Bug #69470 Use "group by", but no aggregate function, get weird result

Description:
In group by statement, if we query some column but without aggregate function, we should get the first value in the group.

But in 5.6.10, sometimes we just get null.

How to repeat:
CREATE TABLE test.self_ref_table
(id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
ref_id INT NOT NULL
)ENGINE = INNODB DEFAULT CHARSET utf8;

Use this statement to insert 50 records.(May be we need more to repeat this problem)
INSERT INTO test.self_ref_table(ref_id) VALUES (ROUND(RAND() * 50));

SELECT
a.id, a.ref_id, b.id
FROM
test.self_ref_table a
LEFT JOIN
test.self_ref_table b
ON a.ref_id = b.id
GROUP BY a.id

I can get one record with null value of "b.id", which should be not null.

Verified as described. Possibly a duplicate of BUG#68897

CREATE TABLE self_ref_table
(id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
ref_id INT NOT NULL
)ENGINE = INNODB DEFAULT CHARSET utf8;
INSERT INTO self_ref_table(ref_id) VALUES (ROUND(RAND() * 50));
INSERT INTO self_ref_table(ref_id) VALUES (ROUND(RAND() * 50));
...

SELECT a.id, a.ref_id, b.id
FROM self_ref_table a LEFT JOIN self_ref_table b ON a.ref_id = b.id
GROUP BY a.id;
id      ref_id  id
1       8       8
...
47      45      45                 <- there obviously is a row with id=47
48      47      NULL               <- but here we seemingly get no join match

Confirmed duplicate of BUG#68897