Bug #12693 join test failure

• Submitted: 19 Aug 2005 22:59
• Modified: 31 Aug 2005 9:29
• Reporter: Jim Winstead
• Status: Can't repeat
• Category: MySQL Server: Optimizer
• Severity: S2 (Serious)
• Version: 5.0-bk
• OS: Linux (Linux (x86_64))
• Assigned to: Sergey Petrunya
• CPU Architecture: Any

[19 Aug 2005 22:59] Jim Winstead

Description:
Then join test fails with a build done using BUILD/compile-amd64-debug-max. Others have reported the same or similar failure on different configurations.

The test passes when the --ps-protocol flag is used.

How to repeat:
Below are the diffs between actual and expected results:
-------------------------------------------------------
*** r/join.result       2005-08-18 02:21:01.000000000 +0300
--- r/join.reject       2005-08-20 01:59:38.273711952 +0300
***************
*** 338,345 ****
  i     i
  2     1
  3     1
- 2     2
  3     2
  select t1.i,t2.i,t3.i from t2 left join t3 on (t2.i=t3.i),t1 order by t1.i,t2.i,t3.i;
  i     i       i
  1     2       2
--- 338,345 ----
  i     i
  2     1
  3     1
  3     2
+ 2     2
  select t1.i,t2.i,t3.i from t2 left join t3 on (t2.i=t3.i),t1 order by t1.i,t2.i,t3.i;
  i     i       i
  1     2       2
***************
*** 362,369 ****
  i     i
  2     1
  4     1
- 2     2
  4     2
  select t1.i,t2.i,t3.i from t2 right join t3 on (t2.i=t3.i),t1 order by t1.i,t2.i,t3.i;
  i     i       i
  1     NULL    4
--- 362,369 ----
  i     i
  2     1
  4     1
  4     2
+ 2     2
  select t1.i,t2.i,t3.i from t2 right join t3 on (t2.i=t3.i),t1 order by t1.i,t2.i,t3.i;
  i     i       i
  1     NULL    4
-------------------------------------------------------

[31 Aug 2005 9:29] Sergey Petrunya

Can't repeat on hammer, 5.0 tree with last cset
ChangeSet@1.1925, 2005-08-31 10:25:00+05:00, [e-mail]
  Merge mysql.com:/usr/home/ram/work/4.1.b9613
  into  mysql.com:/usr/home/ram/work/mysql-5.0, 
tree compiled with BUILD/compile-amd64-debug-max.
Can't repeat on x86 with compile-pentium-debug either.