In this patch, the test suite from html5lib validates the tree-construction
steps in the HTML Processor to ensure that they are behaving according to the
HTML specification. This suite of tests is also used by the servo project to
test its html5ever package.
A new test module in the HTML API transforms HTML Processor output to match
the expected tree shape from the external tests. For cases where there are
tests validating behaviors of unsupported HTML tags and constructs, the tests
are marked as skipped. As the HTML API continues to expand its own support,
the number of skipped tests will automatically shrink down towards zero.
Additional tests are skipped through the `SKIP_TEST` array in the test runner.
Fixes#60227.
See #58517.
Props azaozz, costdev, dmsnell, hellofromtonya, jonsurrell, jorbin, swisspidy.
Built from https://develop.svn.wordpress.org/trunk@58010
git-svn-id: http://core.svn.wordpress.org/trunk@57481 1a063a9b-81f0-0310-95a4-ce76da25c4cd
In PHP 8.4, declaring function or method parameters with a default value of `null` is deprecated if the type is not nullable.
PHP applications are recommended to ''explicitly'' declare the type as nullable. All type declarations that have a default value of `null`, but without declaring `null` in the type declaration, will emit a deprecation notice:
{{{
function test( array $value = null ) {}
}}}
`Deprecated: Implicitly marking parameter $value as nullable is deprecated, the explicit nullable type must be used instead`
**Recommended Changes**
Change the implicit nullable type declaration to a nullable type declaration, available since PHP 7.1:
{{{#!diff
- function test( string $test = null ) {}
+ function test( ?string $test = null ) {}
}}}
This commit updates the affected instances in core to use a nullable type declaration.
References:
* [https://wiki.php.net/rfc/deprecate-implicitly-nullable-types PHP RFC: Deprecate implicitly nullable parameter types]
* [https://php.watch/versions/8.4/implicitly-marking-parameter-type-nullable-deprecated PHP.Watch: PHP 8.4: Implicitly nullable parameter declarations deprecated]
Follow-up to [28731], [50552], [57337], [57985].
Props ayeshrajans, jrf, audrasjb, jorbin.
Fixes#60786.
Built from https://develop.svn.wordpress.org/trunk@58009
git-svn-id: http://core.svn.wordpress.org/trunk@57480 1a063a9b-81f0-0310-95a4-ce76da25c4cd
This fixes the issue where negative values aren't allows in older versions of IE. There is more discussion in another ticket on a wider approach, but for now the decision to commit this was made as an interim solution.
Props mukesh27, SergeyBiryukov, sabernhardt, poena.
Fixes#46771.
Built from https://develop.svn.wordpress.org/trunk@58005
git-svn-id: http://core.svn.wordpress.org/trunk@57479 1a063a9b-81f0-0310-95a4-ce76da25c4cd
While technically only `null`, `WP_User`, or `WP_Error` should be returned from the `authenticate` filter, a plugin might return boolean `false` instead, which would trigger the `authentication_failed` error prior to [57882].
This commit aims to preserve that behavior in case `false` is returned.
Follow-up to [57882].
Props johnbillion.
See #60700.
Built from https://develop.svn.wordpress.org/trunk@57990
git-svn-id: http://core.svn.wordpress.org/trunk@57476 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Stop enqueueing polyfills such as `wp-polyfill-inert` (for the `inert` attribute) and `regenerator-runtime` (for generator functions), as they are no longer needed, considering the WordPress project's [https://make.wordpress.org/core/handbook/best-practices/browser-support/ browser support policy].
In addition to that, `wp-polyfill` (essentially `core-js`) is no longer enqueued as a dependency of `react`. This was added in [43903] to ensure compatibility with IE 11, which is no longer supported by WordPress. Developers requiring `wp-polyfill` need to manually add it as a dependency for their scripts.
Props swissspidy, flixos90, adamsilverstein, youknowriad, gziolo.
Fixes#60962.
Built from https://develop.svn.wordpress.org/trunk@57981
git-svn-id: http://core.svn.wordpress.org/trunk@57467 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Currently, in blocks themes it's possible to use the ${type}_template_hierarchy filter to alter the template hierarchy. However, those filters are not taken into consideration by the Choose a pattern popup screen that appears when creating a new template in the Site Editor, causing a mismatch between the editor and the frontend.
Props aljullu, mukesh27, ntsekouras, jorgefilipecosta, gziolo.
Fixes#60846.
Built from https://develop.svn.wordpress.org/trunk@57944
git-svn-id: http://core.svn.wordpress.org/trunk@57441 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Include the asset version of JavaScript and CSS files when generating the ETag for concatenated assets in `load-scripts.php` and `load-styles.php`. This ensures the ETag is updated as script versions change (for example editor package updates) rather than only when the WordPress version changes.
The `W\` prefix is added to the generated ETag to allow for CDNs and proxy servers modifying the script to add or improve the compression algorithm.
Props azaozz, dav4, ironprogrammer, johnbillion, kkmuffme, monzuralam, peterwilsoncc, sergeybiryukov.
Fixes#58433.
Built from https://develop.svn.wordpress.org/trunk@57943
git-svn-id: http://core.svn.wordpress.org/trunk@57440 1a063a9b-81f0-0310-95a4-ce76da25c4cd
The value of `start` is not fully supported by Opera Mini which has 1.01% usage. There is no material change in functionality with this change.
Follow-up to [55919].
Props davidbaumwald, sabernhardt, khokansardar, devsahadat.
Fixes#60876.
Built from https://develop.svn.wordpress.org/trunk@57933
git-svn-id: http://core.svn.wordpress.org/trunk@57434 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Due to some changes on the WP.com side to compress the requested images on the fly, the exact image size in the response could be different between platforms.
This commit aims to make the affected tests more reliable.
Follow-up to [139/tests], [31258], [34568], [47142], [57903], [57904], [57924].
Props peterwilsoncc, jorbin.
See #60865.
Built from https://develop.svn.wordpress.org/trunk@57931
git-svn-id: http://core.svn.wordpress.org/trunk@57432 1a063a9b-81f0-0310-95a4-ce76da25c4cd
This adds an `is_dir()` check in `_get_block_templates_paths` before trying to run a `RecursiveDirectoryIterator` to avoid errors being reported in New Relic even thought the errors should be handled by a try/catch block.
Follow-up to [57215].
Props iCaleb, sean212, mukesh27, joemcgill.
Fixes#60915.
Built from https://develop.svn.wordpress.org/trunk@57928
git-svn-id: http://core.svn.wordpress.org/trunk@57429 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Since WordPress 6.5, the minimum supported version of MySQL and MariaDB is 5.5.5. This means all supported database servers now support the `utf8mb4` character set and therefore the conditional logic for this is no longer necessary.
Props l1nuxjedi, craigfrancis, OllieJones, johnbillion
Fixes#60096
Built from https://develop.svn.wordpress.org/trunk@57926
git-svn-id: http://core.svn.wordpress.org/trunk@57427 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Some plugins pass invalid values such as `null` instead of a string, which has never been supported by WordPress (no translations are loaded) and was technically undefined behavior. With the introduction of the new l10n library in #59656, which has stricter type hints, this could end up causing warnings or even fatal errors.
This change adds a deliberate short-circuit to `load_textdomain()` & co. to better handle such a case and document that it is not supported.
Props verygoode, swissspidy.
Fixes#60888.
Built from https://develop.svn.wordpress.org/trunk@57925
git-svn-id: http://core.svn.wordpress.org/trunk@57426 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Follow-up to #60865 / [57903] where similar instances were recently updated due to an upstream change.
Particularly the `test_wp_crop_image_with_url()` test recently began to fail on some environments, likely because of the same change.
This updates the image URLs with the aim to bring more consistency and to get the test passing again more broadly.
Fixes#60907.
Built from https://develop.svn.wordpress.org/trunk@57924
git-svn-id: http://core.svn.wordpress.org/trunk@57425 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Previously, when `WP_PLUGIN_DIR` was set to something other than `wp-content/plugins`, e.g. `wp-content/mods`, `load_script_textdomain` was searching for script translations in `wp-content/languages/mods`. However, that is incorrect, as `WP_PLUGIN_DIR` does not affect where translations are stored. The location is always `wp-content/languages/plugins`.
Props coreymckrill, swissspidy.
Fixes#60891.
Built from https://develop.svn.wordpress.org/trunk@57922
git-svn-id: http://core.svn.wordpress.org/trunk@57423 1a063a9b-81f0-0310-95a4-ce76da25c4cd
An excessive amount of autoloaded options is a common cause for slow database responses, sometimes caused by very large individual autoloaded options. As it is not mandatory to provide an autoload value when adding an option to the database, it tends to be ignored, which in combination with a default value of "yes" and lack of documentation can lead to the aforementioned problem.
This changeset enhances the option autoloading behavior in several ways:
* Update the function documentation to encourage the use of boolean `true` or `false` to explicitly provide an autoload value for an option.
* Use new string values `on` and `off` for explicitly provided values stored in the database, to distinguish them from `yes` and `no`, since `yes` does not allow determining whether it was set intentionally by the developer or only as a default.
* Effectively deprecate the values `yes` and `no`. They are still supported for backward compatibility, but now discouraged.
* Use `null` as new default autoload value for `add_option()`. If the developer does not provide an explicit value, this will now trigger WordPress logic to determine an autoload value to use:
* If WordPress determines that the option should not be autoloaded, it is stored in the database as `auto-off`. As part of this changeset, the single heuristic introduced for that is to check whether the option size is larger than a threshold of 150k bytes. This threshold is filterable via a new `wp_max_autoloaded_option_size` filter.
* If WordPress determines that the option should be autoloaded, it is stored in the database as `auto-on`. No logic to make such a decision is introduced as part of this changeset, but a new filter `wp_default_autoload_value` can be used to define such heuristics, e.g. by optimization plugins.
* If WordPress cannot determine whether or not to autoload the option, it is stored in the database as `auto`.
* This effectively means that any option without an explicit autoload value provided by the developer will be stored with an autoload value of `auto`, unless the option's size exceeds the aforementioned threshold. Options with a value of `auto` are still autoloaded as of today, most importantly for backward compatibility. A new function `wp_autoload_values_to_autoload()` returns the list of autolaod values that dictate for an option to be autoloaded, and a new filter `wp_autoload_values_to_autoload` can be used to alter that list.
These behavioral changes encourage developers to be more mindful of autoloading, while providing WordPress core and optimization plugins with additional control over heuristics for autoloading options where no explicit autoload value was provided.
At the same time, the changes are fully backward compatible from a functionality perspective, with the only exception being that very large options will now no longer be autoloaded if the developer did not explicitly request for them to be autoloaded. Neither WordPress core nor plugins are able to override an explicitly provided value, which is intentional to continue giving developers full control over their own options.
Props pbearne, flixos90, joemcgill, azaozz, spacedmonkey, swissspidy, mukesh27, markjaquith.
Fixes#42441.
Built from https://develop.svn.wordpress.org/trunk@57920
git-svn-id: http://core.svn.wordpress.org/trunk@57421 1a063a9b-81f0-0310-95a4-ce76da25c4cd
The `$context` argument passed to filters such as `hooked_block_types`, `hooked_block`, and `hooked_block_{$hooked_block_type}` allows them to conditionally insert a hooked block. If the anchor block is contained in a template or template part, `$context` will be set to a `WP_Block_Template` object reflecting that template or part.
The aforementioned filters are applied when hooked block insertion is run upon reading a template (or part) from the DB (and before sending the template/part content with hooked blocks inserted over the REST API to the client), but also upon writing to the DB, as that's when the `ignoredHookedBlocks` metadata attribute is set.
Prior to this changeset, the `$context` passed to Block Hooks related filters in the latter case reflected the template/part that was already stored in the database (if any), which is a bug; instead, it needs to reflect the template/part that will result from the incoming `POST` network request that will trigger a database update.
Those incoming changes are encapsulated in the `$changes` argument passed to the `reset_pre_insert_template` and `reset_pre_insert_template_part` filters, respectively, and thus to the `inject_ignored_hooked_blocks_metadata_attributes` function that is hooked to them. `$changes` is of type `stdClass` and only contains the fields that need to be updated. That means that in order to create a `WP_Block_Template` object, a two-step process is needed:
- Emulate what the updated `wp_template` or `wp_template_part` post object in the database will look like by merging `$changes` on top of the existing `$post` object fetched from the DB, or from the theme's block template (part) file, if any.
- Create a `WP_Block_Template` from the resulting object.
To achieve the latter, a new helper method (`_build_block_template_object_from_post_object`) is extracted from the existing `_build_block_template_result_from_post` function. (The latter cannot be used directly as it includes a few database calls that will fail if no post object for the template has existed yet in the database.)
While somewhat complicated to implement, the overall change allows for better separation of concerns and isolation of entities. This is visible e.g. in the fact that `inject_ignored_hooked_blocks_metadata_attributes` no longer requires a `$request` argument, which is reflected by unit tests no longer needing to create a `$request` object to pass to it, thus decoupling the function from the templates endpoint controller.
Unit tests for `inject_ignored_hooked_blocks_metadata_attributes` have been moved to a new, separate file. Test coverage has been added such that now, all three relevant scenarios are covered:
- The template doesn't exist in the DB, nor is there a block theme template file for it.
- The template doesn't exist in the DB, but there is a block theme template file for it.
- The template already exists in the DB.
Those scenarios also correspond to the logical branching inside `WP_REST_Templates_Controller::prepare_item_for_database`, which is where `inject_ignored_hooked_blocks_metadata_attributes` gets its data from.
Props tomjcafferkey, bernhard-reiter, gziolo, swissspidy.
Fixes#60754.
Built from https://develop.svn.wordpress.org/trunk@57919
git-svn-id: http://core.svn.wordpress.org/trunk@57420 1a063a9b-81f0-0310-95a4-ce76da25c4cd
Compose V2, which was first released in 2020, is included with all currently supported versions of Docker Desktop. Compose V1 stopped receiving updates in July 2023.
The biggest difference is that the command to interact with Compose changed from `docker-compose` to `docker compose`.
GitHub has now started removing V1 from Ubuntu & Windows images, which caused all Docker-based GitHub Actions workflows to fail.
This change migrates to the new `docker compose` command to address these failures.
Props swissspidy, thelovekesh.
Fixes#60901.
Built from https://develop.svn.wordpress.org/trunk@57918
git-svn-id: http://core.svn.wordpress.org/trunk@57419 1a063a9b-81f0-0310-95a4-ce76da25c4cd
It appears that something has changed on the WP.com side to compress the requested images on the fly, which interfered with the previous expectations in these tests.
This commit uses a direct file URL and updates the expected image size to match the currently returned response.
Follow-up to [139/tests], [31258], [47142].
Props dextorlobo, swissspidy, davidbaumwald, SergeyBiryukov.
See #60865.
Built from https://develop.svn.wordpress.org/trunk@57903
git-svn-id: http://core.svn.wordpress.org/trunk@57404 1a063a9b-81f0-0310-95a4-ce76da25c4cd
dmsnell
Sergey Biryukov
Tammie Lister
Pascal Birchler
John Blackbourn
isabel_brison
jorgefilipecosta
Peter Wilson
Aaron Jorbin
Joe McGill
Felix Arntz
Bernhard Reiter
desrosj
audrasjb
davidbaumwald
Andrew Ozz