also noting that phab was also unavailable, something that hasn't happened before. --Redrose64 🌹 (talk) 08:47, 12 June 2022 (UTC)[reply]
I had limited access to the site in "Desktop mode" on mobile (Chrome/Android), though I didn't check it until a minute or so before the site came back, so I don't know if it was available through the entire outage. – Philosopher Let us reason together. 08:46, 12 June 2022 (UTC)[reply]
I was on desktop mode on Android too. I faced this issue too. The last Wikipedia page to open correctly was at 08:11 per browser history and the next opened at around 08:27 UTC. In the meantime there were errors. This is a different timeframe from what Snævar notes above. In fact, I registered an edit within the outtage period per wikimediastatus.net. CX Zoom[he/him] (let's talk • {CX}) 09:12, 12 June 2022 (UTC)[reply]
Just got this message again, for about 4 of 5 minutes, right now. "Overflow" and no loading of pages. Liz Read! Talk! 03:56, 13 June 2022 (UTC)[reply]

Dead interwiki links

Hello from german wikipedia! For the german wikipedia I wrote some script to identify dead interwiki links, now after 2 month almost all those "redlinks" are gone. So maybe the english wikipedia (or others) are interested in such a list. Currently I find 3,706,692 interwiki links to 401 wikis and there are 129,018 "redlinks" in 39,420 articles (I did not check namespace draft). So 3.5 % of those interwiki links are not working, in german wikipedia we had about 3 % such dead links, so there is no big difference. If you are interested, I can create such a list. --Wurgl (talk) 09:11, 12 June 2022 (UTC)[reply]

I find a surprising amount of hostility on en-wiki to inter-languguage links. I do not understand why, machine translation is so easy and the imperfections are fairly minor. Of course, you must apply some common sense in how you use them. But most of the time, it's more than adequate.
The most mundane approach is just to use an inter-wiki link (e.g.. [[:language-code:articlee_name]]).
In preference to this is ((ill|article_name|language-code...)). This is generally acceptable, but raises a lot of discussion as to the most languages you should ever specify. (I would peronally not go crazy about this, but it seems to generate more heat than light.)
However, I think that is greatly inferior to speicfying the wikidata parameter with ((ill)). This is the most maintainable approach. People seem to complain about the Wikidata interface, and I really have no idea why. I don't doubt the Wikidata interface can be improved, but it's just not that onerous. Of course, it's confusing until you understand what's going on, people who care will figure it out, there are many stumbling blocks in using Wikipedia.
Anyway, I have no understanding of WP auomation or scripting, so I couldn't use it, but it sounds like a great offer and something that somebody in en-wiki should be taking you up on. Fabrickator (talk) 20:26, 12 June 2022 (UTC)[reply]
@Wurgl, if it's bot-related, you might want to post at WP:BOTN. ― Qwerfjkltalk 20:47, 12 June 2022 (UTC)[reply]
The german list: It is very simple de:Benutzer:Wurgl/Interwiki Rotlinks and yes, it is bot-related, but I think I do not need a bot-flag. --Wurgl (talk) — Preceding undated comment added 21:12, 12 June 2022 (UTC)[reply]
(not directly related, but I will ask anyway) What about interlanguage links that link to redirects, and that redirect goes to a page that is present in the same Wikidata item as the original article (where the interlanguage link came from)?--Snævar (talk) 07:39, 13 June 2022 (UTC)[reply]
getting headache … but somehow I like that idea, it's a kind of challenge --Wurgl (talk) 08:31, 13 June 2022 (UTC)[reply]

Good Article Counter missed one (missing 2)

The Good Article counter is not counting my GA for Lambert Gas and Gasoline Engine Company. It gives me 216 GAs, however I have really done 217 GAs. Can it be fixed so it gives me the correct count, by counting everything. This may be a clue Thanks.--Doug Coldwell (talk) 21:39, 12 June 2022 (UTC)[reply]

Perhaps this is related to #System problem? above. NguoiDungKhongDinhDanh 22:23, 12 June 2022 (UTC)[reply]
@Doug Coldwell You mean this GA counter? This appears strange, as this tool doesn't depend on the toolforge databases that were lagged. Will look into it ... – SD0001 (talk) 10:10, 14 June 2022 (UTC)[reply]

Display of numbers goes wrong in this list in mobile view only

Please see this mobile version of List of vehicle speed records. Evenon a wide screen, it geos terribly wrong in some cases (especially in the section "Rail vehicles"), but on a smaller mobile screen, nearly all the numbers get a confused jumble across multipe lines, making it impossible to correctly read this. I have no idea what specific element causes this (perhaps the Template:decimal-align?), I guess the people at VPT are more qualified to answer this. That mobile editing has numerous serious issues is well-known, but mobile viewing should work... Fram (talk) 08:35, 13 June 2022 (UTC)[reply]

@Fram does the first table there look better for you in this revision? The page authors have applied some very specific width allocations on those tables and their columns - in the revision I linked above I let the first table just fill naturally without that specification (left the other tables for comparison). — xaosflux Talk 09:47, 13 June 2022 (UTC)[reply]
Thanks, but the improvement is minor. On a widescreen, it is better, but when reducing the width e.g. the first cell still reads "1,2.98" (first line) "275" (second line)" instead of "1,227.985" (or e.g. "1,227" (first line) ".985" (second line)). For some reason he is line-wrapping the number in this very strange way: while the wrap itself may be caused by the widths, the wrapping in two parts should never happen and seems to be due to either a local template (convert or decimal-align) or a more global technical issue. Fram (talk) 09:54, 13 June 2022 (UTC)[reply]
((decimal-align)) is most likely the culprit. It splits the number into two floats, and each float would wrap independently. Anomie 11:36, 13 June 2022 (UTC)[reply]
Yup. There is a proposal to do this aligning properly in CSS, based on the character, but that's not supported in any browser yet. The best way for now to do this is simply have an equal amount of numbers after the decimal point and right align. These kinds of hacks that decimal-align applies are really shaky. —TheDJ (talkcontribs) 12:55, 13 June 2022 (UTC)[reply]
Okay, thanks! Fram (talk) 13:25, 13 June 2022 (UTC)[reply]
May need both of these to fix it for more screens, add trailing zeros to make the decimals the same length, and get rid of the fixed width allocations. — xaosflux Talk 13:40, 13 June 2022 (UTC)[reply]
If you don't want trailing zeros over concern about implying more significant figures than are present, you could try using figure spaces instead. Anomie 11:29, 14 June 2022 (UTC)[reply]

logged out automatically

Hello. A few minutes ago, I realised I was logged out of my this (usernamekiran) account automatically. I login to Wikipedia/media/tech using four devices. One computer, and three mobiles. All of them are password protected, and only I have access to them. In case somebody wants to use my computer for any purpose, or mobile browser, I always log out. Logging out from on device logs me out from everywhere. For using other accounts, I always use private browsing, and it doesnt have any effect on my main window's login session. It certainly hasn't been 365 days since my last login (I had logged out recently). So it was surprising when I found out I was logged out of the account. And it is not even Thursday. I have a super-duper strong password, which I changed a few minutes ago. I also checked my global contribs, no suspicious activity found. Has this happened with someone else as well? —usernamekiran (talk) 08:57, 13 June 2022 (UTC)[reply]

@Usernamekiran: If you are logged in on two or more devices, logging out of any one of them will always log you out on all of the others, regardless of whether it's a private session or not. It's because the action of logging out sends a message to all servers to invalidate all login cookies with your name on. This is not new, it's been the case for at least as long as I've been around (13 years). --Redrose64 🌹 (talk) 09:08, 13 June 2022 (UTC)[reply]
@Redrose64: yes, I am aware of that. I don't click on "log out" during the private session. I directly close the window, that doesn't log me out from the other sessions. While editing English, and Marathi wikipedia, I regularly do the account switching/hopping from private window for KiranBOT, and KiranBOT II; it doesn't log me out from main window's "usernamekiran" account. —usernamekiran (talk) 09:36, 13 June 2022 (UTC)[reply]
Yes, I've had issues too. Ignore and carry on. I suspect that sometimes when someone's doing some IP 'forensic' work that it is possible to get knocked off. Neils51 (talk) 11:49, 13 June 2022 (UTC)[reply]

Template:District line simple RDT

Can someone please fix the "Paddington" link in Template:District line simple RDT? The correct target is Paddington tube station (Bakerloo, Circle and District lines). Animal lover |666| 11:35, 13 June 2022 (UTC)[reply]

 Done Certes (talk) 12:06, 13 June 2022 (UTC)[reply]
@Animal lover 666: My edit has been reverted, so perhaps you can find consensus for the best link target on the template's talk page. Certes (talk) 21:09, 13 June 2022 (UTC)[reply]
I have no doubt that I'm right here, but that's between me and the reverting user. Thanks for showing me how to do it and for informing me of the revert. Animal lover |666| 06:21, 14 June 2022 (UTC)[reply]

Warning when trying to update a file which prompts that your updated file cannot be accepted.

Upload warning, SVG does not match
Upload warning, SVG does not match

I'd like help with an error that pops up every time I try to update files, it doesn't matter if I match the size of the new image with the old one or the file type.. it will simply give you warnings over and over again, making it impossible to progress, it shouldn't be this hard to update files when you use the correct file type which is shown. Hogyncymru (talk) 15:27, 13 June 2022 (UTC)[reply]

This error suggests you are trying to upload a file with a name that ends in ".svg", but where the file content is actually not actually a Scalable Vector Graphics file. Is that actually wrong? — xaosflux Talk 15:29, 13 June 2022 (UTC)[reply]
Even if it is an oversight, why is it so sensitive?, I shouldn't have to figure out why my file isn't accepted. I'll give you a test if you'd like, it's set up here , if you download the file in question and use that to update the old file see what happens (the file needs updating anyway as it's inaccurate)Hogyncymru (talk) 15:37, 13 June 2022 (UTC)[reply]
So did you download an SVG file, edit it in to a PNG formatted file, then try to upload it to replace the old SVG file? If so, after editing you should render it as SVG first. There are many benefits of different file types that those wanting to use or reuse files may consider, so having the name reflect the content is generally useful. commons:Help:Converting#Converting_images and commons:Help:SVG#Converting_to_SVG have more information on this as well. — xaosflux Talk 15:44, 13 June 2022 (UTC)[reply]
"It's so sensitive" precisely because editors who don't understand what they're doing can ask editors who do and actually receive help with the issue.
I assume what you did was that you downloaded one of the rendered .svg files which are rendered to PNG and saved it without changing the file name. And/or you made edits to the PNG subsequently. You are now trying to update the onwiki version, but because you have a PNG and not an SVG, the system is correctly warning you that you have a PNG. To correct the issue, you will need to get the original SVG and then edit that. Izno (talk) 15:52, 13 June 2022 (UTC)[reply]
Thanks for the information, it does help me understand better, but why does it matter if it changes file type? I'm a little confused, why can't wiki deal with the problem Iv'e ended up with? is there a desire to update it to make it easier to change the file no matter the file type?Hogyncymru (talk) 16:22, 13 June 2022 (UTC)[reply]
why can't wiki deal with the problem Iv'e ended up with Software is not magical. Given the problem you've encountered, no software can fix the issue (at least to the fidelity of the original SVG). This won't be changed in the software. Izno (talk) 16:52, 13 June 2022 (UTC)[reply]
And the change you are trying to do will replace a relatively high-quality image with a lower quality one (svg-->png). You certainly can fork the image and upload your PNG file, then change pages to use your new image - but replacing it with a high quality SVG would be better for most everyone else. — xaosflux Talk 17:27, 13 June 2022 (UTC)[reply]
@Hogyncymru There are two main ways of storing images. The first is called raster graphics, you make a list of every pixel in the image and you note down what colour they are, a hypothetical raster graphics file might look something like "row 1: pixel 1 = red, pixel 2 = light red, pixel 3 = pink ...". The second is called vector graphics, you store the image as a list of instructions that are required to draw it, a hypothetical vector file might look like "start with a green background 10 by 10 cm, draw a red line from (2,3) to (5,6) ...". Vector files have a number of advantages over raster ones, the main one being that you can zoom in indefinitely and the image will never get blurry because it can keep on rendering more detail. SVG is a vector file format. PNG is a raster file format. "wiki" cannot automatically change the file format here because it is very very difficult (bordering on impossible) for a computer to take a finished image and figure out a sensible set of brushstrokes it would need to reproduce it - software isn't magic. Even for cases where is is possible to do an automatic conversion (like converting between different types of raster graphics) it has the potential to be problematic and may involve making editorial decisions that affect how the image looks - in the conversion process you can loose detail and create artefacts, you can mess up colour scales and brightness, you can mess up transparent layers, you can end up needing to pad files with borders or trim off the edges, etc. If you want to edit an image the best thing to do is to keep it in the same file format throughout. 192.76.8.94 (talk) 20:17, 13 June 2022 (UTC)[reply]

Tech News: 2022-24

16:57, 13 June 2022 (UTC)

Outdated data-maps. From 2021. In COVID-19 pandemic in Tennessee infobox...is it ok to delete them?

Yeah, I'll post an RfC at the article if required but...
I don't know how to construct these particular files/maps so have posted elsewhere around Wikipedia asking for help - Wikipedia:Graphics Lab/Map workshop, Talk:COVID-19 pandemic in Tennessee, Wikipedia talk:WikiProject COVID-19, a user talk page - trying to get them updated but have been unsuccessful so far. The information on these data-maps dates from February 2021 and March 2021...I have found sources to update the info but graphics-coding is not my strong suit. I am at the point that I think the maps should be excised from the infobox and am wondering what the Village Pump:Technical community thinks.
If anyone around here can update the 4 maps, YAY! Here are the sources I have found. Some of the initial pandemic data sources have gone dark so getting at June 2022 info isn't as easy as it might have been early on.
These are the 4 2021 data-maps, they are hosted on Commons:

The most recent information can be found at the following sources:
At this time, the State of TN stats seem to be updated weekly, BUT if possible check & confirm with other sources before using.

Thanks, Shearonink (talk) 19:56, 14 June 2022 (UTC)[reply]

The right place to ask is either the talk page of the article or WT:COVID-19. Izno (talk) 20:46, 14 June 2022 (UTC)[reply]
Right place... Well, bless your heart. Did you see that I had mentioned above I had already asked at both those places? (Plus some other WP-venues before I posted here?) Virtually no replies. Oh well, at this point I'm going to just delete the maps - they are so far out of date they are useless. But thanks - Shearonink (talk) 21:49, 14 June 2022 (UTC)[reply]
Honestly, Wikipedia editors seem to have lost interest in regularly maintaining COVID-19 stats. If no one updates them, be bold and remove them from the infobox. A much better position for them will be the February 2021 section in te timeline. CX Zoom[he/him] (let's talk • {CX}) 22:04, 14 June 2022 (UTC)[reply]
Yeah, I know the data is overwhelming and we are all overwhelmed - I get that, there's a g-d- pandemic going on. I just wish I could get someone to do the updated maps to sometime this year.... I'll remove them sometime next week when I can sit down and figure out where to stash them in the timeline of the article. Thanks for your input CX Zoom, that's a good idea. I do hate to lose visible good work off of an article. Shearonink (talk) 23:31, 14 June 2022 (UTC)[reply]

Wrong message when restoring an old version of a page.

For some reason the text of MediaWiki:Postedit-confirmation-restored has been changed from "The page has been restored." to "The page has been undeleted.". When you do a manual revert on a page you therefore get a message about the page being undeleted, rather than being restored to an old version. Is this an intentional change or a bug? I don't think "undeleted" makes much sense in this context. Could someone override the new message locally until this is resolved? 192.76.8.94 (talk) 11:56, 15 June 2022 (UTC)[reply]

This was changed in gerrit:plugins/gitiles/mediawiki/core/+/6abc28230b547cacaffe93df5b54fbd56a45ce99. There are other messages that where changed from "restored" to "undeleted". Snævar (talk) 12:37, 15 June 2022 (UTC)[reply]
phab:T307181. — xaosflux Talk 13:20, 15 June 2022 (UTC)[reply]
192: can you be more specific, I haven't run in to this yet. What do you mean when you say "a manual revert"? When I go to a page history, edit an old version, and publish it ("a manual revert", such as in testwiki:Special:Diff/514665) I'm not seeing this message. What are the exact steps you are doing that is leading you to this? — xaosflux Talk 13:22, 15 June 2022 (UTC)[reply]
@Xaosflux This morning I did a manual revert on a file description page. I opened the version of the page I wanted to restore, Entered an edit summary about reverting sock edits, then saved the old version of the page without changing anything. The green "edit successful" popup was displayed, but instead of saying "The page has been restored." it said "The page has been undeleted.", which doesn't make sense in this context. 163.1.15.238 (talk) 13:48, 15 June 2022 (UTC)[reply]
Thank you! Bug phab:T310708 opened. — xaosflux Talk 14:10, 15 June 2022 (UTC)[reply]
@Xaosflux Thank you for opening a phab ticket. Sorry that my original message was unclear, I'll give more details when reporting problems in the future. 163.1.15.238 (talk) 15:03, 15 June 2022 (UTC)[reply]
No worries, it may have been more obvious to someone else, that you were watching and followed up was most helpful! The dev's seem to agree this is in error and are looking at an upstream fix. — xaosflux Talk 15:06, 15 June 2022 (UTC)[reply]

Malformed entries derived from preloads - is there a better way to do this?

Sometimes, a link opens an edit page where users can fill in content. Sometimes, the new section already has some text in it. Many times, I've seen users write their request/contested speedy deletion/something else under the signature inserted by the preloaded text.

For example:

== Contested speedy deletion ==
This page should not be speedily deleted, because (your reason here) [[User:Example|Example]] ([[User talk:Example|talk]]) 19:08, 15 June 2022 (UTC)

(a reason why the page should not be speedily deleted)

or

 == Edit request ==
 ((request edit))
 [[User:Example|Example]] ([[User talk:Example|talk]]) 19:08, 15 June 2022 (UTC)

 (description of the edit request)

Is there a way for preloads to use the DiscussionTools, which automatically inserts signatures, instead of the edit screen? So that when users write their request at the bottom, it is autosigned.

I'm talking about something like this:

  1. User clicks 'Request edit' button
  2. 'New section' form appears with a preload like this:
((request edit))

<!-- State UNAMBIGUOUSLY your suggested changes below this line, preferably in a "change X to Y" format. Other editors need to know what to add or remove. Blank edit requests will be declined. -->
  1. User fills in edit request below comment and saves the page.
  2. DiscussionTools fills in the signature automatically.

weeklyd3 (message me | my contributions) 19:08, 15 June 2022 (UTC)[reply]

To demonstrate, I found a talk page in CAT:CSD (probably will be deleted soon) that demonstrates the problem I'm talking about: Talk:Leba Chand Tudu. It says 'your reason here' and a signature followed by the user's comment. weeklyd3 (message me | my contributions) 19:10, 15 June 2022 (UTC)[reply]
We've considered this when working on the DiscussionTools new topic tool, you can read the discussions at T269310. (I'm one of the developers.)
It would be possible to add, and we had a mostly working implementation, but we've discovered that this would probably have to be enabled individually for each use of preload, because the existing ones wouldn't always work well with it (there are some that don't expect a signature at the bottom or a heading at the top, and some that don't work well in visual mode – particularly those using <!-- --> comments and subst: templates).
At the time we dropped this work, on the assumption that wiki users probably wouldn't want to do the work to adjust the forms and opt them in, but considering the wider adoption of the tool (e.g. Wikipedia:Village pump (proposals)#Enabling the New Topic Tool by default) maybe we should revisit that. (CC @PPelberg (WMF)) Matma Rex talk 20:06, 15 June 2022 (UTC)[reply]
@Matma Rex It would definitely be nice to have this as an option to opt into using the new topic tool with preloads, there would just need to be a way to specify a fallback. --Ahecht (TALK
PAGE
) 13:10, 16 June 2022 (UTC)[reply]

ReFill down?

Every time I try and use it just says 'Pending. Waiting for an available worker'... GiantSnowman 20:39, 15 June 2022 (UTC)[reply]

I have created a task to cover this. Keith D (talk) 22:25, 15 June 2022 (UTC)[reply]
@GiantSnowman: should be working again now. Keith D (talk) 23:12, 15 June 2022 (UTC)[reply]
Thanks! GiantSnowman 06:11, 17 June 2022 (UTC)[reply]

Query about categories

You guys are so helpful that maybe you can answer this one for me. I occasionally check Category:Wikipedia non-empty soft redirected categories.

If you check it right now, you'll see what I've seen in the past, a category for non-empty categories that is actually full of empty redirect categories! In the past, I have made a minor edit to the redirect category, like adding a blank line or space, and that has removed it from this category. I don't know why basically a null edit would do this but it does. But I really don't want to spend time tonight making pointless edits to 59 redirect categories so I came here looking for a longer term solution. Do you know why some redirect categories would randomly show up in an incorrect category when they, in fact, are empty? Thanks for any help you can provide. Liz Read! Talk! 02:28, 16 June 2022 (UTC)[reply]

@Liz The English wikipedia is too big and gets too many page views to update things in real time. Rendering a page (i.g. turning the wikitext into HTML that can be sent to readers and calculating all the metadata about the results) can take a significant amount of time, 20-30 seconds or more for big or complex pages, so to save computing time the results are calculated and stored in a cache for later use. Mediawiki uses things called "links tables" to keep a record of all the links in a page and what types they are, which are used to fill up categories and pages like Special:WhatLinksHere. To save processing time this table is not updated when the page is viewed or purged, it is only updated when a page is edited (on the assumption that that is the only time a page will have new content added that needs to be scanned for links), hence why you can only update the categories on a page by null editing them.
When it comes to categories there's another complication to consider - it is possible to add or remove categories from millions of pages at once by editing templates. To prevent the servers getting locked up trying to add categories to loads of pages at once mediawiki uses what is called a "job queue" - instructions that pages need to be added to or removed from a category via a template are stored up and processed when the servers are less busy - this is why it can take weeks/months for some template populated categories to empty.
The best solution here is probably a bot. The API has a special option you can set when you can set when purging a page called forcelinkupdate which forces the page to update the links table, something the normal Special:purge page doesn't do. A bot could scan over the category and purge any pages in it a couple of times a day. 192.76.8.94 (talk) 03:20, 16 June 2022 (UTC)[reply]
I don't know if the matter described at Wikipedia:Village pump (technical)/Archive 196#G12 category populated by ghosts was ever fully cleared up - if not, it is probably caused by the same problem.
Generally speaking, whenever you have a page (which we shall call "Foo") that is showing a category at the bottom (which we shall call "Bar"), but the Category:Bar page isn't listing Foo, the fix is to perform a WP:NULLEDIT on Foo. Similarly, if the Category:Bar page lists Foo, but Foo doesn't show Bar among its categories, again perform a null edit on Foo. Of course, you should have enabled Preferences → Appearance → Show hidden categories first. --Redrose64 🌹 (talk) 07:41, 16 June 2022 (UTC)[reply]
In my experience, this used to work reliably, but now I sometimes have to do both a purge and a null edit. Sigh... Peter coxhead (talk) 08:54, 16 June 2022 (UTC)[reply]
@Liz My User:Ahecht/Scripts/refresh script might help here. It can either do a forcelinkupdate or a null edit to every page in a category automatically. --Ahecht (TALK
PAGE
) 13:06, 16 June 2022 (UTC)[reply]

How to get page views for all fungi articles for a partner organisation

Hi all

I've recently started a project with Kew Gardens to try to improve coverage of plants and fungi. I'd really like to give them a number of page views to describe the audience for fungi articles. I've tried using Category:Fungi but this includes people and products eg Marmite and Beatrix Potter. The is a Wikiproject box which should capture all the relevant articles but I don't understand how to get this into a pagepile or similar. If someone could do this for me it would be really really helpful.

https://wp1.openzim.org/#/project/Fungi/articles

Thanks very much

John Cummings (talk) 10:58, 16 June 2022 (UTC)[reply]

One option is to use SPARQL on Wikidata. This query works just for Amanita. (Click the triangle in a square, bottom left.) In theory, you can replace Amanita's id (Q213938) by that of fungus (Q764). In practice, there so many results that it times out. You may be able to fix it by working through all the taxa at some suitable level one at a time. Certes (talk) 13:20, 16 June 2022 (UTC)[reply]

Installing AutoWikiBrowser in Linux

Hi,
I'm having trouble installing AutoWikiBrowser on Linux. I tried installing it with Wine, but it doesn't work. Can you walk me through how to correctly install it on Linux? Thanks, Interstellarity (talk) 12:34, 16 June 2022 (UTC)[reply]

Good luck: it can be a pig. User:Certes/AWB on Ubuntu may be of help. Some of what I did there is unnecessary, but I don't know which bits. It will also need changes if you use a different Linux distribution. I now use wine-5.0.3, which works at least as well as the 4.0.2 I originally had, so any version between should be fine. AWB now works for me, though not smoothly. I usually use JWB, which does simple things just as well, and reserve AWB for the tasks JWB can't handle. Certes (talk) 13:01, 16 June 2022 (UTC)[reply]
Asking on WT:AWB probably would be more fruitful in the future so other users can benefit from your question. :) Izno (talk) 13:50, 16 June 2022 (UTC)[reply]

Thursday issues in Vector-2022?

Hi all, to me, the right-part of the navigation links above a page (Read/Edit source/.../More/TW) appear to have shifted a lot towards the left and now slightly overlapping over the left-part of the navigation links (Article/Talk) on Vector 2022. I'm on mobile using desktop version but this is a brand new issue on it. Anyone else facing this? Something to do with today being Thursday? CX Zoom[he/him] (let's talk • {CX}) 20:15, 16 June 2022 (UTC)[reply]

The paler border beneath the top heading (even in legacy Vector) is also throwing me off... Nardog (talk) 21:42, 16 June 2022 (UTC)[reply]
Thank god I'm not going crazy. I opened a page and the line color change really threw me off balance. SWinxy (talk) 23:14, 16 June 2022 (UTC)[reply]

Preview image for Charles V, Holy Roman Emperor uses signature instead of portrait

I originally brought this issue up in the Teahouse, and was directed to here.

The preview image for Charles V, Holy Roman Emperor's article is the second image—his signature—even though the first image has a full portrait. I was told that maybe the dimensions of the image (1771×3000 px) makes it less preferable to the extension that picks the preview image, but it seems to me that the ratio (0.59 or 1.69, depending on how it's calculated) is well within the allowed ratio by the extension. Is there any way to force the extension to either pick the portrait or not pick the signature?

(By way of suggestion as well, could it help if the extension's criteria be modified to score colored photos higher than pure black-and-white?) — Preceding unsigned comment added by Yo.dazo (talkcontribs) 23:19, 16 June 2022 (UTC)[reply]

I've added notpageimage to the signature block of the infobox. This will remove signatures from being page images on all pages for all signatures. I think that's probably reasonable for signatures. Izno (talk) 23:42, 16 June 2022 (UTC)[reply]

Adiutor (new gadget proposal)

Hi everyone, I would like to propose a new gadget I coded this gadget with OOUI from scratch, this gadget helps various user processes, I first coded Adiutor for Turkish Wikipedia, then I thought about adapting it here, I don't know how the application process works here. I am still developing this gadget.

Features

Here is the gadget page: User:Vikipolimer/Adiutor

Thanks for your reviews. 𝗩𝗶𝗸𝗶𝗽𝗼𝗹𝗶𝗺𝗲𝗿 23:57, 16 June 2022 (UTC)[reply]

Hi Vikipolimer, we don't normally jump right to "gadget" level for new scripts until they have some uptake. Feel free to list at Wikipedia:User scripts/List and advertise it gently in places it could be relevant. I suggest you put a backlink comment in your directions so that you can track the 'what links here' adoptions. If you script starts getting quite popular, then yes this is the venue to ask for it to be moved from a WP:USERSCRIPT to a site gadget. — xaosflux Talk 00:23, 17 June 2022 (UTC)[reply]
Hi @Xaosflux thanks for your kindly response, I've already listed it in Wikipedia:User scripts/List. Well, you can archive this proposal I get the point now :) 𝗩𝗶𝗸𝗶𝗽𝗼𝗹𝗶𝗺𝗲𝗿 00:30, 17 June 2022 (UTC)[reply]
Twinkle already does speedy deletion and proposed deletion. I don't think you'd get approval for activity status for a gadget, but I think there is already a script for that. Recent diffs can be obtained from (ancient) popups gadget. (I don't know if there are others.) Copyright violations might reasonably be scripted, but OTOH I know that our copyright violations processes already suffer from insufficient processing access.
And any way, I'm not really sure all those different functions should be in one script.
I don't mind the idea that there should be competitors to our existing gadgets and scripts, but it might be more useful to work on those directly instead. (I understand that's maybe difficult to think about since you have already started and it was originally made for Turkish WP.)
As an aside, there's a general expectation that OOUIjs will at some point in the nearish future be deprecated in favor of Codex, which is also designed by the WMF and is based on Vuejs, so it uses modern Javascript. Izno (talk) 00:30, 17 June 2022 (UTC)[reply]
Thanks for your review and advice @Izno, I plan to add some key features to this script that will distinguish the new tool from Twinkle, of course, I'm not asking for new ideas here, but it's a good start to create a roadmap, thank you again. Also, I had no idea about Wikipedia will pass the VueJS. 𝗩𝗶𝗸𝗶𝗽𝗼𝗹𝗶𝗺𝗲𝗿 00:54, 17 June 2022 (UTC)[reply]
Wholeheartedly agree in not seeing the point of having these disparate features in one script. Nardog (talk) 01:00, 17 June 2022 (UTC)[reply]
Copyright violations might reasonably be scripted
Indeed it is (plus 4 others at WP:US/L). Guarapiranga  02:43, 17 June 2022 (UTC)[reply]

Infobox settlement

As a person who regularly works with categorization cleanup, I've noticed a recurring problem that I wanted to ask if anybody can solve.

What happens, over and over again, is that somebody adds a ((citation needed)) tag for the population of a community that's using ((Infobox settlement)) without explicitly referencing the population there — but because they frequently pop the tag directly onto the population number in population_total = instead of placing it in the population_footnotes = field where it really belongs, the automatic comma-delimiter function in population_total automatically also comma-delimits the year in the citation tag's transcluded maintenance category, resulting in Special:WantedCategories constantly getting cluttered up with nonsense categories like Category:Articles with unsourced statements from June 2,022 (note the comma in the year) where it really should have been Category:Articles with unsourced statements from June 2022.

But this obviously isn't a thing that those of us who work on category cleanup should be expected to just put up with repeatedly mopping up over and over again — it's a thing that by rights we should never be seeing at all, because it should be flatly impossible for the template to even be able to cause such a nonsense category to get generated in the first place. WP:TEMPLATECAT even explicitly states that "When templates are used to populate administration categories, ensure that the code cannot generate nonsensical or non-existent categories, particularly when the category name depends on a parameter".

Obviously we don't want to completely shut off comma-delimiting in that field, because it's necessary in the population numbers themselves, but is there any way that some sort of "firewall" can be coded so that it can continue to comma-delimit the population numbers while simultaneously being prevented from fucking up the maintenance category if somebody places the citation tag in the wrong field? Bearcat (talk) 18:46, 17 June 2022 (UTC)[reply]

If cn is particularly common, you could add some Module:String based code to remove a citation needed after display but before categorization. Izno (talk) 19:00, 17 June 2022 (UTC)[reply]
I don't know how to do that, which is precisely why I'm asking for help. Bearcat (talk) 01:56, 18 June 2022 (UTC)[reply]
@Bearcat, something like:
((#invoke:String|replace|source=((subst:cn))|pattern=%[%[Category:All articles with unsourced statements%]%]%[%[Category:Articles with unsourced statements from .+?%]%]%<sup class="noprint Inline-Template Template-Fact" style="white-space:nowrap;"%>%[%<i%>%[%[Wikipedia:Citation needed%((!))%<span title="This claim needs references to reliable sources%. %(.+?%)"%>citation needed%<%/span%>%]%]%<%/i%>%]%<%/sup%>|plain=false|replace=s))

(though there's some error as this doesn't work). ― Qwerfjkltalk 07:36, 18 June 2022 (UTC)[reply]

Help to reset a password?

A long-time editor I know (since '02!) who took a break for many years, now needs a password reset. They didn't set an email in user prefs, but list their email on their userpage. Is emailing ca@ still the right way to ask a dev to set the email for the account? – SJ + 20:41, 17 June 2022 (UTC)[reply]

@Sj: They (or you) can file a task on Phabricator. If they can prove that the account belongs to them, devs will likely act. NguoiDungKhongDinhDanh 01:51, 18 June 2022 (UTC)[reply]
@Sj: yes, emailing ca@ is the best way, especially if you can vouch for them being the legitimate account owner. Legoktm (talk) 05:10, 18 June 2022 (UTC)[reply]
File a task and reference it through ca@. 🐶 EpicPupper (he/him | talk) 05:11, 18 June 2022 (UTC)[reply]
Email ca@wikimedia.org. Trisbendo (talk) 14:02, 18 June 2022 (UTC)[reply]
The "right way" is just start a new account. A dev or wmf MAY help, but they may not. — xaosflux Talk 16:54, 18 June 2022 (UTC)[reply]