Conflict with Login Tobaggon

It's actually not a conflict with Logintoboggan specifically, it is a conflict with user_save and the regsitration_role module not properly setting a new role.

user_save(http://api.drupal.org/api/function/user_save/5) calls the insert hook expecting modules that want to alter a user insert to use the $edit array to set new variables, after the hook is called is when the roles are updated, and if there exists any new roles in the $edit array, all old roles are dropped, and since registration_role is using a direct SQL statement to alter the users roles the user_save is pretty much immediately deleting that change. Any module that alters user roles will break registration_role.

Attached is a fix to use the $edit array to properly set the role at registration.

status update

That hardcodes the role ID as role 1. Also has anyone tested this approach?

Thanks to a patch supplied by abhaga and christefano in issue http://drupal.org/node/320087 this was in fact fixed long ago, but i failed to create a release. New release now, 5.x-1.3. My apologies.

^{benjamin, Agaric Design Collective}

> That hardcodes the role ID as role 1. Also has anyone tested this approach?

No, the "1" is a truth(boolean) value, not an RID, print_r() a user object and look at the format of the roles array. The value of [$rid] makes no difference, I put up the link to the api documetation for reference, all that matters is that the $user[roles] array exists and has a list of keys that represent the RIDs you want to add/modify.

    // Save user roles (delete just to be safe).
    if (is_array($array['roles'])) {
      db_query('DELETE FROM {users_roles} WHERE uid = %d', $array['uid']);
      foreach (array_keys($array['roles']) as $rid) {
        if (!in_array($rid, array(DRUPAL_ANONYMOUS_RID, DRUPAL_AUTHENTICATED_RID))) {
          db_query('INSERT INTO {users_roles} (uid, rid) VALUES (%d, %d)', $array['uid'], $rid);
        }
      }
    }

The other patch does the job just fine except for instantiating an empty array for no reason, but I don't see a use-case where the difference would be noticeable.

Yup, my bad– as noted, this was fixed in head a good while ago, and a new release has been made now.