Details
Major Description
I've noticed several times failures of the main.cte_nonrecursive test. They usually don't repeat when restarting the test, so they seem sporadic.
Example:
main.cte_nonrecursive w3 [ fail ]
|
Test ended at 2021-11-27 23:03:54
|
|
|
CURRENT_TEST: main.cte_nonrecursive
|
mysqltest: At line 1220: query 'drop database test' failed: ER_DB_DROP_RMDIR (1010): Error dropping database (can't rmdir './test', errno: 39 "Directory not empty")
|
|
|
The result from queries just before the failure was:
|
< snip >
|
# MDEV-18460: Server crashed in strmake / tdc_create_key /
|
# THD::create_tmp_table_def_key
|
#
|
connect con1,localhost,root,,;
|
CREATE TEMPORARY TABLE test.t (a INT);
|
WITH cte AS (SELECT 1) SELECT * FROM cte;
|
1
|
1
|
WITH t AS (SELECT 1) SELECT * FROM t;
|
1
|
1
|
WITH cte AS (SELECT 1) SELECT * FROM t;
|
ERROR 3D000: No database selected
|
DROP TABLE test.t;
|
connection default;
|
disconnect con1;
|
#
|
# MDEV-22781: create view with CTE without default database
|
#
|
drop database test;
|
|
|
More results from queries before failure can be found in /tmp/tmp.SRpydvPo6F/var/3/log/cte_nonrecursive.log
|
|
|
- saving '/tmp/tmp.SRpydvPo6F/var/3/log/main.cte_nonrecursive/' to '/tmp/tmp.SRpydvPo6F/var/log/main.cte_nonrecursive/'
|
Occurences:
- Salsa-CI mariadb-10.6 on amd64: https://salsa.debian.org/mariadb-team/mariadb-server/-/jobs/2225231/raw
- Debian build mariadb-10.6 on armhf: https://buildd.debian.org/status/fetch.php?pkg=mariadb-10.6&arch=armhf&ver=1%3A10.6.5-1&stamp=1638053889&raw=0
I've seen this failure on MariaDB 10.4 a couple of times too, so I don't think it is specific for 10.6.
Attachments
Activity
otto, would the following change to the test fix the failure?
diff --git a/mysql-test/main/cte_nonrecursive.result b/mysql-test/main/cte_nonrecursive.result
|
index 17c73f74a49..3d56754b2a8 100644
|
--- a/mysql-test/main/cte_nonrecursive.result
|
+++ b/mysql-test/main/cte_nonrecursive.result
|
@@ -1693,7 +1693,9 @@ disconnect con1;
|
#
|
# MDEV-22781: create view with CTE without default database
|
#
|
-drop database test;
|
+create database db;
|
+use db;
|
+drop database db;
|
create database db1;
|
create table db1.t1 (a int);
|
insert into db1.t1 values (3),(7),(1);
|
@@ -1723,7 +1725,6 @@ a
|
drop view db1.v1;
|
drop table db1.t1;
|
drop database db1;
|
-create database test;
|
use test;
|
#
|
# MDEV-24597: CTE with union used multiple times in query
|
diff --git a/mysql-test/main/cte_nonrecursive.test b/mysql-test/main/cte_nonrecursive.test
|
index 8ab3bddc410..82afcdb9b8e 100644
|
--- a/mysql-test/main/cte_nonrecursive.test
|
+++ b/mysql-test/main/cte_nonrecursive.test
|
@@ -1217,7 +1217,9 @@ DROP TABLE test.t;
|
--echo # MDEV-22781: create view with CTE without default database
|
--echo #
|
|
-drop database test;
|
+create database db;
|
+use db;
|
+drop database db;
|
create database db1;
|
create table db1.t1 (a int);
|
insert into db1.t1 values (3),(7),(1);
|
@@ -1239,7 +1241,6 @@ drop view db1.v1;
|
drop table db1.t1;
|
drop database db1;
|
|
-create database test;
|
use test;
|
|
--echo # |
I pushed the test case cleanup to 10.3 and merged to all later-vesion branches.
otto, I see that the test is executing several DROP DATABASE. The problematic one in your report is
I remember that monty has stated earlier (related to versioning tests) that the test database should never be dropped or re-created. That is, this failure could be worked around by slightly changing the test.
It would be helpful to know which files existed in the directory when it was not empty, so that we would know which earlier test might have left some garbage files behind in that directory. In 10.6, the culprit should not be InnoDB (
MDEV-21283).I made a quick attempt based on https://salsa.debian.org/mariadb-team/mariadb-server/-/jobs/2225231/raw as follows:
Unfortunately, the test main.cte_nonrecursive always passed on my system. I attempted it 3 times.