Bug #18980 Test 'rpl_row_blob_innodb' fails randomly
Submitted: 10 Apr 2006 19:04 Modified: 4 May 2007 14:01
Reporter: Kent Boortz Email Updates:
Status: Can't repeat Impact on me:
Category:MySQL Server: Row Based Replication ( RBR ) Severity:S3 (Non-critical)
Version:5.1-bk OS:Linux (SuSE SLES9 x86)
Assigned to: Chuck Bell CPU Architecture:Any

[10 Apr 2006 19:04] Kent Boortz
Test 'rpl_row_blob_innodb' fails randomly with

 mysqltest: In included file "./extra/rpl_tests/rpl_row_blob.test":
 At line 42: could not sync with master
 ('select master_pos_wait('master-bin.000001', 18169)' returned NULL)

How to repeat:
Run the test case
[5 Jul 2006 1:25] Greg Lehey
Run the test on Fedora Core 5/i386 in a loop for an estimated 3000 times without failure.  It looks as if this one might have gone away or gone into hiding.  It doesn't sound very likely that it's related to the specific Linux distro.
[5 Jul 2006 1:39] Greg Lehey
Test was run with:

 ./mysql-test-run --mysqld="--binlog-format=row" --skip-ndb rpl_row_blob_innodb
[3 Aug 2006 7:43] Greg Lehey
Repeated on sles9-x86, where the test failed immediately.  This is an assertion failure:

mysqld: field.cc:3565: virtual longlong Field_long::val_int(): Assertion `!table || (!table->read_set || bitmap_is_set(table->read_set, field_index))' failed.
[7 Nov 2006 18:52] Chuck Bell
Initially, cannot reproduce errors on SUSE or Windows using clone from mysql-5.1-new-rpl. A new error occurred on line #182 during the diff of the output of the master versus the slave.
[7 Nov 2006 18:57] Chuck Bell
The new error from the test is:

Errors are (from d:/source/c++/mysql-5.1-new-rpl/mysql-test/var/log/mysqltest-time) :
mysqltest: In included file ".\extra\rpl_tests\rpl_row_blob.test": At line 182:
command "diff $MYSQLTEST_VARDIR/tmp/rpl_row_blob_master.sql $MYSQLTEST_VARDIR/tmp/rpl_row_blob_slave.sql" failed
[21 Nov 2006 14:09] Chuck Bell
I have added test files for reproducing the error with the rpl_row_blob_innodb test. See attached files.
[21 Nov 2006 14:10] Chuck Bell
Correct results for test. Place in ./mysql-test/r.

Attachment: rpl_bug_18980.result (text/plain), 3.32 KiB.

[21 Nov 2006 14:11] Chuck Bell
Test include file (contains actual test). Place in ./mysql-test/extra/rpl_tests.

Attachment: rpl_row_blob_bug_18980.test (application/octet-stream, text), 3.54 KiB.

[21 Nov 2006 14:11] Chuck Bell
Option file (required). Place in ./mysql-test/t.

Attachment: rpl_bug_18980-slave.opt (text/plain), 9 bytes.

[21 Nov 2006 14:12] Chuck Bell
Test file. Place in ./mysql-test/t. Run with --mysqld=--binlog-format=row

Attachment: rpl_bug_18980.test (application/octet-stream, text), 478 bytes.

[21 Nov 2006 14:14] Chuck Bell
I have tested this bug usng the attached test files on Windows XP, Windows 2003 64-bit, and SUSE 10.0. Please verify bug exists on 64-bit Linux platforms. The result should be a failed test with incorrect output on the slave.
[22 Nov 2006 18:00] Chuck Bell
This bug is fixed in BUG#24403. Please refer to BUG#24403 for more details.
[27 Nov 2006 18:11] Paul DuBois
Test case change. No changelog entry needed.
[20 Mar 2007 12:59] Chuck Bell
Per Jani Tolonen's email on 20 Mar 2007. Reevaluation of bug shows the problem has not been corrected. Bug report reopened.
[4 May 2007 14:01] Chuck Bell
This bug was tested with the latest clone from mysql-5.1-new-rpl. The bug described was not reproduced. Platforms tested include Windows XP Pro, Windows 2003 Server, SUSE 10.0, and Fedora Core 4. 

A patch was pushed that enabled the test in the disabled.def file.
[1 Jun 2007 19:22] Bugs System
Pushed into 5.1.20-beta