MySQL Bugs: #40379: Server crash by "show slave status" (platform-specific)

Bug #40379	Server crash by "show slave status" (platform-specific)
Submitted:	28 Oct 2008 20:23	Modified:	14 Nov 2008 15:18
Reporter:	Joerg Bruehe	Email Updates:
Status:	Can't repeat	Impact on me:	None
Category:	MySQL Server: Replication	Severity:	S1 (Critical)
Version:	5.0.72	OS:	Linux (IA64 / ICC only)
Assigned to:	Sven Sandberg	CPU Architecture:	Any

Description:
This happened during the 5.0.72 build,
same symptom in three different tests,
all specific to Linux/IA64, on RedHat 5, compiled by ICC, "debug" build.

The oldest builds (in the archive) with this symptom were in 5.0.62,
sadly it seems that didn't get reported.

=====
rpl_bug33931                   [ fail ]

mysqltest: In included file "./include/wait_for_slave_to_stop.inc": At line 20: Error running query ' SHOW SLAVE STATUS': 2013 Lost connection to MySQL server during query

The result from queries just before the failure was:
reset master;
stop slave;
reset slave;
start slave;

More results from queries before failure can be found in /PATH/mysql-test/var/log/rpl_bug33931.log
=====

=====
rpl_change_master              [ fail ]

mysqltest: At line NNN: query 'show slave status' failed: 2013: Lost connection to MySQL server during query

The result from queries just before the failure was:
stop slave;
drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9;
reset master;
reset slave;
drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9;
start slave;
create table t1(n int);
select * from t1;
n
stop slave sql_thread;
insert into t1 values(1);
insert into t1 values(2);
stop slave;
show slave status;

More results from queries before failure can be found in /PATH/mysql-test/var/log/rpl_change_master.log
=====

=====
rpl_log                        [ fail ]

mysqltest: In included file "./include/wait_for_slave_to_stop.inc": At line 20: Error running query ' SHOW SLAVE STATUS': 2013 Lost connection to MySQL server during query

The result from queries just before the failure was:
stop slave;
drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9;
reset master;
reset slave;
drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9;
start slave;
stop slave;

More results from queries before failure can be found in /PATH/mysql-test/var/log/rpl_log.log
=====

How to repeat:
See description.

This is a debug server, but when compiling it the same flags as for the optimized one (which does not crash) were used.

Nonetheless, I try again and remove all optimizer flags when compiling the debug server.

See also BUG#40459.

In particular, note that "lost connection" during "show slave status" happens in the current 6.0-rpl tree on debx86-b for a specific test, namely rpl_ndb.rpl_ndb_denote_gap. That seems almost debuggable.

I followed Sinisa's suggestion and changed the RPM spec file to compile the debug server without optimization.
Running the tests again, all these crashes were gone.

So it seems the ICC compiler cannot handle optimization flags and the result of "--with-debug" when they occur simultaneously.

I have replaced the original RPMs by the newly built ones and will publish only those.

If anybody is interested to analyze this, or if it becomes relevant again in the future: The original (= faulty) RPMs will still be archived in a "RPMS-debug-optim/" directory.

After I discussed this with Joerg and Sven I will close this bug report now.