https://www.php.net/manual/en/function.strtotime.php is mostly to blame. T178175 more or less covers this.

@Addshore

When navigating multiple property talk pages, it is hard to remember where you are, as the title of the page doesnt show the label

In comparison the main page (non talk)

It would be great to include the label somehow on the talk page

No, but it's deprecated anyway.

Probably a duplicate of T130504: Some users are recognized as bots in recent changes and watchlists.

In T111649#1612725, @Bene wrote:

I guess the best option is to just ignore the clear parameter in the summary. The entity mostly doesn't get cleared but only some data is added and some is removed.

The patch got merged, I just unintentionally removed the reference to this task from it.

I'm not 100% sure if the patch resolves the issue. I found this independently, and now I think it could.

AbuseFilter does not treat this as a delete action, but as a move action. Something like action == 'move' & moved_to_prefixedtitle == "File:Example.png" & moved_to_id > 0 should catch it.

I am not sure if your experiment (dis)proves anything. The ChangesList interface does not allow users to run queries like these. Also, such a simple query could have been optimized by using subtraction (SELECT COUNT(*) FROM recentchanges - SELECT COUNT(*) FROM recentchanges WHERE rc_source IN (...)).

I think this is an oversight that stems from the unfortunate state of the codebase where thanks "eligibility" checks are done by each interface separately. Hence, if a new check is to be added (such as T407056), it needs to be duplicated everywhere where the eligibility for thanking (definitive or tentative) is checked.

Yes, that is an appropriate replacement in practice. user_type also allows 'external' and 'unknown', but these are usually found in retrospect.

This is how GeneratorsMixin.loadrevisions documents starttime and endtime:

:keyword starttime: retrieve revisions starting at this Timestamp
:keyword endtime: stop upon reaching this Timestamp
:keyword rvdir: if false, retrieve newest revisions first (default);
    if true, retrieve oldest first

I've come across a filter created by me in 2024 where I forgot user_type === 'ip'. Given that the variable has been available for over a year, it may be worth scanning all filters for this and similar expressions as well.

Types (a or rdf:type) for wikibase:Item, wikibase:Statement, wikibase:Reference, wikibase:Lexeme, wikibase:Form, wikibase:Sense are currently omitted for performance reasons.

In T334940#11214087, @Theklan wrote:

most things that aren't supported today in Charts will remain unsupported

Are you saying that our engineers took a software that will be unable to do "plot a point at x=2, y=5 and another one at x=3, y=6". Because that's a very wild declaration.

There is a slight ambiguity in the task scope.

Did it myself.

This was fortunately very obvious after it happened, and editing DISPLAYTITLE is pretty uncommon, so people have cleaned up most incidents. I fixed two more on enwiki that I found via this search, and did a quick check on other big wikipedias and didn't find any more.

Not sure if for this or the next issue. Communities should be made aware that some VE users could unintentionally insert artifacts into articles, and a cleanup is necessary.

Not sure if this 2020 task is still relevant today. Logging in was reworked server-side with SUL3, and we are now bound to use BotPasswords. (By the way, I filed T343187 in 2023.)

But, I do think it is rare for a number to "matter" (worth including) and yet not be worth elevating or simplifying in one of the above ways for greater usability/accessibilty or a higher quality translation.

Dirty quick solution: raise the limit of cached revisions (e.g., 10, 50, 100...).

Repurposing a little bit, as we now have the new PatrolManager service.

The problem with the existing ("free form SQL") framework is that it is based on exclusion, not inclusion. I have presented this in my analysis in T307328#10715066.

In T399996#11084249, @Umherirrender wrote:

Some future use: Instead of a bool, use a string and allow it to be changed by extensions. For example the MobileApp extension adds change tags for app edits and could set this value as well. Similiar to VisualEditor extension. But this makes it more complex, so it could be better to start with the bool.

It is preferred to submit patches via Gerrit.

I realize the task might not have been understood correctly. The title is "Revisit HTML...", that is, the "code" stored in the clipboard prior to copy-pasting it into a wiki (like T376306#10197664 or T379908#10322254).

Tone Check could have already been a small step in this direction. Some aspects of the language of machine-generated texts might overlap with the scope of Tone Check, or Tone Check could be adapted to that. See also https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing (in development).

It took around 13 seconds to compute all the variables for the 100 actions, and around 9 seconds of that was parsing content.

That's not a different font, it's just an excessive use of a combining character.

Edited the original post.

One ChatGPT artifact that started to emerge: :contentReference[oaicite:1]{index=1} or similar (example). Confirmed by... ChatGPT itself.

Please follow T399188 now.

Yes, everything looks correct now.

We have tickets with longer inactivity. I agree this should be fixed. Not only those, but also the other messages used in EditPage::showHeader.