Hi Federico,

I think it is more about BEFORE than about TRIGGER:

DROP TABLE IF EXISTS t;

CREATE TABLE t (a INT DEFAULT 999) ENGINE = InnoDB;

DELIMITER ||

CREATE TRIGGER t_bi

	BEFORE INSERT

	ON t

	FOR EACH ROW

BEGIN

	SELECT DEFAULT(a) INTO @default_a_bi FROM t LIMIT 1;

END;

CREATE TRIGGER t_ai

	AFTER INSERT

	ON t

	FOR EACH ROW

BEGIN

	SELECT DEFAULT(a) INTO @default_a_ai FROM t LIMIT 1;

END;

||

DELIMITER ;

INSERT INTO t VALUES (0);

SELECT * FROM t;

SELECT @default_a_bi, @default_a_ai;

MariaDB [test]> SELECT @default_a_bi, @default_a_ai;

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

| @default_a_bi | @default_a_ai |

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

| NULL          |           999 |

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

1 row in set (0.00 sec)

So, it might have something to do with not having the information in a before trigger, although I haven't found anything about it in the documentation.
Since it's an upstream issue, as you know, normally we would refile it for MySQL and see what they say. However, this syntax has been removed in MySQL 5.6 and further, you cannot use there DEFAULT anymore.
It was apparently done by this change:

revno: 3858

revision-id: alexander.nozdrin@oracle.com-20120516123801-7wmajugqw48tq5jl

committer: Alexander Nozdrin <alexander.nozdrin@oracle.com>

timestamp: Wed 2012-05-16 16:38:01 +0400

message:

  WL#6230: Remove 'SET = DEFAULT'.

  This patch forbids assigning the DEFAULT keyword

  to stored program local variables or stored procedure

  parameters.

The worklog itself is not public, so it's hard to understand why it was done, maybe the problem you described was one of the reasons.

So, undoubtedly, MySQL won't do anything about it in 5.5.
I suppose eventually this change will make it to 10.x as well, so I'm not sure fixing it there, even if it's possible, is worth the trouble.

Do you agree, or do you want a further analysis?

Hi elenst, thank you for the useful information. I feel bad if a new limitation will be added to the stored programs, which already have many limitations. But in this particular case, DEFAULT already does not work at least in BEFORE INSERT, and after all it's just a cleaner way to do something I could do without the DEFAULT keyword. So, don't mind too much about this bug.

OK, thanks.
Then I will close it as "Won't fix" for now, if you later have more thoughts and want to re-open it, please comment.

I retried this to check if something changed - it didn't. However, shouldn't MariaDB return an error/warning to inform us that we are trying to do something that will not work?