Check if we still need details.css

catch → credited jacine → .

I can not reproduce this issue:

- Using Safari 18.3 on mac OS 15.3.1, full keyboard access active (otherwise Safari won't tab focus links at all)
- Codepen with three details elements, each with a paragraph of text and one with an additional link inside
- all collapsed: tab just highlights the three summaries, no additional invisible focus in between
- with the details with the link expanded: tab highlights the three summaries and also the link in between

Not sure when it got fixed though.

So the CSS was preserved in 2013 when details support was added. There is no reference to why it was preserved, but not all browsers supported details at that point so we didn't even use the proper html5 element for it yet, was a JavaScript quasi-polyfill instead with a @todo to add a real html5 polyfill later.

This is the diff that preserved it:

--- a/core/modules/system/system.base.css
+++ b/core/modules/system/system.base.css
@@ -38,25 +38,13 @@
 }
 
 /**
- * Collapsible fieldsets.
+ * Collapsible details.
  *
  * @see collapse.js
  */
-.js fieldset.collapsed {
-  border-bottom-width: 0;
-  border-left-width: 0;
-  border-right-width: 0;
-  height: 1em;
-}
-.js fieldset.collapsed .fieldset-wrapper {
+.js details:not([open]) .details-wrapper {
   display: none;
 }

That issue doesn't mention any safari accessibility bugs, and I just noticed now that the css uses the .js class which means it was specifically for the js implementation of details (not the browser-native one that came later).

Given all that and the testing from @feuerwagen I think it's safe to drop.

Postponed 📌 Move details.css to the collapse library Active on this issue and put up an MR.

While this is a tiny amount of CSS, we duplicate individual CSS files quite a lot in aggregates still even after recent improvements, so files in the system/base library can end up getting downloaded multiple times by the same visitor from different pages. Also in situations like the installer where there's no aggregation, it's one less http request.

Seemed to cause a number of test failures.

Could only see one known random failure, but did a rebase anyway.

Yea not sure what was happening when I looked but all seems good now.

nod_ → closed merge request !11449

Committed 365250d and pushed to 11.x. Thanks!

Automatically closed - issue fixed for 2 weeks with no activity.