galera.galera_query_cache_invalidate fails: Note 4190 missing after RESET SLAVE ALL

galera.galera_query_cache_invalidate fails on the main (13.2) branch: the output lacks the Note 4190 lines that r/galera_query_cache_invalidate.result expects after RESET SLAVE ALL.

Build Run Info

Failure text

Latest full-text run on main (worker number, timestamps and paths differ between runs):

galera.galera_query_cache_invalidate     w11 [ fail ]

        Test ended at 2026-11-02 03:59:48

CURRENT_TEST: galera.galera_query_cache_invalidate

--- /home/buildbot/s390x-ubuntu-2204/build/mysql-test/suite/galera/r/galera_query_cache_invalidate.result	2026-11-01 22:17:11.000000000 +0000

+++ /home/buildbot/s390x-ubuntu-2204/build/mysql-test/suite/galera/r/galera_query_cache_invalidate.reject	2026-11-02 03:59:47.921050129 +0000

@@ -109,8 +109,6 @@

 connection node_3;

 STOP SLAVE;

 RESET SLAVE ALL;

-Warnings:

-Note	4190	RESET SLAVE is implicitly changing the value of 'Using_Gtid' from 'Current_Pos' to 'Slave_Pos'

 connection node_1;

 SET SESSION WSREP_ON=OFF;

 include/reset_master.inc

Result length mismatch

 - saving '/home/buildbot/s390x-ubuntu-2204/build/mysql-test/var/11/log/galera.galera_query_cache_invalidate/' to '/home/buildbot/s390x-ubuntu-2204/build/mysql-test/var/log/galera.galera_query_cache_invalidate/'

Retrying test galera.galera_query_cache_invalidate, attempt(2/3)...

Analysis

The test and result files are identical on 12.3, 13.1 and main. The test sets master_use_gtid=current_pos on node_3 (line 32) and runs RESET SLAVE ALL (line 113); result lines 112 and 113 expect Warnings: and Note 4190 RESET SLAVE is implicitly changing the value of 'Using_Gtid' from 'Current_Pos' to 'Slave_Pos' after it. Commit c7032278d40b (Remove C++ STL machinery from sql/rpl_info_file. and CHANGE MASTER parser; author date 2026-05-29, committed to main on 2026-10-16) removes the WARN_OPTION_CHANGING Note from reset_slave() in sql/sql_repl.cc, so main no longer raises it, while 12.3 and 13.1 still do (their heads do not contain the commit). The same commit removed the Note from the expected output of rpl_galera_to_mariadb_clone_slave_using_mariadb-backup.result, but not from this result file. The first failure on main is at commit 327437ade1d6 (2026-10-16), two commits above it, with none on main in the 15 days before, and the staging branch bb-cleanup-monty fails from 2026-10-15; the first and the latest failing build on main both contain the commit. Between 2026-10-01 and 2026-11-02 the failure was recorded 568 times (59 on main). Eight builders report it (aarch64-debian-12, aarch64-debian-13, amd64-freebsd-14, ppc64le-ubuntu-2204, s390x-ubuntu-2204, amd64-openssl3-fips, amd64-debian-12-deb-autobake-migration and amd64-rhel-9-rpm-autobake-migration); the last three store only the tail of the output, and their build logs (builds 15715, 15330 and 15701 at main 5d26b3be259d) show the same two missing lines. The same diff on four CPU architectures points at the result file, not at a builder. galera_3nodes.galera_2_cluster and galera_3nodes.galera_gtid_2_cluster fail on main with the same missing Note. MDEV-31300 (Open) is an older failure of this test on 11.0 with the opposite diff, extra warnings in the output. The expected output of this test and of the two siblings needs the same update that rpl_galera_to_mariadb_clone_slave_using_mariadb-backup.result got.

Created via the mtr-failure-create-bug skill.