[MDEV-22047] binlog.binlog_mysqlbinlog2 Created: 2020-03-26  Updated: 2020-04-28  Resolved: 2020-04-28

Status: Closed
Project: MariaDB Server
Component/s: Platform Debian, Tests
Affects Version/s: 5.5.67
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Andrei Elkin Assignee: Alice Sherepa
Resolution: Won't Fix Votes: 0
Labels: sprint-week-18


 Description   

Starting from 407b0a6ae7c3 commit BB shows a massive result mismatch:

binlog.binlog_mysqlbinlog2 'mix'         w1 [ fail ]
 
        Test ended at 2020-03-24 19:09:10
 
 
 
CURRENT_TEST: binlog.binlog_mysqlbinlog2
 
--- /usr/share/mysql/mysql-test/suite/binlog/r/binlog_mysqlbinlog2.result	2019-01-29 21:39:35.000000000 +0300
 
+++ /dev/shm/var/1/log/binlog_mysqlbinlog2.reject	2020-03-25 02:09:10.359788889 +0300
 
@@ -272,11 +272,8 @@
 
 /*!50003 SET @OLD_COMPLETION_TYPE=@@COMPLETION_TYPE,COMPLETION_TYPE=0*/;
 
 DELIMITER /*!*/;
 
 ROLLBACK/*!*/;
 
-BEGIN
 
-/*!*/;
 
-SET INSERT_ID=3/*!*/;
 
 use `test`/*!*/;
 
-SET TIMESTAMP=1579609944/*!*/;
 
+SET TIMESTAMP=1579609942/*!*/;
 
 SET @@session.pseudo_thread_id=999999999/*!*/;
 
 SET @@session.foreign_key_checks=1, @@session.sql_auto_is_null=0, @@session.unique_checks=1, @@session.autocommit=1/*!*/;
 
 SET @@session.sql_mode=0/*!*/;
 
@@ -285,6 +282,30 @@
 
 SET @@session.character_set_client=8,@@session.collation_connection=8,@@session.collation_server=8/*!*/;
 
 SET @@session.lc_time_names=0/*!*/;
 
 SET @@session.collation_database=DEFAULT/*!*/;
 
+create table t1 (a int auto_increment not null primary key, b char(3))
 
+/*!*/;
 
+BEGIN
 
+/*!*/;
 
+SET INSERT_ID=1/*!*/;
 
+SET TIMESTAMP=1579609942/*!*/;
 
+insert into t1 values(null, "a")
 
+/*!*/;
 
+SET TIMESTAMP=1579609942/*!*/;
 
+COMMIT
 
+/*!*/;
 
+BEGIN



 Comments   
Comment by Sergei Golubchik [ 2020-03-26 ]

407b0a6ae7c3, really? it's "MDEV-10466 Server crashed in SEL_ARG::store_min() with extended_keys=on" and the only code change is

--- a/sql/opt_range.cc
 
+++ b/sql/opt_range.cc
 
@@ -8837,7 +8837,7 @@ key_and(RANGE_OPT_PARAM *param, SEL_ARG *key1, SEL_ARG *k>
 
       if (key2->next_key_part)
 
       {
 
        key1->use_count--;                      // Incremented in and_all_keys
 
-       return and_all_keys(param, key1, key2, clone_flag);
 
+        return and_all_keys(param, key1, key2->next_key_part, clone_flag);
 
       }
 
       key2->use_count--;                       // Key2 doesn't have a tree
 
     }

Comment by Andrei Elkin [ 2020-03-26 ]

I also checked the commit's patch which was provided by crossref search. The earliest failure occurred
in the following build:
http://buildbot.askmonty.org/buildbot/builders/kvm-deb-jessie-x86/builds/8510.
Naturally it's not to blame just to set up a margin for further searching.

Comment by Sergei Golubchik [ 2020-04-28 ]

This is happening only on 5.5 and only on jessie. If you take a closer look at the log, you'll see that it tests 10.0 not 5.5, and 10.0 from debian repositories. That's why tests fail, 10.0 is very old and its tests don't work in 2020.

Jessie used to be disabled for 5.5, precisely because jessie has 10.0 in its repos. But mid-March 2020 it someone has enabled it, apparently, by mistake. That's when it started failing and was always failing ever since.

There's no need to fix this, 5.5 has reached its EOL, so after this week's release there will be no more of 5.5. For the next few days simply pretend that jessie isn't there

Generated at Thu Feb 08 09:11:47 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.