Elements API updates backport

[MaggieCabrera]

This PR backports the following changes:

• Theme Json: Don't output double selectors for elements inside blocks gutenberg#40889
• Global Styles: Load block CSS conditionally gutenberg#41160
• Add support for button elements to theme.json gutenberg#40260
• Global Styles: Button Element: update button element selector gutenberg#41240
• Elements: Add an API make it easier to get class names gutenberg#41753
• Elements: Button - Fix element selectors gutenberg#41822
• Support pseudo selectors on elements in theme json gutenberg#41786
• Global Styles: Allow references to values in other locations in the tree gutenberg#41696
• Global Styles: Add support for caption elements gutenberg#41140
• Fix link element hover bleeding into button element default styles gutenberg#42072
• Add visited to link element allowed pseudo selector list gutenberg#42096
• Global Styles: Add support for heading elements gutenberg#41981
• Link elements: Add a :where selector to the :not to lower specificity gutenberg#42669
• Theme JSON: Add a static $blocks_metadata data definition to the Gutenberg instance of WP_Theme_JSON gutenberg#42776 ⚠️ I don't think there's anything to backport on this one
• Pseudo elements supports on button elements gutenberg#43088
• Theme_JSON: Use existing append_to_selector for pseudo elements gutenberg#43167
• Duotone: Fix CSS Selectors rendered by theme.json duotone/filter settings for blocks on public pages gutenberg#41335
• Block styles: Account for style block nodes that have no name gutenberg#41446
• Styles API: Fixed selectors for nested elements gutenberg#43988

Trac ticket: https://core.trac.wordpress.org/ticket/56467

---

This Pull Request is for code review only. Please keep all other discussion in the Trac ticket. Do not merge this Pull Request. See GitHub Pull Requests for Code Review in the Core Handbook for more details.

[MaggieCabrera]

While trying to fix the tests I found a bug that wasn't caught in GB because the test is only on core, not in the GB repo. I'm going to try and patch that so we can add it to this one or we could "hack" the snapshot and then patch afterwards. I'm more inclined to do the former.

[MaggieCabrera]

This is the fix for the issue that the tests uncovered.

Edit: merged and backported here too

[MaggieCabrera]

this only needs a review and check what's breaking the tests (they are all passing locally)

[SergeyBiryukov]

The tests previously failed on PHP 5.6 with a fatal error:
Fatal error: Cannot use isset() on the result of an expression (you can use "null !== expression" instead) in src/wp-includes/class-wp-theme-json.php on line 1526

0015d99 should resolve the issue. We already used array_key_exists( $element, static::VALID_ELEMENT_PSEUDO_SELECTORS ) in one place, and that resolves the error in other instances in my testing.

[SergeyBiryukov]

This looks good to me 👍 Made some minor adjustments to documentation and unit tests for consistency with the rest of core. Going to merge shortly.

[SergeyBiryukov]

Thanks for the PR! Merged in r54118.

[ockham]

Thanks a lot, folks!

I was alerted that unfortunately, r54118 caused a fatal error on websites running Core trunk and a current version of the Gutenberg plugin 😕

PHP Fatal error:  Cannot redeclare wp_add_global_styles_for_blocks() (previously declared in wp-includes/global-styles-and-settings.php:202) in wp-content/plugins/gutenberg/lib/compat/wordpress-6.1/get-global-styles-and-settings.php on line 12

We'll need to add a ! function_exists() wrapper around the function definition in Gutenberg to avoid this.

Unfortunately, this won't fix the problem immediately: WP sites running Core trunk with the GB plugin active will continue to be affected until a new version of the plugin is released (as planned for next Thursday); or until the change is merged into GB trunk (if the site is running GB trunk). And finally, sites running Core trunk and an older version of GB (that has the unguarded wp_add_global_styles_for_blocks function definition) will continue to be affected 😕

I'm not sure how to best cover the latter case. For mitigation, we might want to consider reverting r54118 (as much as it pains me to suggest that) -- at least until a GB version with the fix is released.

---

Going forward, we should probably put some strategies in place to avoid collisions like this:

• Require wp_ function definitions in GB to be guarded with a ! function_exists() check (and ideally enforce with a WPCS lint).
• Add a check in wordpress-develop's CI to install and enable GB alongside Core (and run some basic tests to make sure it doesn’t fatal).

Anyway, this is something we'll discuss separately. I'll file one or two issues tomorrow.

[ockham]

Update: Unbeknownst to me, the wp_ prefix has been changed to gutenberg_ in the GB plugin: WordPress/gutenberg#44052

[ockham]

Update #2: Let's hold off on the revert. We might release a Gutenberg point release with this fix as a stop-gap.