Opened 10 years ago
Last modified 8 years ago
#36851 new defect (bug)
Widgets don't remove hooks after being unregistered
| Reported by: |
|
Owned by: | |
|---|---|---|---|
| Milestone: | Future Release | Priority: | normal |
| Severity: | normal | Version: | 2.8 |
| Component: | Widgets | Keywords: | needs-patch |
| Focuses: | Cc: |
Description
In WP_Widget_Recent_Comments::__construct(), there is this bit of code:
<?php if ( is_active_widget( false, false, $this->id_base ) || is_customize_preview() ) { add_action( 'wp_head', array( $this, 'recent_comments_style' ) ); }
If unregister_widget( 'WP_Widget_Recent_Comments' ) is called, this added wp_head action is still going to persist unexpectedly. At the moment, I believe the only way to remedy this inside such widgets themselves would be to check to see if the widget is still among $wp_widget_factory->widgets when the action callback is called (here the recent_comments_style method). From outside the widget, the alternative is would be to do:
<?php $widget = $wp_widget_factory->widgets['WP_Widget_Recent_Comments']; unregister_widget( get_class( $widget ) ); remove_action( 'wp_head', array( $widget, 'recent_comments_style' ) );
Neither of these options are great.
Perhaps there should be new widget_registered and widget_unregistered actions that widgets could listen to to do this cleanup? Or there could be a new unregister method on WP_Widget that a subclass could have this logic inside of. (We wouldn't be able to use the PHP destructor since it would never be called since the reference to the class would be still captured among the registered hooks.) Likewise, instead of adding the hooks inside of the constructor, perhaps there should also be a WP_Widget::register() method that gets called inside of the faux-private WP_Widget::_register() (and _register should be final, no?)
@westonruter I like the idea of the
unregistermethod coupled with ado_action( 'widget_unregistered' )whenever a widget is unregistered.Adding a
WP_Wiget:: register()method is also a great idea, however, I don't think we can assume that hooks added in a subclass will be there moving ahead. Perhaps it's something that we add and then make recommendations to use that instead of in the constructor in the docs.