Hi, Sergey! Just one comment: On Oct 18, Sergey Vojtovich wrote:
diff --git a/mysql-test/r/lock_tables_lost_commit.result b/mysql-test/r/lock_tables_lost_commit.result index 769e973..394ef0a 100644 --- a/mysql-test/r/lock_tables_lost_commit.result +++ b/mysql-test/r/lock_tables_lost_commit.result @@ -9,7 +9,6 @@ disconnect con1; connection con2; SELECT * FROM t1; a -10
This test doesn't make sense after your fix. and the original bug#578 reappears. The problem is that mysqlimport issues LOCK TABLES, but does not issue UNLOCK TABLES at the end. I suspect this has to be fixed now.
Alternatively, you can make LOCK TABLE to commit (not rollback) a transaction at disconnect. But that would be weird, wouldn't it? This change is in new patch too.
You're right, we should fix mysqlimport. I just wonder why this trivial fix wasn't introduced before?
Will you do it as a part of MDEV-7660? Better not.
Could you make an MDEV for that? Type=Bug, FixVersion=10.2. Otherwise all ok, I think. Regards, Sergei Chief Architect MariaDB and security@mariadb.org