User talk:Legobot

I stumbled across this Linter fix, which seemed like an easy one that Legobot should have gotten to. Is Legobot still patrolling the obsolete tag lists for all possible font tag fixes? – Jonesey95 (talk) 23:52, 4 December 2025 (UTC)

Hey @Jonesey95, it was busted for the same reason the MFD archiving task above was also broken. I should have fixed it, we'll find out tomorrow. Also, I finally deployed the code I worked on back in January(!!) to support using arbitrary regexes for fixes. Hopefully shouldn't be too difficult to expand it if everything else is working again... Legoktm (talk) 05:22, 21 December 2025 (UTC)

Fingers crossed. We have a couple million more Linter errors to fix, and many of them need a regex bot or script to handle efficiently. – Jonesey95 (talk) 17:08, 21 December 2025 (UTC)

@Jonesey95: Legobot is still skipping all userspace pages, based on a suggestion you made during the BRFA (though now that I re-read it, I think you might've meant for that to be temporary). Do you think it should keep skipping those or should I remove that restriction? Legoktm (talk) 17:27, 30 December 2025 (UTC)

I recommend working on User pages at this point. My caveat was for the initial run and deliberately used the words "for a while" to indicate that there was plenty of work to be done in other namespaces. When I edit User pages, I typically amend my edit summary to include a phrase like "I hope you don't mind this minor cleanup edit in your user space" or "Feel free to restore, preferably if errors are fixed", since User space is generally allowed more latitude for experimental and draft content and formatting. – Jonesey95 (talk) 17:33, 30 December 2025 (UTC)

I also think it's safe to let the bot fix userspace. I've hardly ever been reverted fixing issues. Gonnym (talk) 19:42, 30 December 2025 (UTC)

Thanks both, it's running now. :) Legoktm (talk) 04:05, 31 December 2025 (UTC)

How's that regex code implementation looking? I keep looking at these 4,000 pages, and the Welcome messages without ending /div tags, and some of the other giant blobs at Wikipedia:Linter/Signature submissions that require a bit of regex, and hoping that Legobot will be able to tackle some of them. I have a bunch of the regexes ready, but I don't have bot-making ability, and doing thousands of identical replacements manually gets old fast. – Jonesey95 (talk) 00:28, 14 January 2026 (UTC)

It should've been running but after digging in a bit because of your comment, I realized I made a silly error, it was only pulling the list of pages with obsolete-tag errors and not also missing-end-tag, so it wouldn't actually find any unless they also happened to have an obsolete tag, which I guess none did. Now fixed and started a run. Legoktm (talk) 03:51, 21 January 2026 (UTC)

Also just to be upfront, my school semester just started again so I'm probably going to be dipping back into hibernation mode once the workload picks up. Legoktm (talk) 04:02, 21 January 2026 (UTC)

I appreciate anything you can do. Good luck with school. I'll try to remember that you'll be busy IRL.

If you're looking for one last list of maybe 50,000 pages to feed to the bot, the query at User:Jonesey95/Pages with many strike tags should be a fruitful one, and the last two search links for the Welcomeg and WelcomeMenu messages could provide a lot to work on. – Jonesey95 (talk) 01:32, 22 January 2026 (UTC)

Fixed a different issue in the code and here we go! Legoktm (talk) 18:34, 24 January 2026 (UTC)

Excellent. The only way we are going to reduce the 2.5 million error backlog of Linter errors in a reasonable amount of time is with bots. Thanks for continuing to work on this even while school is in session. – Jonesey95 (talk) 21:16, 24 January 2026 (UTC)

This is amazing. The bot just keep chugging away, fixing these missing /div tags in welcome messages at a rate of about 9,000 per day. It has fixed tens of thousands so far, and there are still tens of thousands remaining, apparently. Go, bot, go! – Jonesey95 (talk) 20:36, 2 February 2026 (UTC)

I've kind of been waiting for it to run out of pages...nope! :) 77k+ missing end tag edits so far. I have it capped at 6 edits per minute, so it should be ~8.6k edits per day. Legoktm (talk) 03:12, 3 February 2026 (UTC)

Yeah, this is pretty amazing. Without a bot this would be VERY tedious. Another good candidate once this is done is maybe to take over Wikipedia:Bots/Requests for approval/SheepLinterBot 2 which stopped working 4 months ago and the search results still shows over 11k pages. Gonnym (talk) 11:54, 3 February 2026 (UTC)

Agreed about the Sheep task. Many of those 11,000 pages have six or seven Linter errors each. – Jonesey95 (talk) 15:24, 3 February 2026 (UTC)

OK, I've implemented a basic fixer for the TWA/Earth pages: example. It's just enabled in the web interface for now; if it looks reasonable then I can enable it in the bot once the current run finishes. Legoktm (talk) 08:59, 8 February 2026 (UTC)

That looks like what is needed on those TWA/Earth pages. I have fixed a few, and they all had the same stripped /div tags at the top and missing div tags at the bottom. I think some of them do not have the center tags. – Jonesey95 (talk) 15:09, 8 February 2026 (UTC)

This is just insane how many pages the WelcomeMenu error was on. We've gone down 370k in errors in 2.5 months, and this is probably a large percent of it. Gonnym (talk) 08:01, 18 February 2026 (UTC)

Agreed. It is bonkers. When I posted the "Welcomeg" section at Wikipedia:Linter/Signature submissions, I noted that 320K pages appeared to contain the substed templates, but I did not think that there could really be that many affected pages. Now I'm thinking that there were probably more. All I knew at that time was that multiple insource searches were failing to pull up all of the pages. I have noticed that the two search links with the word "pattern" in their link text have decreased over the past month or so as Legobot has been working, but they still pull up thousands of pages, decreasing only a little each day despite Legobot's 9,000 edits per day for the last ~25 days (since 24 January, per "here we go" above). – Jonesey95 (talk) 22:00, 18 February 2026 (UTC)

And another crazy thing is that Legobot only saves fixes to pages when there are no remaining Linter errors on them. That means that a page with a missing closing p tag, of which there are zillions, or any other missing or stripped or misnested or obsolete tag that Legobot is not programmed to fix, will be left alone. That means that when the bot finally gets through this batch of pages, there will probably be a few thousand pages that are still missing /div tags at the end of welcome messages.

I'm surprised but thrilled that there are so many of these pages and that so many of them are bot-fixable. Two million Linter errors is daunting. If a bot can get rid of a few hundred thousand over the course of a month or two, that's amazing and saves us humans a lot of work. – Jonesey95 (talk) 21:58, 18 February 2026 (UTC)

Some of them would appear at first glance to be my fault. Examples from the last two days: Al C1964; Ayrshirelou; NickCatford; Phill.d; Axonite2; TJWH. But of course I was merely using {{subst:WelcomeMenu}} as it then stood. --Redrose64 🌹 (talk) 22:35, 18 February 2026 (UTC)

Nobody is really to blame for these missing end tags. The software let us (and still lets us!) create and insert code without balanced tags, and then we went around welcoming people, which we were encouraged to do. So it goes. At least this one is easy to fix with a bot. Loads of one-off errors need manual intervention. – Jonesey95 (talk) 05:58, 19 February 2026 (UTC)

I don't think I've ever had a bot task run uninterrupted for this long, it's been 25 days and counting of non-stop fixing 😅.

I've been spot checking pages and my impression is that at least these user talk pages tend to be much simpler markup wise and often some other Linter bots have already made a pass. Legoktm (talk) 00:51, 20 February 2026 (UTC)

Seems it finally ran out of fixes. The next run with the TWA/Earth fixes will get us under 2 million. Gonnym (talk) 09:05, 23 February 2026 (UTC)

Here we go. I also started looking at some of the welcome messages that were skipped; this change should pick up some more pages that were incorrectly skipped earlier.

I don't understand why this potential fix isn't working as expected in case either of you can spot it. Legoktm (talk) 06:31, 24 February 2026 (UTC)

This is the fix. The div is not for the full table but for the column cell. Gonnym (talk) 07:03, 24 February 2026 (UTC)

This, respectfully, is a better fix. This was an edge case where there was a ! at the end of the user name, so the ! at the end of the text combined to make !!, which was interpreted as table markup. I boldly removed one of the ! characters, which nobody will care about, in order to preserve the formatting intent of the template. I could also have inserted a nowiki/ tag between the two ! characters, which I have done on other pages with this problem. It's a nasty one. – Jonesey95 (talk) 22:11, 24 February 2026 (UTC)

Incredible, that's cursed. Legoktm (talk) 23:51, 24 February 2026 (UTC)

Indeed, but at least it's valid parsing of unlikely input. Compared with T403931, T410345, and T417792, it's a tasty treat. – Jonesey95 (talk) 00:16, 25 February 2026 (UTC)

The TWA/Earth pages are now mostly finished, per this search there's a little under 400 left, which I think just have extra text before the 3 unclosed divs at the top (just need to loosen the regex). A bunch of welcome messages were also fixed but I don't have a count for those (yet). Legoktm (talk) 04:41, 27 February 2026 (UTC)

Super. That's something like 60K to 70K Linter errors resolved just from those TWA/Earth pages. See below for another batch of 18,000 TWA pages, most with three missing end tags each, along with two batches of welcome messages. – Jonesey95 (talk) 06:17, 27 February 2026 (UTC)

If I recall correctly, Legobot checks its proposed revisions before saving to see if there are any remaining Linter errors. Is it still doing that? I support the methodology; I'm asking because of the follow-up question below.

If it is doing that, is it ignoring duplicate id and background color errors? It probably should, since those are basically down-the-road errors that don't need to be fixed right now. They are also often caused by templates, so editing a given page that transcludes one of those templates won't fix the errors on that page. Am I making sense? Basically, if Legobot is checking for remaining errors, it should check only for those currently listed on the Firefly table. – Jonesey95 (talk) 18:10, 31 December 2025 (UTC)

Correct, it does check there are no more lint errors, and doesn't ignore any type. Matrix mentioned the same thing regarding the dark mode/background color ones. I haven't had the chance to fully read up on it but I was also leaning that way.

I think the duplicate ID ones Legobot shouldn't ignore, since it's using an HTML parser and that's invalid HTML. For ones where the duplication isn't via a template, I was wondering if we could just auto-renumber them, e.g. id="Foo", id="Foo2", id="Foo3", etc. Legoktm (talk) 03:15, 2 January 2026 (UTC)

The problem with duplicate ids is that many (most?) of them are caused by CS1 citation templates, and there is no desire at all to change the behavior of those templates to fix what is seen as a non-problem. And the problem with night mode errors, as I said above, is that many of them are caused by templates and can't be resolved in the wikitext of a given page. I'd be curious to know how many "Publish" operations are skipped solely because there are night mode or duplicate id errors remaining on a page. It seems a shame to force humans to perform bot-fixable edits.

Does Legobot at least ignore the "hidden" Linter errors like "large-tables" (id 20) and "missing-image-alt-text" (id 23)? Those are not reported in "Page information", primarily because the WMF developers are still working out the kinks to figure out whether the tracking of those conditions even makes sense to perform. – Jonesey95 (talk) 04:53, 2 January 2026 (UTC)

I agree with Jonesey here, the bot should ignore the duplicate id and background color as those aren't things we can really fix now. They require much more investigation and discussion (and code changes) so holding off on fixing fixable issues for them isn't really helpful to the project (and won't be a spam on watchlist as again, no fix for those two issues is on the current radar). Gonnym (talk) 19:51, 2 January 2026 (UTC)

@Jonesey95, @Gonnym: Yes, the hidden Linter error categories are ignored because they just don't appear in the API output.

As for numbers, I created a new webpage (warning: large page, will take some time to load) that groups all the pending edits by which errors are preventing it. In short: duplicate-ids is blocking 213 pages, night mode is 15k pages, duplicate-ids + night mode is 503 pages.

I do want to look at the duplicate-ids problem a bit more, e.g. this one seems trivially fixable by a bot just adjusting one of the IDs. My spot checking is that the majority are from templates, but I'm guessing there's some minority not from templates and just bad copy-paste issues. Legoktm (talk) 05:25, 4 January 2026 (UTC)

Fascinating list. I find it odd that it lists so many html5 misnesting errors that aren't actually present. We eliminated those errors from all of en.WP long ago. [edited to add: I think I get it, after working on some of the pages in the list: The html5 misnesting errors are present after Legobot applies its fixes. It is happening to me as well. They usually start out as regular misnesting errors, but the parser sees them differently after some wikitext changes are applied.] – Jonesey95 (talk) 14:49, 4 January 2026 (UTC)

Hmm, I need to look at this closer. In e.g. this page it's turning a <font> into a <span> instead of a <div>, which is causing the html5-misnesting issue...but I already have code to check for block elements, which should have it use a div. Legoktm (talk) 20:12, 4 January 2026 (UTC)

Why does it want to change [[Clay Felker]] to [[Clay Felker|<span style="color: yellow;">Clay Felker</span>]]? Gonnym (talk) 10:02, 5 January 2026 (UTC)

There's a <font size="+3" color="yellow" face="Comic Sans MS">...</font> element enclosing the whole of the page. All child elements inherit the size and face, but not all inherit the colour - a notable case being the <a>...</a> element (because clickable links should be distinguishable from plain text by their colours). Some years ago, a previous version of MediaWiki, upon encountering a construct like

<font color="yellow">[[Clay Felker]]</font>

would twist this around and emit HTML like

<a href="/wiki/Clay_Felker" title="Clay Felker"><span style="color:yellow;">Clay Felker</span></a>

MediaWiki no longer does that, which is why we launched the good ship delinter some years back. It's not just for picking up unclosed inline elements. --Redrose64 🦌 (talk) 12:12, 5 January 2026 (UTC)

Filed T415233, which from spot checking will unblock fixes on a thousand-ish pages once resolved. Legoktm (talk) 04:42, 22 January 2026 (UTC)

@Jonesey95, @Gonnym: Legobot is now ignoring dark mode errors. Legoktm (talk) 03:52, 21 January 2026 (UTC)

Super. That should help a lot, especially on discussion pages. – Jonesey95 (talk) 05:31, 21 January 2026 (UTC)

Amazing, thanks! Gonnym (talk) 18:44, 21 January 2026 (UTC)

Legobot modified my user page this morning. This violates the general convention of not modifying an editor's User page unless you are the owner. If my userpage somehow has a catastrophic effect on the functioning of wikipedia by using obsolete html tags - one day taking down the entire site for days because of my malfeasance - then sure. Otherwise, I wear my extraordinarily weak capabilities with html tags as a badge of honor. If my page alone breaks from this gross error, well then, I can live with it, and will likely attend to it within a few years.

But a little more seriously - is there any actual harm that can derive from this, other than my user page someday appearing wonky? cheers. anastrophe, an editor he is. 19:26, 21 February 2026 (UTC)

That is a bit surprising. You can opt out of bots with Template:Nobots Dw31415 (talk) 19:59, 21 February 2026 (UTC)

That’s generally speaking. I can’t promise Legobot respects that or there is something about your previous edit that is potentially harmful Dw31415 (talk) 20:01, 21 February 2026 (UTC)

Thanks! I'll give it a whirl. cheers. anastrophe, an editor he is. 20:08, 21 February 2026 (UTC)

Bots and other editors are allowed to edit pages in User space to fix problems. See WP:UOWN, including except when it is likely edits are expected and/or will be helpful. The bot clearly stated what it was doing and why, and linked to a page with further explanations that explain why such edits are helpful. The bot did not change the appearance of the page, and it fixed an error that appeared in the "Page information" page for your user page. This bot and other bots have cleaned up millions of these errors in every namespace. I hope that explanation helps you understand what may have been a surprising edit. – Jonesey95 (talk) 21:08, 21 February 2026 (UTC)

Yes, I read the edit summary it provided. I linked to WP:UOWN in my first message. I was specifically referencing "Bots and other users may edit pages in your user space or leave messages for you, though by convention others will not usually edit your user page itself, other than (rarely) to address significant concerns or place project-related tags."

'Helpful' is in the eye of the beholder with non-critical changes, in my opinion. But, whatever. I've added nobots, hopefully it will be another twenty years before a bot decides to fix my user page. cheers. anastrophe, an editor he is. 21:35, 21 February 2026 (UTC)

p.s. please don’t trust my answer. I’ve just been following Legobot for a while and don’t know what I’m talking about. Sorry for over stepping Dw31415 (talk) 21:40, 21 February 2026 (UTC)

Hi @Anastrophe. As mentioned, the nobots system will cause Legobot to skip your userpage. Can you explain why you want to keep using <font> tags?

In addition to the links Jonesey provided, I've also written an essay that explains why cleaning this up is a good thing. To answer your point, no, it's unlikely anything bad will happen from continuing to use obsolete HTML tags, but things could be much much better if we got rid of them and could rely on pages being clean of any lint errors. Legoktm (talk) 22:42, 21 February 2026 (UTC)

I think I already explained my rationale; I'm an obstinate bastard! If someday down the road the HTML tags become non-functional and my page (mildly in the exteme) 'breaks', i'll fix them then. This isn't a big deal; it was just a bit of a surprise, because the only previous edits by any entities other than myself had been by vandals, and those who smacked down those vandals. To be absolutely crystal clear, I know this wasn't vandalism and I didn't even slightly consider that it was vandalism. Just a surprise. cheers. anastrophe, an editor he is. 22:52, 21 February 2026 (UTC)

@Anastrophe: that's pretty disappointing to hear. If someone crossed paths with a project you'd been working on for a decade, undid your work and when asked why, merely said "I'm an obstinate bastard!" I think you'd be justifiably upset.

Wikipedia is a collaborative project and we all make compromises for its betterment; I hope you'll reconsider. Legoktm (talk) 23:12, 21 February 2026 (UTC)

I'm unclear how I "undid your work". I did not. My user page is my own. It's my choice to configure it within the limits of policy, which I have done, and it's my choice what changes I allow others to make (which is none). The lack of compliance with what Legobot makes its decisions by is not critical, nor meaningful, to my userpage. I did not "Merely say 'I'm an obstinate bastard!'" (which was supposed to be lighthearted). I said I had already explained it. Please reread my initial post.

I'm sorry you're taking this personally, but legobot's usefulness, in my opinion, is in improving the public encyclopedia. That it does. My user page is irrelevant to the public encyclopedia. It should not matter what I do there, within the constraints of policy. I have opted out of the bot. Everybody should be happy. cheers. anastrophe, an editor he is. 00:29, 22 February 2026 (UTC)

In addition, there was already an RfC on this which validated this process. You reverting a bot, will only cause additional editors to redo this. Gonnym (talk) 08:54, 22 February 2026 (UTC)

Very interesting. Reviewed the deprecated tags page. It seemed a little wishy-washy about HTML5. I’d be interested to read the RfC and would appreciate any tips on finding it. Dw31415 (talk) 09:07, 22 February 2026 (UTC)

@Anastrophe, have you read about the evils of the font tag and the blessings of HTML5? Dw31415 (talk) 09:09, 22 February 2026 (UTC)

I'm uninterested in such matters, so no. cheers. anastrophe, an editor he is. 09:55, 22 February 2026 (UTC)

What do you make of @Gonnym‘s comment that other editors can (and should) clean up your tags? I guess it should be clarified if you have right to your HTML4 or are expecting to get on the 5 train. I’ts an odd hill to defend, especially if it’s not out of a deeply held conviction of the html spec Dw31415 (talk) 10:03, 22 February 2026 (UTC)

Even in HTML 4, the font element was deprecated. Some of the most significant features of HTML 4 as compared to HTML 3.2 were the class=, id= and style= attributes. Alongside these was the full introduction of the <style>...</style> element (having merely been proposed but not implemented in 3.2). Taken together with the existing <link /> element, these features allow CSS styling to be incorporated into HTML documents when rendered by the browser. As stated in the CSS 2.1 spec: By separating the presentation style of documents from the content of documents, CSS 2.1 simplifies Web authoring and site maintenance. --Redrose64 🌹 (talk) 23:12, 22 February 2026 (UTC)

Thanks @Redrose64, do you know if it’s consensus the the OP is entitled to his font tags or if other editors are encouraged to make the same edit Legobot made? Personally, I just ask because I’m curious. Dw31415 (talk) 02:53, 23 February 2026 (UTC)

I might consider buying a T-Shirt with “I already explained my rationale; I'm an obstinate bastard!” Let me know where to send $14.95 and a self-addressed and stamped envelope ;-) Joking aside, I’ll just express gratitude to the folks keeping Wikipedia clean and tidy and allowing us to drop support for IE-9. Dw31415 (talk) 23:50, 21 February 2026 (UTC)

The bot removed the use of active templates while clearing up lint errors on my user talkpage - this has been reverted. Link to diff Nightfury 22:20, 24 February 2026 (UTC)

(talk page stalker) Nightfury, that is a strange one. I have fixed the bot's edit for you, since the font tags did need to be updated. I believe that the page is working as intended now. I'll leave the explanation of the edit to the bot operator. – Jonesey95 (talk) 00:02, 25 February 2026 (UTC)

@Nightfury: sorry about that, it looks like an edge case that I didn't realize Parsoid treated differently. I've put in a temporary fix to have it skip pages if it finds that condition until I get a proper fix working. Thanks for letting me know! Legoktm (talk) 03:39, 25 February 2026 (UTC)

It looks like the bot could take another pass at articles with strike tags. I'm seeing Talk:Electoral system/Archive 6, Talk:Diver navigation, Talk:Diving reflex and more that seem like the bot would be able to fix. I'm using this simple query to find them. There are currently 3,686 pages with strike tags; it would be great if the bot could cut that down to the point where we could use User:Jonesey95/Pages with many strike tags to eliminate the rest. – Jonesey95 (talk) 00:21, 25 February 2026 (UTC)

The main limitation Legobot has by using Parsoid HTML to make changes is that it can't edit through templates, because even with VisualEditor, templates are edited as wikitext. So in Electoral system, the strike tag is inside archivetop/archivebottom templates, and then for Diver navigation and Diving reflex, there are raw #if parser functions surrounding the B-class review which is where the strike tag is. :/

At some point I'll need to figure out how to overcome this limitation by editing the wikitext directly but really haven't figured out how I want to go about that. Legoktm (talk) 03:15, 25 February 2026 (UTC)

Oof, that's a big limitation. The LintHint script is subject to a similar issue in that it will highlight a big block (like a block wrapped in atop/abot templates) as having a Linter error, making it more difficult to pinpoint the error. I will probably forget about this limitation on your bot and ask you about it again, for which I ask your forbearance in advance. I hope that you can find a way around it. – Jonesey95 (talk) 04:22, 25 February 2026 (UTC)

Constructive contributions are appreciated and strongly encouraged, but your recent edit to the userpage of another user may be considered vandalism. Specifically, your edit to User:BalaM314 may be offensive or unwelcome. In general, it is considered polite to avoid substantially editing others' user pages without their permission. Instead, please bring the matter to their talk page and let them edit their user page themselves if they agree on a need to do so. Please refer to Wikipedia:User page for more information on user page etiquette. Thank you. 『π』BalaM314〘talk〙 18:21, 27 February 2026 (UTC)

@BalaM314: You realize this is a bot right? Why do you want to keep using a <center> tag? Would you be okay with using {{center}} instead? Legobot and/or other humans are just going to keep trying to clean it up... Legoktm (talk) 19:21, 27 February 2026 (UTC)

The center tag is notorious to web developers. It's mentioned in this famous StackOverflow post. It's also one of the 4 HTML tags I used in my first experiment with web development. I don't mind if it doesn't render correctly. I don't approve of this bot editing my user page. 『π』BalaM314〘talk〙 19:30, 27 February 2026 (UTC)

@BalaM314: please keep in mind that you do not own your userpage, it's in service of the collaborative project, which wants to eliminate these tags. In any case, {{bots|deny=Legobot}} will stop this bot but at best it's a temporary stay, it's going to get removed at some point. Legoktm (talk) 20:34, 27 February 2026 (UTC)

Hi @BalaM314, I just started a discussion at Wikipedia talk:User pages#c-Dw31415-20260227234100-Deprecated format tags. This bot is just carrying out the approved replacement of tags. May I suggest we take it to that venue? If not, I thinks it’s fine for you to deny this bot for now, but better if you help improve the accessibility of Wikipedia. Dw31415 (talk) 23:48, 27 February 2026 (UTC)

Hi, recently I went through WP:DASHBOARD and removed the fixed colour attributes from the dashboard elements that had the blue background colour, so as to make the text readable to dark mode users. It seems Legobot is set to restore these values each time it updates the page Wikipedia:Dashboard/Relisted AfD debates.

Could changes be made so it stops adding the colour attribute? Here is the diff of the changes I was trying to make. Thanks! SnowyRiver28 (talk) 18:20, 10 March 2026 (UTC)

https://github.com/legoktm/harej-bots/pull/20 * Pppery * _{it has begun...} 21:08, 10 March 2026 (UTC)

Thanks both; should be updated on the next bot run. Legoktm (talk) 01:43, 11 March 2026 (UTC)