@@warning_count gets reset before execution of SELECT

Note: It may well be expected behavior. However, as recent experience has shown, it is counter-intuitive, so if it works so by design, it should be explicitly documented, probably here.

create table t1 (a int);

create table if not exists t1 (b int);

select @@warning_count or (select count(*) from t1);

MariaDB [test]> create table t1 (a int);

Query OK, 0 rows affected (0.234 sec)

MariaDB [test]> create table if not exists t1 (b int);

Query OK, 0 rows affected, 1 warning (0.001 sec)

MariaDB [test]> select @@warning_count or (select count(*) from t1);

+----------------------------------------------+

| @@warning_count or (select count(*) from t1) |

+----------------------------------------------+

|                                            0 |

+----------------------------------------------+

1 row in set (0.003 sec)

The intuitive expectation here is that when the SELECT is executed, the value of @warning_count is still 1, and SELECT will return true without even executing the right part of OR. However, it doesn't happen, at the moment of execution the value of @@warning_count is already 0. Apparently, it gets reset before SELECT starts.

Also reproducible on active versions of MySQL, but equally unclear from MySQL documentation whether it's expected behavior.