If I'm reading it right, an integration test would have probably caught the bug in T331412 during code review by Verified -1ing the patch. It was a simple PHP InvalidArgumentException. Plus I need to write an integration test to safely refactor (gerrit 865581). So that's my idea behind this ticket.

Reopening per T329300#8681686. Sounds like this wasn't intentional.

Is deprecating the use of gray in skins really a good idea? This change has now created a situation where some icons are black and some icons are gray, which is inconsistent. This is certainly not an aesthetic improvement. I think an exception should be considered for Vector 2010, returning the bell and tray icons back to their original gray color.

Explanation of what probably happened on wikitech-l mailing list:

Thanks for the feedback. I created T331598: Fix Special:NewPages Feed "Review" button word wrapping in Chinese and T331595: Rename $wgPageTriageEnableEnglishWikipediaFeatures to $wgPageTriageEnableExtendedFeatures. Feel free to create your own tickets in the future as you run across PageTriage bugs, feature requests, and ideas.

@Tgr Works for me at your history link. But did you try watchlist?

Thanks for the suggested code. I've created a Twinkle ticket, and we will investigate adding that.

Could go even further and remove the custom namespace code completely ($wgPageTriageNamespaces), hard-coding PageTriage to only work in mainspace, simplifying the code. I'd probably be in favor of this, now that I think about it. YAGNI.

It's a really rare bug. It's been in production since 2018. Probably OK to skip backport unless you have the free time for it.

@Chlod suggests this might be a ResourceLoader issue:

I am also getting JavaScript errors on that page.

Thanks for that link Kosta. This one is buried in there but also looks relevant: https://en.wikipedia.org/wiki/Wikipedia:WikiProject_Articles_for_creation/AfC_Process_Improvement_May_2018/Copyvio_solutions_comparison_report

Below is what happens if a signature is included. #1 is no signature, #2 is signature. With that screenshot in mind, I think "no signature" is best.

Good point. Maybe a good next question is, what is the hypothetical most expensive query if all check boxes are checked? We should run it through quarry and see how long it takes.

Title::exists does not show up in a search of the codebase.

Confirmed that these are usually bluelinks instead of redlinks outside of PageTriage (I just tested in a sandbox). Legoktm's comment above appears to be the fix.

Howdy. I'm a Twinkle maintainer. Here's a couple options.

In T329952#8628983, @Samwalton9 wrote:

A related confusion they had was why these four checkboxes are separate from other article issues, like being an orphan or having no categories. My assumption is that this is because the That: filters are definitive, whereas the Potential issues: filters may be inaccurate?

I'm able to reproduce. This is not fixed for logged out users. Reopening.

The issue is fixed for me when using the link in the original post. Interesting that CMD is still getting it. I wonder what's different about their environment.

https://en.wikipedia.org/wiki/Wikipedia_talk:WikiProject_Articles_for_creation#ChatGPT_and_other_AI_generated_drafts has some additional discussion. @Querfjkl offers to write a tool or bot to run this on new drafts, and it is mentioned that GPTZero has an API. See bottom of https://gptzero.me/

In T330346#8640155, @kostajh wrote:

Could you list a few? OpenAI's classifier for detecting AI-written text has a 26% true positive rate.

No idea. Would like to discuss in this ticket. Any of the "implementation idea" bullets above are possible technical solutions.

action=purge didn't work for me at the time.

Steps to replicate the issue (include links if applicable):

On further examination, Special:Log already has this. You simply click on the date and time to see that revision. This feature isn't needed. Closing.

In T329952#8626572, @Tgr wrote:

Something being spam, copyvio and attack page at the same time seems like a very rare situation.

I'm slightly against this one. Not sure it's worth the clutter and complexity.

The "blocked" notice is probably most used as a hint to check for CSD G5. https://en.wikipedia.org/wiki/Wikipedia:Criteria_for_speedy_deletion#G5._Creations_by_banned_or_blocked_users

In T329660#8616570, @Etonkovidova wrote:

and legacy Vector

Good idea. Technical note: probably need to code this so that 3 conditions are met before trying to subscribe the user: