Bug #39544 rpl.rpl_charset sporadically crashes the server in ps_stm / stmt mode on windows
Submitted: 19 Sep 2008 17:08 Modified: 14 Apr 2009 5:55
Reporter: Alexander Nozdrin Email Updates:
Status: Can't repeat Impact on me:
None 
Category:MySQL Server: Replication Severity:S2 (Serious)
Version:6.0-TRUNK OS:Windows
Assigned to: Zhenxing He CPU Architecture:Any
Tags: crash, pushbuild, sporadic, test failure

[19 Sep 2008 17:08] Alexander Nozdrin
Description:
Starting from Sep-16 rpl_charset failed twice in mysql-6.0.
The push on Sep-16 has nothing to do with the code base
(it was a change of .bzr-mysql/default.conf).

Probably something changed on the box.

Symptoms:
--------------------------------------------------------------
mysqltest: In included file ".\extra\rpl_tests\rpl_charset.test": At line 136: failed in 'select master_pos_wait('master-bin.000001', 2948)': 2013: Lost connection to MySQL server during query

The result from queries just before the failure was:
< snip >
@@character_set_server
latin5
select @@character_set_server;
@@character_set_server
latin5
set one_shot max_join_size=10;
ERROR HY000: The 'SET ONE_SHOT' syntax is reserved for purposes internal to the MySQL server
set character_set_client=9999999;
ERROR 42000: Unknown character set: '9999999'
set collation_server=9999998;
ERROR HY000: Unknown collation: '9999998'
use test;
CREATE TABLE t1 (c1 VARBINARY(255), c2 VARBINARY(255));
SET CHARACTER_SET_CLIENT=koi8r,
CHARACTER_SET_CONNECTION=cp1251, 
CHARACTER_SET_RESULTS=koi8r;
INSERT INTO t1 (c1, c2) VALUES ('��, �� �������','��, �� �������');
select hex(c1), hex(c2) from t1;
hex(c1)	hex(c2)
CDF32C20E7E020F0FBE1E0EBEAF3	CDF32C20E7E020F0FBE1E0EBEAF3
--------------------------------------------------------------

Occurred so far:
  - http://tinyurl.com/4oedus (Sep-16, vm-win2003-64-b)
  - http://tinyurl.com/44cejs (Sep-19, vm-win2003-64-b)

How to repeat:
http://tinyurl.com/4wl7ls
[29 Sep 2008 11:49] Sven Sandberg
Most likely a server crash, but there is no way to debug it since to stack trace is available.

There is a version of mtr in 5.1-rpl and 6.0-rpl which prints stack traces for core files it finds. Once that is merged to main, chances are that we can get a useful error message for this failure.

Setting to "To be fixed later" in the meantime.