I already created the new -footer page on dewiki a few days ago. I intented to proceed exactly the way you stated above, creating first and then history-merging later. Is everything ready for the change? Then I could do this for dewiki already.

Thanks, I'll have a look at it.

Default background has been added to top-level images in T370074 - which doesn't apply to the example because it's inside an infobox.

Is there a recommendation on how to proceed onwiki?

Maybe it would be beneficial to switch 'Groups you can edit' with 'Groups you cannot edit' to have the area which you visited the special page for more on top.

In T191156#10055466, @Dogu wrote:

Before	After

This is a local wiki issue. The module got deleted after discussion some years ago, so before recreating it, there should be consensus on whether this module is needed as it brings duplication to the existing wikidata module in some parts.

Thanks for applying the background. What about images in galleries? The gallery in https://de.wikipedia.org/wiki/Butylgruppe?useskin=vector-2022&vectornightmode=1 could also benefit from this.

The scroll offset attribute you are looking for should be scroll-padding-top

Sure. I was referring to the current approach using skin-invert in codemirror and the 2010 wikitext editor.

I reported this already in T371681. The fix should go live this week.

The temporary color inversion doesn't work with os clientpref by the way.

For zebra, there is also a css solution in place, but that doesn't work in all circumstances, e.g. with rowspans or repeating table headings. Adding the white background to the image instead of the background would work, but it's sometimes better to have some kind of whitespace around the actual image to avoid an odd look in both day and night mode.

I'm not sure I understand the question. #f8f9fa is light gray, #ffffff is plain white. Both colours are used for example to create a light "zebra style". In some infoboxes they are used to distinguish key and value parts. Sometimes plain white is needed to display transparent images correctly. Editors use plain white in cases "plain white" is actually needed (white jersey in Tour de France). Sometimes it's used to have a better contrast from gray.

The default background colour in wikitable tables is #f8f9fa, not #ffffff. The background shouldn't be removed. That's an editor's choice.

Left a message on https://www.mediawiki.org/wiki/Help_talk:Lint_errors/night-mode-unaware-background-color.

I wanted to wait a bit in case there is some caching involved (don't know), but removal is already scheduled.

Is this really useful as a public error category? I don't think we want thousands of hardcoded text colors in wikitext. The dark mode WikimediaMessages stylesheet fixes almost everything there.

Thanks!

There shouldn't be any stylistic rules in infobox.less, the additional padding was always breaking tables on minerva causing weird look and overflows when used with fixed inline width. Time to get rid of it and let TemplateStyles do it where wanted.

Hm, creating an y-scroll on the parent element indeed fixes the stickyness of the table header and I know it's probably the only way to maintain the sticky header with horizontal scrolling in a css-only way, but I'm not sure if this really creates a good ux:

• if the table height is smaller than the viewport (see https://de.wikipedia.org/wiki/Walensee which has a rather short table with sticky headers) there will be a lot of additional whitespace that fills the remaining viewport height.
• having many elements with vertical scrolling makes the page look cluttered, especially when at some point there will be horizontal and vertical scrolling combined.
• this is based just on my personal experience, but I find it quite annoying having two vertical scrolls in one page especially where the 'inner' vertical scroll size exceeds my viewport.

Maybe the sticky header issue (T366314#9873064) could be another task.

The italics rule is not present anymore in the current stylesheet.

Thanks Izno for filing the documentation task, I was about to suggest something similar if the floatleft/right classes will become mandatory for positioning. I guess the non-javascript version will be to directly output the wrapper div from parser?

The float-left and float-right classes are quite old, introduced in January 2007 (https://de.wikipedia.org/w/index.php?title=MediaWiki%3ACommon.css&diff=25858783&oldid=25637989) when sortable tables were new and it became clear that the now-deleted Template:Prettytable that was used for common template styling was too unflexible as it didn't support adding new classes for tables (Template:Prettytable had siblings Template:Prettytable-L and Template:Prettytable-R that did the floating stuff).

Side note: with https://pl.wikipedia.org/w/index.php?title=MediaWiki%3AGadgets-definition&diff=73889807&oldid=73887321 and directly removing the code from mobile.css in https://pl.wikipedia.org/w/index.php?title=MediaWiki:Mobile.css&diff=prev&oldid=73890354 you will probably be running into caching issues for anons like in T362747, so maybe it makes sense to postpone opting out of the global styles until the cache has cleared.

See https://www.mediawiki.org/wiki/Extension:WikimediaMessages#Site_admin_helper on how to disable WikimediaMessages styles locally. It's the infobox bit.

Maybe related to T274359

For now this is intended behavior, see T361934#9692764

With the upcoming dark mode, TemplateStyles stylesheets get bigger and also more complex (see this discussion at dewiki for an example). It would be nice to have TemplateStyles abilities extended with some more modern css. Browser support shouldn't be that much of a problem anymore five years after this task was created.

It's fine now, could be marked as resolved IMO

Nice to see that a "fix" (it's not really a bug) is on the way. Thanks for having a look at this.

Before I start using this onwiki: is there a risk of this later being revoked following the security review?

For using table syntax inside of a parser function, you need:

{| class="wikitable"
{{#if: {{{param|}}}
  | {{!}} foo {{!}}{{!}} bar
}}
|}

which renders as two table cells if param is set.

{{!}} needs to be treated as part of the table syntax for usage in parser functions.

Seems like the cache got me, now I'm seeing the correct styling, too. At first I checked with disabling/enabling the responsive skin option and saw the difference after doing that, so I thought anons don't have the responsive mode activated by default and didn't check the presence of the class. Sorry for inconvenience.

Don't know if that helps for investigation, but while the "stash parsoid html" error occurred today, I also got an OAuth E004 missing request token error when trying to login to Phabricator. It worked again the same time as the parsoid error disappeared.

This happened also in dewiki for every page and for discussiontools, now seems fine again.

That's the same as T361940 which already has a connected patch to revert the change.

Other revisions without text change like protections and moves are also affected.

Could it be that FlaggedRevs removes the tags applied by core? I saw something in maybeMakeEditReviewed but I don't fully understand everything there.

There are four changes that got the tag ("Zurückgesetzt" is reverted):

• https://de.wikipedia.org/w/index.php?title=Jassir_Arafat&curid=2493&diff=243768744&oldid=243066759
• https://de.wikipedia.org/w/index.php?title=Lanterne_Rouge&curid=3723391&diff=243768858&oldid=243768411
• https://de.wikipedia.org/w/index.php?title=Blixa_Bargeld&curid=61609&diff=243769643&oldid=243621320
• https://de.wikipedia.org/w/index.php?title=Fallschirmspringen&curid=30040&diff=243782947&oldid=243482341

Can confirm this in template namespace that also has FlaggedRevs activated. https://gerrit.wikimedia.org/r/c/mediawiki/extensions/FlaggedRevs/+/1015176 has changes to FlaggedRevs and reverts but seems unrelated to me at first glance.

Yes, the content looks different depending on skin (and that's okay), but my impression is that especially the wikitable styling is pretty much the same across all wikimedia deployed skins, and because the background styling is already in ResourceLoader skin module, it makes sense to also apply the background consistently. If from a design point of view it makes sense to have some kind of visual connection between table and caption (I don't know), then I'm fine with it, but it should be considered that some tables make use of completely different background colors in the entire table which would leave the caption area as the only one with the initial background - then the visual connection seems rather weak and random to me.

This should be a duplicate of T361775

I think this is caused by https://en.wikipedia.org/wiki/User:Headbomb/unreliable.js which applies #ffdddd background color as inline style to a.external when a link to wikipedia gets matched. plainlinks overrides the background color, but the inline style persists and gets chatched by the night mode rule.

I think this got fixed by https://gerrit.wikimedia.org/r/c/mediawiki/core/+/833413, right?

Thanks a lot @matmarex and everyone else involved!

Thanks for clarification, then it's the additional styling causing the issues. We're discussing approaches for fixes at https://de.wikipedia.org/wiki/Wikipedia:Technik/Skin/MediaWiki/%C3%84nderungen#Ge%C3%A4nderte_%C3%9Cberschriftenstruktur but it doesn't seem so easy at first glance - maybe you have an idea, your input would be appreciated. For dewiki's main page I did this: https://de.wikipedia.org/w/index.php?title=Wikipedia%3AHauptseite%2F%21box&diff=240191659&oldid=213295348

Since the update of 1.42.0-wmf.9 it seems like DiscussionTools already applies the heading html transformation - I had to hotfix dewiki's main page as it looked broken. You can see the broken look still on other pages like https://de.wikipedia.org/wiki/Wikipedia:Autorenportal or https://de.wikipedia.org/wiki/Hilfe:%C3%9Cbersicht. I'm also not sure why DiscussionTools is active on these page(s), though.

D is the marker of a wikidata item and K stands for minor edit in German.

Caused by T350443? Database seems to have high lags

Sorry, I restored the test case and updated my comment.

Should't the category be named wikihiero-usage-tracking-category?

The DOM looks odd, we have div and p inside span now and the mw-fr-reviewformlegend span got an 'additional' cdx-card__text id which should be in class= I guess.

reproduced here using the original article contents:

Thanks! Task should be resolved then.

jobs are all moved to kubernetes, however it seems there are 16 running jobs on GridEngine left that are stuck, don't know if and how I can abort them.