Just noticed that while this function *does* clean up table content_access correctly, it fails to clean up the acl schema properly ...

For whatever reason (I don't have time to investigate it) the statement
if (module_exists('acl')) fails to return TRUE... I don't think that it's really necessary to have this test in the first place but I've only given it a quick look. I believe it's superfluous because the module dependencies would have been called out long before this point ... however, commenting out this module test causes the right behavior: 

/**
 * Deletes all custom per node settings, so that content type defaults are used again.
 */
function content_access_delete_per_node_settings($node) {
  db_query("DELETE FROM {content_access} WHERE nid = %d", $node->nid);
  // Clear the cache.
  content_access_per_node_setting(NULL, $node, FALSE);
  // Delete possible acl settings
  // if (module_exists('acl')) {   <=== Comment this line out
    foreach (array('view', 'update', 'delete') as $op) {
      $acl_id = content_access_get_acl_id($node, $op);
      acl_delete_acl($acl_id);
    }
  // }                            <=== Comment this line out
}

Comments

sjz’s picture

sjz commented
Status: Active » Closed (fixed)

Took more time to understand this - Ignore (part of) the base message ... I took more time to look into the interaction and where there are remnants in the acl schema after the "reset to defaults" takes place.

The mechanism here:

function content_access_page_reset($form, &$form_state) {
  content_access_delete_per_node_settings($form_state['node']);
  node_access_acquire_grants($form_state['node']);
  cache_clear_all();
  drupal_set_message(t('The permissions have been reseted to the content type defaults.'));
}

... shows the reason for the artifacts left in the acl schema.

node_access_acquire_grants causes re-insertion of the information in the database. It's unnecessary but harmless. I still suggest that Content Access should clean up completely when resetting to defaults.

I'll still stand by the notion that calling module_exists('acl') is superfluous in the original post - but it *is*, indeed, working properly.