[tarantool-patches] Re: [PATCH 2/5] schema: add new system space for FK constraints

[Vladislav Shpilevoy <v.shpilevoy@tarantool.org>]

Hi!

On 06/08/2018 03:28, n.pettik wrote:
> Firstly, I don’t understand some of your fixes:
> 
> @@ -3855,11 +3853,9 @@ on_replace_dd_fk_constraint(struct trigger * /* trigger*/, void *event)
>                  memset(fkey, 0, sizeof(*fkey));
>                  fkey->def = fk_def;
>                  fkey->index_id = fk_index->def->iid;
> +               rlist_add_entry(&child_space->child_fkey, fkey, child_link);
> +               rlist_add_entry(&parent_space->parent_fkey, fkey, parent_link);
>                  if (old_tuple == NULL) {
> -                       rlist_add_entry(&child_space->child_fkey, fkey,
> -                                       child_link);
> -                       rlist_add_entry(&parent_space->parent_fkey, fkey,
> -                                       parent_link);
>                          struct trigger *on_rollback =
>                                  txn_alter_trigger_new(on_create_fkey_rollback,
>                                                        fkey);
> @@ -3868,10 +3864,6 @@ on_replace_dd_fk_constraint(struct trigger * /* trigger*/, void *event)
>                          struct fkey *old_fk =
>                                  fkey_grab_by_name(&child_space->child_fkey,
>                                                    fk_def->name);
> -                       rlist_add_entry(&child_space->child_fkey, fkey,
> -                                       child_link);
> -                       rlist_add_entry(&parent_space->parent_fkey, fkey,
> -                                       parent_link);
> 
> In case of replace we must firstly remove entry and only then insert new one.
> It is easy to check that the way you suggest doesn’t work (just modify test you provided):

Yes, your are right. This was a buggy hunk. Thanks for the fix!