In T395937#10883927, @Esanders wrote:

it should be achieved with a more localised wrapper around the menu itself

Per the commit message, this alignment appears to be new. Looking at any patchdemo before March you can see the toolbar did not go off the page:
https://patchdemo.wmcloud.org/wikis/a02452fc4f/w/index.php?title=Special:UserLogin&returnto=The+Hitchhiker%27s+Guide+to+the+Galaxy&returntoquery=mobileaction%3Dtoggle_view_mobile

This seems high priority, as it means users often have to scroll much further for an edit button, but probably not UBN. cc @ppelberg

In T395937#10881014, @Jdlrobson-WMF wrote:

Removing the overflow adds horizontal scroll to all mobile browsers.

It looks like the fix was designed to hide the horizontal scrollbars caused by the personal tools and overflow menus going off screen slightly:

Since that post there are now other links generated by DT that should be excluded:
Links to the latest comment on the page / in the thread:

As we implemented an API in VE for this I would recommend the following approach in future:

• Use a regex to extract a list of unique URLs on the page
• In a debounced content change listener:
  • Recompute this list, and check to see if any new URLs have been added
  • Pass those URLs off to the VE API and show any resulting warning

and renaming it to _code isn't very helpful

Can you give some examples of the text you are copying, and some screenshots of the errors?

I'm not opposed to consolidating and then migrating away from templatedata-doc-subpage

Caused by https://gerrit.wikimedia.org/r/c/VisualEditor/VisualEditor/+/644554 (T269145)

In T379341#10822938, @DLynch wrote:

Note that we'd need to think fairly hard about when to invalidate these checks. Showing someone a stale check is probably worse than showing them no check at all, after all.

I'm still getting access denied when trying to log in to spiderpig: Service access denied due to missing privileges.

The above patch demonstrates rendering the chart with a <div> and updating it with the wikipage.content hook, without adding any complexity.

If be interested to hear what these benefits are, because custom tags have caused problems for both parsing and editing teams. We have lots of other dynamically infused content on the page and the wikipage.content is the standard way to support this.

The greylink functionality doesn't actually exist yet, so QA will be done in T347209.

I think having both "Get shortened URL" and "Download QR code" in the menu is unnecessary now. If we can come up with a short label that explains you get both I think we can get rid of the download link from the toolbar.

Fixed in https://gerrit.wikimedia.org/r/c/mediawiki/extensions/UrlShortener/+/1141090

Block information is fetched in EditorGateway#getContent, which uses a GET request so that could be the issue...

Note that when opening a new private tab, the error goes away - which suggest some sort of caching is to blame. (Maybe network cache?)

On further investigation, it appears this only applies to the first row. Presumably the code that "cleans up" this optional first line of |- is not checking if it has meaningful attributes.

This is a Parsoid bug. Updated description.

I think there's another case when a table consists of a header row followed by data rows. I can imagine in these cases the header row has different styles to the data row. If clicked on the header row and did insertBelow, would I want to clone styles from the data row (as the new row will be a data row)? What if I wanted to add a second header row?

Discovered while implementing T389961

In T52098#1848459, @Jdforrester-WMF wrote:

Sure. But even if Parsoid does (even more) magic stuff, we should do it properly in VE.

<a rel="mw:WikiLink" href="./Foo">Foo</a>
<a rel="mw:WikiLink" href="./Foo">foo</a>
<b><a rel="mw:WikiLink" href="./Foo">Foo</a></b>
<b><a rel="mw:WikiLink" href="./Foo">foo</a></b>
<a rel="mw:WikiLink" href="./Foo"><b>Foo</b></a>
<a rel="mw:WikiLink" href="./Foo"><b>foo</b></a>

->

[[Foo]]
[[foo]]
'''[[Foo]]'''
'''[[foo]]'''
'''[[Foo]]'''
[[Foo|'''foo''']]

T393008 seemed to describe a simple failure case where a whole word is linked and formatted.

This sounds like a duplicate of T55590

We moved from throwing-methods to null in T54482. The reasoning given at the time was:

... this avoids us having pointless code that is never executed, and the replacement notification to the user (that ve.blah.UnicornAnnotation.foo is null and not a method) is sufficiently clear that the issue lies in the lack of existence of the ve.blah.UnicornAnnotation#foo method.
This removes 46 such methods from our codebase (1970 -> 1924), and so incidentally boosts nominal code coverage for the functions...

See also:

In T339806#9112293, @TripleCamera2022 wrote:

The layout issue still exists: The "Edit template data" button is split into two lines.

https://www.mediawiki.org/wiki/Template:!!?uselang=en&safemode=1

Besides, the current style may be misleading: "Template parameters" is not a heading, but a caption.

In T339806#9112293, @TripleCamera2022 wrote:

The layout issue still exists: The "Edit template data" button is split into two lines.

https://www.mediawiki.org/wiki/Template:!!?uselang=en&safemode=1

Besides, the current style may be misleading: "Template parameters" is not a heading, but a caption.

There is plenty of precedent for this, e.g. T136628 (cswiki) T370158 (eswiki) T116603 (frwiki). As long as discussion pages are in the minority, which it appears they are, this shouldn't be a problem.

@Quiddity @UOzurumba still holding unfortunately.

@Ryasmeen is this good now? Did you mean to add Verified?

The message was changed in T384865. The rest of this task is complete.

Solved by T269145

Looks like a regression from https://gerrit.wikimedia.org/r/c/design/codex/+/1053365, which added a __wrapper div, thus breaking the child selector (&:hover > &__input:enabled).

Looking at the code this looks unintentional and/or a regression. binary-input.less which is used just by Checkbox and Radio tries to apply cursor:pointer on hover, but it the selectors are incorrect.

This was fixed in T314952, which was a duplicate of this task.