MySQL Bugs: #17762: Connect fails in all tests of standard build

Bug #17762	Connect fails in all tests of standard build
Submitted:	28 Feb 2006 9:25	Modified:	13 Dec 2007 16:31
Reporter:	Joerg Bruehe	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: Compiling	Severity:	S3 (Non-critical)
Version:	5.1.7-beta + 5.0.19	OS:	Linux (Linux (ia64, icc) + SCO)
Assigned to:		CPU Architecture:	Any
Tags:	icc

Description:
Release build of 5.1.7-beta, based on ChangeSet
  1.2168 06/02/27 22:14:46 lars@mysql.com +1 -0
  BUG#17752: Non-allocated memory caused core in rpl_multi_engine test

Effect is specific to machine "rx2620b",
and does not occur in "debug" build:

=====
alias                          [ fail ]

Errors are (from /home/mysqldev/rx2620b-icc-glibc23/test/mysql-5.1.7-beta-linux-ia64-icc-glibc23/mysql-test/var/log/mysqltest-time) :
mysqltest: Could not open connection 'default': 1045 Access denied for user 'root'@'localhost' (using password: NO)
(the last lines may be the most important ones)

Ending Tests
Shutting-down MySQL daemon

Master(s) shutdown finished
Slave(s) shutdown finished
Resuming Tests
=====

Same problem occurs in all following tests, until the test suite timeout stops this.

How to repeat:
Detected by the test suite.

Same symptom on SCO OSR6 in the release build of 5.0.19:

=====
grant2                         [ fail ]

Errors are (from PATH/mysqltest-time) :
mysqltest: At line NNN: Could not open connection 'n5': 1045 Access denied for user 'test'@'localhost' (using password: NO)
(the last lines may be the most important ones)

Ending Tests
=====

Here, subsequent tests passed, it was a once-only incident.

July 17, 2006

Hello,

I have nearly the same thing.  I just downloaded and installed from mysql-5.0.22-win32.zip.

Here I intend to 
1) Show the command line for invoking mysqltest
2) Show the results
3) Make a comment and ask for help

I am using Windows XP Pro.
--------First the invocation command and the results:

C:\>"C:\Program Files\MySQL\MySQL Server 5.0\bin\mysqltest"
mysqltest: Could not open connection 'default': 1045 Access denied for user 'ODBC'@'localhost' (usin
g password: NO)
not ok
--------End invocation command and the results

Comment: I am not really familiar with mysql, but, where on earth did the mysqltest get the idea that there was a user named ODBC?  I know I did not create such a user.  I know the buzzword ODBC really means Open DataBase Connectivity but that is all I know about it.

Request for help: So,   So, what do I do about this ODBC thing?  Help!

By the way, I could not help but notice that the Severity has been given a rating of S3 (Non-critical).  Since I cannot get beyond this stage, I have to say, well, it may not be critical but I will still ask for some suggestions.

I checked with the Windows Control Panel.  Indeed, there is no such user as "ODBC".  In addition, I found a second test in the bin directory that seems to fail for the same reason.  The invocation line and results are once again pasted below.  Thanks in advance for your help.

C:\Program Files\MySQL\MySQL Server 5.0\bin>mysql_client_test.exe

#####################################
client_connect
#####################################

 Establishing a connection to '' ...
 [MySQL-5.0.22-community-nt][1045] Access denied for user 'ODBC'@'localhost' (using password: NO)

 Check the connection options using --help or -?

C:\Program Files\MySQL\MySQL Server 5.0\bin>

There is an even easier test that it fails.
---------paste of command invocation and the results
C:\>mysql
ERROR 1045 (28000): Access denied for user 'ODBC'@'localhost' (using password: NO)
---------end paste
As I said, there is no such user as ODBC, and about that two letter password "NO".  I haven't a clue where that comes from.  I really don't know what is going on here.

Please help.

Thanks

This still happens on our 5.1 builds, and we're getting closer and closer to RC. So could it please get assigned?

[msvensson@rx2620b my51-bug17762]$ icc test.cc && ./a.out 
str: 'localhost'  wildstr: 'localhost'
FAILED: res: 1
[msvensson@rx2620b my51-bug17762]$ icc -O0 test.cc && ./a.out 
str: 'localhost'  wildstr: 'localhost'
OK
[msvensson@rx2620b my51-bug17762]$ icc -O1 test.cc && ./a.out 
str: 'localhost'  wildstr: 'localhost'
FAILED: res: 1

There is two warnings produced by icc when compiling with -Wall but fixing them won't fix this. However changing the DBUG_RETURN macro
to
#define DBUG_RETURN(a1) return(a1)
will

Small test programs that fails when compiled with -O1 or greater

Attachment: bug17762.cc (text/x-c++src), 1.72 KiB.

I escalated the issue to Intel Compiler engineering. The bug is caused by an incorrect pointer incrementation in the context of a  Speculative Load for the second argument. 
For reference: Intel Bug Number: Q435671
I will update this case as soon as a fix is available

Closing because we're using icc 10 now, and the failure does not occur in more recent 5.1 builds anymore.