Define missing object properties on non-testing classes for PHP 8.2

1. +++ b/core/lib/Drupal/Component/Diff/Engine/DiffEngine.php
   @@ -31,6 +31,16 @@ class DiffEngine {
      const MAX_XREF_LENGTH = 10000;
    
   +  private array $xchanged;
   +  private array $ychanged;
   +  private array $xv;
   +  private array $yv;
   

   so #3274474: Fix 'Access to an undefined property' PHPStan L0 errors took the easy way out and dropped them from that issue, but we still need to fix it.

   needs a core maintainer decision but I expect we still need minimal docs on this one (or also just put the dynamic property annotation on it and make it a problem for another day?)

2. +++ b/core/lib/Drupal/Core/Config/Entity/ConfigEntityType.php
   @@ -47,6 +47,10 @@ class ConfigEntityType extends EntityType implements ConfigEntityTypeInterface {
      protected $mergedConfigExport = [];
   +  protected $label_callback;
   +  protected $controllers;
   +  protected $module;
   +  protected $list_path;
    
      /**
   

   The fix IMHO is still to add the AllowedDynamicProperties annotation to \Drupal\Component\Plugin\Definition\PluginDefinition. plugin definitions are imho expected to be extensible. just adding these might fix the core tests for now, but it will not fix contrib. We can have a follow-up to use magic get/set or something, but that should be ok for now.

3. +++ b/core/lib/Drupal/Core/Session/UserSession.php
   @@ -66,6 +66,23 @@ class UserSession implements AccountInterface {
      protected $timezone;
   +  protected $langcode;
   +  protected $pass;
   +  protected $status;
   +  protected $created;
   +  protected $changed;
   +  protected $login;
   +  protected $init;
   

   this is an absolute mess. we need to clean this up, but instead of adding all these properties, lets also use the annotation on this class.

4. +++ b/core/modules/block/src/Entity/Block.php
   @@ -137,6 +137,7 @@ class Block extends ConfigEntityBase implements BlockInterface, EntityWithPlugin
       */
      protected $theme;
   +  protected $provider;
   

   this is a config entity class, so this can be removed.

5. +++ b/core/modules/field/tests/modules/field_test/src/Plugin/Field/FieldWidget/TestFieldWidget.php
   @@ -23,6 +23,11 @@
    
   +  /**
   +   * \Drupal\Tests\field_ui\Kernel\EntityFormDisplayTest.
   +   */
   +  public $randomValue;
   +
   

   ok, so that's this part:

       $random_value = $this->randomString();
       $widget->randomValue = $random_value;
       $widget = $form_display->getRenderer($field_name);
       $this->assertEquals($random_value, $widget->randomValue);
   

   Isn't there a better way to check that it's the same object? According to https://www.php.net/manual/en/language.oop5.object-comparison.php, we can simply do a $renderer === $renderer2 comparison, then we can drop this. same for the widget above.

6. +++ b/core/modules/migrate/src/Plugin/Migration.php
   @@ -327,6 +327,19 @@ class Migration extends PluginBase implements MigrationInterface, RequirementsIn
   +  protected $class;
   +  // phpcs:disable Drupal.Classes.PropertyDeclaration
   +  protected $_discovered_file_path;
   +  protected $deriver;
   +  protected $target_types;
   +  protected $field_plugin_method;
   +
   +  /**
   +   * Drupal\Tests\migrate\Kernel\MigrateSkipRowTest.
   +   */
   +  protected $load;
   +  public $provider;
   

   I think that load thing in MigrateSkipRowTest is a bogus leftover. the test still passes if you comment it out, lets just remove that there.

   the others we either need to document properly or clean up. What I'm not 100% clear about is the different between the plugin definition and the plugin class here. Ah, I see. \Drupal\migrate\Plugin\Migration::__construct() is the culprit, it moves everything from configuration and plugin definition directly to the class. We either need to limit that to existing properties or add the annotation as well, because this could contain anything then.

7. +++ b/core/modules/migrate_drupal/src/Plugin/migrate/FieldMigration.php
   @@ -21,6 +21,7 @@ class FieldMigration extends Migration implements ContainerFactoryPluginInterfac
      protected $init = FALSE;
   +  protected $field_plugin_method;
    
      /**
       * The migration field discovery service.
   

   same here, this is already on the parent class now.

8. +++ b/core/modules/system/src/Entity/Action.php
   @@ -79,6 +79,12 @@ class Action extends ConfigEntityBase implements ActionConfigEntityInterface, En
      protected $pluginCollection;
   +  protected $url;
   +  protected $submit;
   +  protected $form_build_id;
   +  protected $form_token;
   +  protected $form_id;
   +  protected $confirm;
    
   

   config entity, safe to remove.

9. +++ b/core/modules/system/tests/modules/entity_test/src/TypedData/ComputedString.php
   @@ -11,6 +11,11 @@
    class ComputedString extends TypedData implements CacheableDependencyInterface {
    
   +  /**
   +   * Drupal\Tests\jsonapi\Kernel\Normalizer\FieldItemNormalizerTest.
   +   */
   +  protected $value;
   +
   

   this is kind of a weird one.

   the bug is arguably in \Drupal\Core\TypedData\TypedData::setValue, this sets $this->value but doesn't actually define such a property. but that's more or less because the type there isn't defined.

   as a computed string, I don't think this expects to have a value set, but all properties get set to a value through \Drupal\Core\TypedData\Plugin\DataType\Map::setValue(), even if nothing was set for them.

   I think this for now is correct, lets just set a generic docblock like on \Drupal\Core\TypedData\PrimitiveBase.

10. +++ b/core/modules/user/src/Entity/User.php
    @@ -74,6 +74,8 @@ class User extends ContentEntityBase implements UserInterface {
        */
       protected static $anonymousUser;
    +  // phpcs:disable Drupal.Classes.PropertyDeclaration
    +  public $_skipProtectedUserFieldConstraint;
     
    

    this _should_ go through magic get/set, do tests really fail without this?

> I think this for now is correct, lets just set a generic docblock like on \Drupal\Core\TypedData\PrimitiveBase.

Alternatively, we could overwrite setValue() and ignore the input, as we do not support setting a value.

Re #4.1 the way out was meant to be #3299678: Deprecate DiffEngine and replace with sebastian/diff, but I reckon that is too late for 9.5. IMHO better go for the dynamic property annotation for now, then directionally work to remove DiffEngine.

@andypost dunno… it was like that in #3274474: Fix 'Access to an undefined property' PHPStan L0 errors before being expuncted. I guess we need guidance here.

edit: commented on the wrong issue.

+++ b/core/modules/views/src/ManyToOneHelper.php
@@ -21,6 +21,15 @@
+  /**
+   * Should the field to use formula or alias.
+   *
+   * @see \Drupal\views\Plugin\views\argument\StringArgument::query()
+   *

I think the "to" here is not necessary and not valid english, can be removed.

@Berdir,
Made changes as per comment #19.
Needs review.

Thanks.

+++ b/core/modules/views/src/ManyToOneHelper.php
@@ -21,6 +21,15 @@
+  public $formula = FALSE;

Given we added handler with a typehint I think we should be adding this with one too...

so
public bool $formula = FALSE;

Discussed with a @catch - we agreed to make a commit to add the typehint and be done.

Committed and pushed 4a16f2a856 to 10.1.x and 19bd9092bd to 10.0.x. Thanks!

--- a/core/modules/views/src/ManyToOneHelper.php
+++ b/core/modules/views/src/ManyToOneHelper.php

@@ -38,6 +38,15 @@ public function buildOptionsForm(&$form, FormStateInterface $form_state) {
+  /**
+   * Should the field use formula or alias.
+   *
+   * @see \Drupal\views\Plugin\views\argument\StringArgument::query()
+   *
+   * @var bool
+   */
+  public $formula = FALSE;

If we're going to backport this then I think we should backport [##3295157] too

This should add the declaration before the constructor, as in 10.1.x.

Adding declaration above constructor.

I agree with #30 and I'm at the same time against backporting #3295157: Fix 'Access to an undefined property' PHPStan L0 errors - public properties, I was wrong with an earlier slack thread, I believe *that* issue is what broke search_api in D10 and backporting would break it there too. So, IMHO, we should close this and leave PHP 8.2 support for D10+.