Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.11, 11.4, 11.8
-
None
Description
binlog.binlog_parallel_replication_marks_stm_mix fails in buildbot on release branches 10.11, 11.4 and 11.8, with a mysqltest result content mismatch in the replicated binlog event stream.
Build Run Info
| Branch | Builder name | Date | Commit | Variant |
|---|---|---|---|---|
| 10.11 | amd64-msan-clang-22 | 2026-09-21T13:58:29Z | bdff5b8f2455acd226d242e16e8c2107f741d3da | mix |
| 11.4 | amd64-msan-clang-22 | 2026-09-21T17:13:43Z | 6c951fea5f62e1eb0ed7483ad8aa2cdcbfff5dd1 | mix |
| 11.8 | amd64-msan-clang-22 | 2026-09-18T19:03:53Z | ccf21ce23870f8122ee4cb1f184453b893797a78 | mix |
Failure text
The buildbot failure cross-reference captures two context hunks per occurrence (no +/- diff markers preserved), identical across all matched occurrences:
--- hunk 1 ---
|
# server id 1 end_log_pos # CRC32 0x######## Write_rows: table id # flags: STMT_END_F
|
/* GTID */ UPDATE t1 SET c=1 WHERE a=7
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (8, 5, 0)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (9, 5, 1)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (10, 6, 0)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (11, 7, 0)
|
COMMIT/*!*/;
|
DROP TABLE t1;
|
 |
--- hunk 2 ---
|
/* GTID */ INSERT INTO t1 VALUES (8, 5, 0)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# ddl thread_id=#
|
/* GTID */ ALTER TABLE t4 ADD b INT
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (9, 5, 1)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# trans thread_id=#
|
START TRANSACTION
|
/* GTID */ INSERT INTO t1 VALUES (10, 6, 0)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# ddl thread_id=#
|
START TRANSACTION
|
/* GTID */ CREATE TEMPORARY TABLE t5 (a INT PRIMARY KEY) ENGINE=InnoDB
|
/* GTID */ INSERT INTO t1 VALUES (11, 7, 0)
|
COMMIT/*!*/;
|
# server id 1 end_log_pos # CRC32 0x######## GTID #-#-# ddl thread_id=#
|
DROP TEMPORARY TABLE IF EXISTS `t5` /* generated by server */
|
DROP TABLE t1;
|
Analysis
All 5 matching occurrences in the window (across 10.11, 11.4 and 11.8) occur on a single builder, amd64-msan-clang-22, pointing at this specific MSAN test/build path rather than a broad per-branch regression. The mismatch is around the presence of an ALTER TABLE t4 ADD b INT statement and a CREATE TEMPORARY TABLE t5 / DROP TEMPORARY TABLE t5 pair between the two captured hunks; the cross-reference data does not preserve +/- diff markers, so which side is expected vs actual could not be determined from this data alone - the full .reject file from the buildbot build log is needed to confirm.
A related historical ticket, MDEV-13248 (Closed/Fixed, 2017), covered a "Result length mismatch" on this exact test (and its sibling binlog_parallel_replication_marks_row), but that root cause was a timestamp string leaking into mysqlbinlog output and is unrelated to today's content difference - not a duplicate.
Created via the mtr-failure-bug skill.
Attachments
Issue Links
- relates to
-
MDEV-40415 Address replication MTR test failures
-
- Open
-