[MDEV-22263] main.func_debug fails on a valgrind build with wrong result Created: 2020-04-16  Updated: 2020-04-20  Resolved: 2020-04-20

Status: Closed
Project: MariaDB Server
Component/s: Tests
Affects Version/s: 10.3, 10.4
Fix Version/s: 10.3.23, 10.4.13

Type: Bug Priority: Major
Reporter: Elena Stepanova Assignee: Anel Husakovic
Resolution: Fixed Votes: 0
Labels: regression


 Description   

The failure below only happens for me on a Valgrind build (-DCMAKE_BUILD_TYPE=Debug -DWITH_VALGRIND=YES), not on a plain debug or ASAN builds.
On a Valgrind build, it currently happens deterministically, at least on two different machines with different gcc versions.
The test doesn't have to be run with --valgrind, it happens with and without the option alike; it's only the build that matters.

10.3 6577a7a8 Valgrind

$ perl ./mtr main.func_debug --valgrind
 
main.func_debug                          [ fail ]
        Test ended at 2020-04-16 21:47:01
 
CURRENT_TEST: main.func_debug
--- /data/bld/10.3-valgrind-nightly/mysql-test/main/func_debug.result	2020-04-16 06:58:36.000000000 +0300
+++ /data/bld/10.3-valgrind-nightly/mysql-test/main/func_debug.reject	2020-04-16 21:47:01.119451507 +0300
@@ -1671,6 +1671,6 @@
 2
 9
 Warnings:
-Note	1105	DBUG: Item_subselect::exec (select max(`test`.`t1`.`a`) from `test`.`t...
+Note	1105	DBUG: Item_subselect::exec (select max(`test`.`t1`.`a`) from `test`.`t1`)
 DROP TABLE t1;
 SET SESSION debug_dbug="-d,Item_subselect";
 
mysqltest: Result content mismatch

Reproducible on 10.3 and on 10.4 after the recent merge, so it must have been introduced recently. I didn't bisect for the exact commit.
10.2 doesn't have the test.



 Comments   
Comment by Anel Husakovic [ 2020-04-16 ]

Hi, here is a patch sanja.

Comment by Oleksandr Byelkin [ 2020-04-17 ]

Ok to push!

Thanks for the fix!

Comment by Anel Husakovic [ 2020-04-20 ]

Pushed to 10.3 with 42af2b1d8b8a85.
Thanks.

Generated at Thu Feb 08 09:13:26 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.